From e83462842093fef6e6eb295fca39beaacc1948c2 Mon Sep 17 00:00:00 2001 From: Juan Sugg Date: Sat, 28 Mar 2026 19:51:06 -0300 Subject: [PATCH] Update .md files - cosmetics --- QUICKSTART.md | 86 ++++-------- README.md | 122 ++++-------------- SETUP_GUIDE.md | 106 ++++----------- USAGE_GUIDE.md | 61 +++------ docs/testing/README.md | 19 ++- docs/testing/authoring.md | 44 ++----- docs/testing/operations.md | 21 ++- platform/docs/BACKUP_AND_RECOVERY.md | 7 +- platform/docs/CI_AND_SECURITY.md | 94 ++++---------- platform/docs/CONTEXT_SERVICE.md | 55 ++------ platform/docs/DEVFLOW.md | 3 +- platform/docs/HOST_PLATFORMS.md | 47 +++---- platform/docs/HYPERMEMORY.md | 28 +--- platform/docs/MODEL_ROUTING.md | 3 +- platform/docs/POLICY_ENGINE_AND_WRAPPERS.md | 21 +-- platform/docs/SECRETS_AND_ENV.md | 10 +- platform/docs/SECURITY_MODEL.md | 3 +- .../linux-runtime-user-and-systemd.md | 3 +- .../runbooks/macos-service-user-and-ssh.md | 6 +- platform/plugins/memory-lancedb-pro/README.md | 77 ++++------- .../plugins/memory-lancedb-pro/README_CN.md | 77 ++++------- .../plugins/memory-lancedb-pro/README_DE.md | 77 ++++------- .../plugins/memory-lancedb-pro/README_ES.md | 77 ++++------- .../plugins/memory-lancedb-pro/README_FR.md | 77 ++++------- .../plugins/memory-lancedb-pro/README_IT.md | 77 ++++------- .../plugins/memory-lancedb-pro/README_JA.md | 77 ++++------- .../plugins/memory-lancedb-pro/README_KO.md | 77 ++++------- .../memory-lancedb-pro/README_PT-BR.md | 77 ++++------- .../plugins/memory-lancedb-pro/README_RU.md | 77 ++++------- .../plugins/memory-lancedb-pro/README_TW.md | 77 ++++------- .../memory-lancedb-pro/STRONGCLAW_VENDOR.md | 6 +- .../docs/memory_architecture_analysis.md | 3 +- .../hook/enqueue-lesson-extract/HOOK.md | 9 +- .../memory-lancedb-pro/skills/lesson/SKILL.md | 5 +- .../skills/local/browser-lab-safety/SKILL.md | 3 +- platform/skills/quarantine/README.md | 3 +- platform/workers/acpx/README.md | 8 +- platform/workers/browser-lab/README.md | 3 +- platform/workspace/admin/IDENTITY.md | 10 +- platform/workspace/reader/IDENTITY.md | 10 +- .../platform/docs/BACKUP_AND_RECOVERY.md | 7 +- .../assets/platform/docs/CI_AND_SECURITY.md | 94 ++++---------- .../assets/platform/docs/CONTEXT_SERVICE.md | 55 ++------ src/clawops/assets/platform/docs/DEVFLOW.md | 3 +- .../assets/platform/docs/HOST_PLATFORMS.md | 47 +++---- .../assets/platform/docs/HYPERMEMORY.md | 28 +--- .../assets/platform/docs/MODEL_ROUTING.md | 3 +- .../docs/POLICY_ENGINE_AND_WRAPPERS.md | 21 +-- .../assets/platform/docs/SECRETS_AND_ENV.md | 10 +- .../assets/platform/docs/SECURITY_MODEL.md | 3 +- .../linux-runtime-user-and-systemd.md | 3 +- .../runbooks/macos-service-user-and-ssh.md | 6 +- .../plugins/memory-lancedb-pro/README.md | 77 ++++------- .../plugins/memory-lancedb-pro/README_CN.md | 77 ++++------- .../plugins/memory-lancedb-pro/README_DE.md | 77 ++++------- .../plugins/memory-lancedb-pro/README_ES.md | 77 ++++------- .../plugins/memory-lancedb-pro/README_FR.md | 77 ++++------- .../plugins/memory-lancedb-pro/README_IT.md | 77 ++++------- .../plugins/memory-lancedb-pro/README_JA.md | 77 ++++------- .../plugins/memory-lancedb-pro/README_KO.md | 77 ++++------- .../memory-lancedb-pro/README_PT-BR.md | 77 ++++------- .../plugins/memory-lancedb-pro/README_RU.md | 77 ++++------- .../plugins/memory-lancedb-pro/README_TW.md | 77 ++++------- .../memory-lancedb-pro/STRONGCLAW_VENDOR.md | 6 +- .../docs/memory_architecture_analysis.md | 3 +- .../hook/enqueue-lesson-extract/HOOK.md | 9 +- .../memory-lancedb-pro/skills/lesson/SKILL.md | 5 +- .../skills/local/browser-lab-safety/SKILL.md | 3 +- .../platform/skills/quarantine/README.md | 3 +- .../assets/platform/workers/acpx/README.md | 8 +- .../platform/workers/browser-lab/README.md | 3 +- .../platform/workspace/admin/IDENTITY.md | 10 +- .../platform/workspace/reader/IDENTITY.md | 10 +- tests/fixtures/README.md | 20 +-- 74 files changed, 818 insertions(+), 2035 deletions(-) diff --git a/QUICKSTART.md b/QUICKSTART.md index 9923c0ac..a2995c11 100644 --- a/QUICKSTART.md +++ b/QUICKSTART.md @@ -2,9 +2,7 @@ This quickstart gets you from zero to a verified secure baseline. -For host-native runtime-user provisioning and service-manager activation, use -[`platform/docs/HOST_PLATFORMS.md`](platform/docs/HOST_PLATFORMS.md) alongside -this guide. +For host-native runtime-user provisioning and service-manager activation, use [`platform/docs/HOST_PLATFORMS.md`](platform/docs/HOST_PLATFORMS.md) alongside this guide. ## 1. Install the runtime package @@ -12,18 +10,11 @@ this guide. make install ``` -`make install` prefers Python `3.12` for the managed environment on supported -Darwin/Linux hosts so the default hypermemory rerank stack stays on the most -compatible interpreter path. Python `3.13` remains supported when you choose it -explicitly. +`make install` prefers Python `3.12` for the managed environment on supported Darwin/Linux hosts so the default hypermemory rerank stack stays on the most compatible interpreter path. Python `3.13` remains supported when you choose it explicitly. -If you plan to develop on this repo, install `uv` and use `make dev` plus -`make test` separately. The companion-tool test entrypoints run through -`uv run`, and the bootstrap flow installs `uv` when the host does not -already provide it. +If you plan to develop on this repo, install `uv` and use `make dev` plus `make test` separately. The companion-tool test entrypoints run through `uv run`, and the bootstrap flow installs `uv` when the host does not already provide it. -If you want shorter commands in an interactive shell, sync the dev environment -once and activate `.venv` before running tools directly: +If you want shorter commands in an interactive shell, sync the dev environment once and activate `.venv` before running tools directly: ```bash uv sync --locked @@ -34,13 +25,7 @@ deactivate ## 2. Prepare the Varlock env contract -You can prepare the env contract either manually or through the guided setup -flow. `clawops setup` and `clawops varlock-env configure` create the managed -Varlock env under the StrongClaw config dir, repair missing keys, generate -required local secrets, and prompt for missing runtime or provider-auth input -when needed. On Linux that default path is usually -`~/.config/strongclaw/varlock`; on macOS it is usually -`~/Library/Application Support/StrongClaw/config/varlock`. +You can prepare the env contract either manually or through the guided setup flow. `clawops setup` and `clawops varlock-env configure` create the managed Varlock env under the StrongClaw config dir, repair missing keys, generate required local secrets, and prompt for missing runtime or provider-auth input when needed. On Linux that default path is usually `~/.config/strongclaw/varlock`; on macOS it is usually `~/Library/Application Support/StrongClaw/config/varlock`. Manual path: @@ -49,8 +34,7 @@ clawops varlock-env configure --non-interactive $EDITOR ~/.config/strongclaw/varlock/.env.local ``` -Before you continue, decide how OpenClaw should authenticate to an LLM provider. -StrongClaw supports two setup paths: +Before you continue, decide how OpenClaw should authenticate to an LLM provider. StrongClaw supports two setup paths: - guided/OpenClaw-managed: `make setup`, `uv run --project . clawops setup`, or `clawops setup` can launch `openclaw configure --section model` - env-driven: set provider keys plus optional model overrides in the managed `.env.local` @@ -59,25 +43,17 @@ StrongClaw supports two setup paths: - for local models, set `OLLAMA_API_KEY=ollama-local` and `OPENCLAW_OLLAMA_MODEL=` - a fully local dev baseline can use `OPENCLAW_OLLAMA_MODEL=llama3:latest` -StrongClaw now defaults to `hypermemory`, so set -`HYPERMEMORY_EMBEDDING_MODEL=` before you run the -no-arg setup path. The hypermemory setup path uses loopback defaults for -`HYPERMEMORY_EMBEDDING_BASE_URL` and `HYPERMEMORY_QDRANT_URL` unless you override -them. +StrongClaw now defaults to `hypermemory`, so set `HYPERMEMORY_EMBEDDING_MODEL=` before you run the no-arg setup path. The hypermemory setup path uses loopback defaults for `HYPERMEMORY_EMBEDDING_BASE_URL` and `HYPERMEMORY_QDRANT_URL` unless you override them. -For a fully local dev stack, `HYPERMEMORY_EMBEDDING_MODEL=ollama/nomic-embed-text` -plus `HYPERMEMORY_EMBEDDING_API_BASE=http://host.docker.internal:11434` is a -working local baseline when that Ollama model is already pulled. +For a fully local dev stack, `HYPERMEMORY_EMBEDDING_MODEL=ollama/nomic-embed-text` plus `HYPERMEMORY_EMBEDDING_API_BASE=http://host.docker.internal:11434` is a working local baseline when that Ollama model is already pulled. -If you want the built-in OpenClaw path instead, use the explicit -`openclaw-default` profile: +If you want the built-in OpenClaw path instead, use the explicit `openclaw-default` profile: ```bash clawops config memory --set-profile openclaw-default ``` -If you want the built-ins plus the experimental QMD backend, use -`openclaw-qmd`: +If you want the built-ins plus the experimental QMD backend, use `openclaw-qmd`: ```bash clawops config memory --set-profile openclaw-qmd @@ -132,13 +108,9 @@ That setup flow: - prepares sidecar config and service manifests - runs the baseline verification gate -StrongClaw-generated runtime data does not default into the repository checkout. -Setup now places compose state, harness artifacts, ACP summaries, the managed -`lossless-claw` checkout, and QMD package files under OS-appropriate app -data/state directories instead. +StrongClaw-generated runtime data does not default into the repository checkout. Setup now places compose state, harness artifacts, ACP summaries, the managed `lossless-claw` checkout, and QMD package files under OS-appropriate app data/state directories instead. -If you intentionally want repo-local compose state during development, keep it -explicit instead of relying on stale container mounts: +If you intentionally want repo-local compose state during development, keep it explicit instead of relying on stale container mounts: ```bash clawops ops sidecars up --repo-local-state @@ -147,8 +119,7 @@ clawops ops prune-qdrant-test-collections clawops ops reset-compose-state --component qdrant ``` -You can rerun the host doctor directly after any local change that might affect -the rendered config or CLI toolchain: +You can rerun the host doctor directly after any local change that might affect the rendered config or CLI toolchain: ```bash clawops doctor-host @@ -160,10 +131,7 @@ For the full post-bootstrap readiness sweep, run: clawops doctor ``` -If Linux bootstrap just added the runtime user to the `docker` group, setup -pauses with clear remediation. Open a fresh login shell as that user and rerun -the same `make setup` or `clawops setup` command; completed bootstrap work is -auto-detected and skipped. +If Linux bootstrap just added the runtime user to the `docker` group, setup pauses with clear remediation. Open a fresh login shell as that user and rerun the same `make setup` or `clawops setup` command; completed bootstrap work is auto-detected and skipped. ## 4. Rerender the OpenClaw config when you change profiles @@ -171,8 +139,7 @@ auto-detected and skipped. clawops render-openclaw-config ``` -This now renders the default StrongClaw profile, `hypermemory`, -and writes the result to `~/.openclaw/openclaw.json`. +This now renders the default StrongClaw profile, `hypermemory`, and writes the result to `~/.openclaw/openclaw.json`. For the explicit built-in OpenClaw path, render `openclaw-default`, which merges: @@ -183,8 +150,7 @@ For the explicit OpenClaw + QMD path, render `openclaw-qmd`, which adds: - a rendered form of `platform/configs/openclaw/40-qmd-context.json5` -For placeholder-backed variants, rerender by profile instead of merging raw -JSON5 overlays: +For placeholder-backed variants, rerender by profile instead of merging raw JSON5 overlays: ```bash clawops render-openclaw-config --profile openclaw-default @@ -202,11 +168,7 @@ The `openclaw-qmd` profile enables QMD-backed memory retrieval and indexes: - the managed StrongClaw workspace Markdown tree - the managed upstream checkout when that directory exists -The default `hypermemory` profile enables the combined -`lossless-claw` + `strongclaw-hypermemory` runtime, points the plugin at -the rendered runtime config under `~/.config/strongclaw/memory/hypermemory.yaml`, -enables `autoRecall`, keeps -`autoReflect` disabled, and does not inherit the QMD overlay. +The default `hypermemory` profile enables the combined `lossless-claw` + `strongclaw-hypermemory` runtime, points the plugin at the rendered runtime config under `~/.config/strongclaw/memory/hypermemory.yaml`, enables `autoRecall`, keeps `autoReflect` disabled, and does not inherit the QMD overlay. ## 5. Verify the baseline again on demand @@ -235,15 +197,13 @@ clawops devflow resume --run-id --approved-by operator clawops devflow audit --run-id ``` -See [`platform/docs/DEVFLOW.md`](platform/docs/DEVFLOW.md) for the operator run -layout, recovery flow, and audit bundle details. +See [`platform/docs/DEVFLOW.md`](platform/docs/DEVFLOW.md) for the operator run layout, recovery flow, and audit bundle details. - `clawops verify-platform sidecars --skip-runtime` - `clawops verify-platform observability --skip-runtime` - `clawops verify-platform channels` - companion-tool smoke tests -For the deeper StrongClaw readiness scan, including model/provider validation -and platform verification in one place, run: +For the deeper StrongClaw readiness scan, including model/provider validation and platform verification in one place, run: ```bash make doctor @@ -257,13 +217,13 @@ Add these only in order: 2. Repo codebase context: `clawops context codebase index --scale small --config platform/configs/context/codebase.yaml --repo .` 3. QMD prewarm: `qmd status` 4. Built-in OpenClaw memory fallback: - `clawops setup --profile openclaw-default` +`clawops setup --profile openclaw-default` 5. Built-in OpenClaw plus experimental QMD: - `clawops setup --profile openclaw-qmd` +`clawops setup --profile openclaw-qmd` 6. Opt-in local LanceDB durable memory with Ollama-backed smart extraction by rerendering - `clawops render-openclaw-config --profile memory-lancedb-pro` +`clawops render-openclaw-config --profile memory-lancedb-pro` 7. Migration-only standalone overlay reference: - `platform/configs/openclaw/75-strongclaw-hypermemory.example.json5` +`platform/configs/openclaw/75-strongclaw-hypermemory.example.json5` 8. Telegram: `platform/docs/channels/telegram.md` 9. WhatsApp: `platform/docs/channels/whatsapp.md` 10. OTel/Langfuse: `clawops verify-platform observability` diff --git a/README.md b/README.md index 250b259b..a4489e1b 100644 --- a/README.md +++ b/README.md @@ -1,8 +1,6 @@ # Strongclaw / ClawOps -This repository is the **Strongclaw** bootstrap for a hardened, -production-oriented OpenClaw deployment and ships the **ClawOps** companion -tooling. It is intentionally broader than a simple install pack. It includes: +This repository is the **Strongclaw** bootstrap for a hardened, production-oriented OpenClaw deployment and ships the **ClawOps** companion tooling. It is intentionally broader than a simple install pack. It includes: - a hardened OpenClaw control plane baseline - a separate execution plane for ACP/acpx coding workers @@ -68,27 +66,11 @@ make install make setup ``` -`make install` and the managed bootstrap path prefer Python `3.12` on supported -Darwin/Linux hosts as the compatibility baseline. That keeps the default -hypermemory local-rerank path installable on Intel macOS while remaining valid -across Apple Silicon Macs, Linux laptops, and supported 64-bit Raspberry Pi -setups. Python `3.13` remains supported for explicit use and CI coverage. - -`make setup` runs the guided `clawops setup` workflow inside the managed -environment. It bootstraps host prerequisites, creates or repairs -the managed Varlock env under the StrongClaw config dir, offers local or -managed Varlock secret backends for provider auth, prompts for missing setup -input when needed, configures OpenClaw model/provider auth, activates -services, and runs the baseline verification gate. The lower-level CLI -entrypoint remains available at `clawops setup` for manual or partial -bring-up, and you can call the CLI directly with `uv run --project . clawops setup`. -For a render-only pass that does not activate services yet, use -`clawops setup --no-activate-services`; that path now defers model/provider -auth until you are ready to start the gateway. - -The wheel now ships the runtime `platform` asset bundle, so package-safe -commands such as `clawops render-openclaw-config`, `clawops setup`, and -`clawops verify-platform ...` work outside a cloned StrongClaw checkout. +`make install` and the managed bootstrap path prefer Python `3.12` on supported Darwin/Linux hosts as the compatibility baseline. That keeps the default hypermemory local-rerank path installable on Intel macOS while remaining valid across Apple Silicon Macs, Linux laptops, and supported 64-bit Raspberry Pi setups. Python `3.13` remains supported for explicit use and CI coverage. + +`make setup` runs the guided `clawops setup` workflow inside the managed environment. It bootstraps host prerequisites, creates or repairs the managed Varlock env under the StrongClaw config dir, offers local or managed Varlock secret backends for provider auth, prompts for missing setup input when needed, configures OpenClaw model/provider auth, activates services, and runs the baseline verification gate. The lower-level CLI entrypoint remains available at `clawops setup` for manual or partial bring-up, and you can call the CLI directly with `uv run --project . clawops setup`. For a render-only pass that does not activate services yet, use `clawops setup --no-activate-services`; that path now defers model/provider auth until you are ready to start the gateway. + +The wheel now ships the runtime `platform` asset bundle, so package-safe commands such as `clawops render-openclaw-config`, `clawops setup`, and `clawops verify-platform ...` work outside a cloned StrongClaw checkout. Boundary override flags are now literal: @@ -97,9 +79,7 @@ Boundary override flags are now literal: - use `--source-root` for source-tree-only verification such as `clawops baseline` - use `--repo-root` only for repo-contract tooling such as `clawops repo`, `clawops worktree`, and `clawops supply-chain` -By default, StrongClaw now renders and provisions the -`hypermemory` stack. Set one embedding model name before you run -the no-arg setup path: +By default, StrongClaw now renders and provisions the `hypermemory` stack. Set one embedding model name before you run the no-arg setup path: ```bash export HYPERMEMORY_EMBEDDING_MODEL=openai/text-embedding-3-small @@ -107,58 +87,31 @@ uv run --project . clawops setup clawops doctor ``` -The shipped hypermemory configs now also enable planner-stage reranking with a -local `sentence-transformers` provider first and a `compatible-http` fallback. -Plain `uv sync` installs the local rerank dependency on the host/Python -combinations where upstream wheels are known to work: macOS arm64, macOS -x86_64 on Python 3.12, and Linux x86_64 or aarch64/arm64 on Python 3.12 or -3.13. For Raspberry Pi, that explicitly includes Raspberry Pi 4/5 running -64-bit Raspberry Pi OS or Ubuntu arm64. On unsupported combinations such as -macOS x86_64 with Python 3.13 or 32-bit Raspberry Pi Linux, StrongClaw skips -the local dependency and relies on `compatible-http` or fail-open behavior -instead of breaking installation. -On Intel macOS/Python 3.12 that compatibility path is pinned to -`sentence-transformers==3.4.1`, `torch==2.2.2`, and `numpy<2`. The shipped -config defaults `rerank.local.device` to `auto`, which selects `cuda` on -supported GPU hosts, `mps` on supported Apple Silicon hosts, and `cpu` -otherwise. If auto-selected acceleration fails at runtime, the local reranker -falls back to CPU automatically. - -If you want the HTTP fallback available, set `HYPERMEMORY_RERANK_BASE_URL` and, -optionally, `HYPERMEMORY_RERANK_MODEL` / `HYPERMEMORY_RERANK_API_KEY`. - -If you want the legacy OpenClaw built-ins instead, use the explicit -`openclaw-default` profile: +The shipped hypermemory configs now also enable planner-stage reranking with a local `sentence-transformers` provider first and a `compatible-http` fallback. Plain `uv sync` installs the local rerank dependency on the host/Python combinations where upstream wheels are known to work: macOS arm64, macOS x86_64 on Python 3.12, and Linux x86_64 or aarch64/arm64 on Python 3.12 or 3.13. For Raspberry Pi, that explicitly includes Raspberry Pi 4/5 running 64-bit Raspberry Pi OS or Ubuntu arm64. On unsupported combinations such as macOS x86_64 with Python 3.13 or 32-bit Raspberry Pi Linux, StrongClaw skips the local dependency and relies on `compatible-http` or fail-open behavior instead of breaking installation. On Intel macOS/Python 3.12 that compatibility path is pinned to `sentence-transformers==3.4.1`, `torch==2.2.2`, and `numpy<2`. The shipped config defaults `rerank.local.device` to `auto`, which selects `cuda` on supported GPU hosts, `mps` on supported Apple Silicon hosts, and `cpu` otherwise. If auto-selected acceleration fails at runtime, the local reranker falls back to CPU automatically. + +If you want the HTTP fallback available, set `HYPERMEMORY_RERANK_BASE_URL` and, optionally, `HYPERMEMORY_RERANK_MODEL` / `HYPERMEMORY_RERANK_API_KEY`. + +If you want the legacy OpenClaw built-ins instead, use the explicit `openclaw-default` profile: ```bash uv run --project . clawops config memory --set-profile openclaw-default ``` -If you want the built-ins plus the experimental QMD backend, use -`openclaw-qmd`: +If you want the built-ins plus the experimental QMD backend, use `openclaw-qmd`: ```bash uv run --project . clawops config memory --set-profile openclaw-qmd ``` -StrongClaw-generated runtime artifacts no longer default into the git checkout. -Harness output, ACP summaries, compose sidecar state, QMD package files, and -the managed `lossless-claw` checkout are written to OS-appropriate app -data/state directories instead. +StrongClaw-generated runtime artifacts no longer default into the git checkout. Harness output, ACP summaries, compose sidecar state, QMD package files, and the managed `lossless-claw` checkout are written to OS-appropriate app data/state directories instead. -The bootstrap path reuses any existing Docker-compatible runtime that already -provides `docker` plus `docker compose`, and only installs Docker as a fallback -when no compatible runtime is detected. +The bootstrap path reuses any existing Docker-compatible runtime that already provides `docker` plus `docker compose`, and only installs Docker as a fallback when no compatible runtime is detected. Then continue with ACP workers, repo lexical context indexing, channels, and observability using the step order in [`SETUP_GUIDE.md`](SETUP_GUIDE.md). -If Linux bootstrap just added the runtime user to the `docker` group, setup now -pauses with a clear message. Open a fresh login shell as that user and rerun -the same `make setup` or `clawops setup` command; completed bootstrap work is -detected automatically. +If Linux bootstrap just added the runtime user to the `docker` group, setup now pauses with a clear message. Open a fresh login shell as that user and rerun the same `make setup` or `clawops setup` command; completed bootstrap work is detected automatically. -For placeholder-backed variants, rerender through the profile-aware entrypoint -instead of merging raw overlays: +For placeholder-backed variants, rerender through the profile-aware entrypoint instead of merging raw overlays: ```bash clawops render-openclaw-config --profile acp @@ -166,13 +119,9 @@ clawops render-openclaw-config --profile hypermemory clawops render-openclaw-config --profile memory-lancedb-pro ``` -The StrongClaw-managed `memory-lancedb-pro` profile keeps smart extraction on, -but still ships with `autoRecall = false`, `sessionStrategy = "none"`, -`selfImprovement.enabled = false`, and `enableManagementTools = false`. +The StrongClaw-managed `memory-lancedb-pro` profile keeps smart extraction on, but still ships with `autoRecall = false`, `sessionStrategy = "none"`, `selfImprovement.enabled = false`, and `enableManagementTools = false`. -The bootstrap verification flow keeps the `clawops verify-platform` entrypoints -on the operator path: sidecars can be probed directly, while the baseline gate -re-runs the sidecar, observability, and channel checks in static mode. +The bootstrap verification flow keeps the `clawops verify-platform` entrypoints on the operator path: sidecars can be probed directly, while the baseline gate re-runs the sidecar, observability, and channel checks in static mode. For a deeper post-setup scan, run: @@ -180,16 +129,11 @@ For a deeper post-setup scan, run: make doctor ``` -Supported toolchain versions are documented in -[`platform/docs/HOST_PLATFORMS.md`](platform/docs/HOST_PLATFORMS.md). The -current validated baseline is Python `3.12` or `3.13`, Node `22.16.x` or -`24.x`, `uv 0.10.9`, Varlock `0.5.0`, OpenClaw `2026.3.13`, ACPX `0.3.0`, -QMD `2.0.1`, and `lossless-claw v0.3.0`. +Supported toolchain versions are documented in [`platform/docs/HOST_PLATFORMS.md`](platform/docs/HOST_PLATFORMS.md). The current validated baseline is Python `3.12` or `3.13`, Node `22.16.x` or `24.x`, `uv 0.10.9`, Varlock `0.5.0`, OpenClaw `2026.3.13`, ACPX `0.3.0`, QMD `2.0.1`, and `lossless-claw v0.3.0`. ## Development -For local development, use `uv` for the project environment and install the -pre-commit hooks once: +For local development, use `uv` for the project environment and install the pre-commit hooks once: ```bash make dev @@ -206,8 +150,7 @@ That syncs the `dev` extra into `.venv/` and installs hooks for: - actionlint for GitHub Actions - basic repository hygiene checks -If you want shorter commands in a shell session, sync the dev environment once -and activate it before running tools directly: +If you want shorter commands in a shell session, sync the dev environment once and activate it before running tools directly: ```bash uv sync --locked @@ -217,10 +160,7 @@ clawops --help deactivate ``` -`uv run ...` remains the default documented path for one-off commands from the -repo root. Activating `.venv/` is optional and is mainly useful when you want -plain `pytest`, `clawops`, and other project commands without repeating -`uv run`. +`uv run ...` remains the default documented path for one-off commands from the repo root. Activating `.venv/` is optional and is mainly useful when you want plain `pytest`, `clawops`, and other project commands without repeating `uv run`. Useful follow-up commands: @@ -236,23 +176,15 @@ make precommit make dev-check ``` -`make precommit` runs the mutating formatter/import/hygiene hooks first and then -verifies the full pre-commit stack in one final pass. +`make precommit` runs the mutating formatter/import/hygiene hooks first and then verifies the full pre-commit stack in one final pass. -`make dev-check` builds on `make precommit` and then runs the test suite plus a -compile smoke. That keeps the two targets distinct: `make precommit` is the -repository normalization and hook gate, while `make dev-check` is the deeper -development verification pass. +`make dev-check` builds on `make precommit` and then runs the test suite plus a compile smoke. That keeps the two targets distinct: `make precommit` is the repository normalization and hook gate, while `make dev-check` is the deeper development verification pass. -`make shellcheck` and `make precommit` expect a local `shellcheck` binary on -`PATH` now that the repo hook uses the system binary instead of a Docker-backed -wrapper. `brew install shellcheck` and `sudo apt-get install shellcheck` both -work. +`make shellcheck` and `make precommit` expect a local `shellcheck` binary on `PATH` now that the repo hook uses the system binary instead of a Docker-backed wrapper. `brew install shellcheck` and `sudo apt-get install shellcheck` both work. ## Devflow -The repository now exposes a production devflow surface for multi-stage -planning, execution, recovery, and audit: +The repository now exposes a production devflow surface for multi-stage planning, execution, recovery, and audit: ```bash clawops devflow plan --goal "Fix regression and add coverage" diff --git a/SETUP_GUIDE.md b/SETUP_GUIDE.md index 679c759e..911f7043 100644 --- a/SETUP_GUIDE.md +++ b/SETUP_GUIDE.md @@ -65,12 +65,9 @@ make dev make test ``` -`make test` runs in the locked managed environment via `uv run`. The baseline -verifier also uses that managed test path, and bootstrap installs `uv` when -the host does not already provide it. +`make test` runs in the locked managed environment via `uv run`. The baseline verifier also uses that managed test path, and bootstrap installs `uv` when the host does not already provide it. -If you want shorter commands in an interactive shell, sync the dev environment -once and activate `.venv` before running tools directly: +If you want shorter commands in an interactive shell, sync the dev environment once and activate `.venv` before running tools directly: ```bash uv sync --locked @@ -81,11 +78,7 @@ deactivate ## 4. Prepare the Varlock env contract -You can let the guided setup path create and repair the managed Varlock env -contract for you, or you can prepare it manually. The managed path defaults to -`~/.config/strongclaw/varlock` on Linux and -`~/Library/Application Support/StrongClaw/config/varlock` on macOS. The manual -path is: +You can let the guided setup path create and repair the managed Varlock env contract for you, or you can prepare it manually. The managed path defaults to `~/.config/strongclaw/varlock` on Linux and `~/Library/Application Support/StrongClaw/config/varlock` on macOS. The manual path is: ```bash clawops varlock-env configure --non-interactive @@ -98,14 +91,9 @@ If `varlock` is already installed on the host, you can validate the contract now varlock load --path ~/.config/strongclaw/varlock ``` -The managed env contract also carries the local Neo4j sidecar credentials -used by the codebase context provider. `clawops varlock-env configure` repairs -them automatically; for manual edits keep `NEO4J_USERNAME=neo4j` unless you -also rotate the compose-side username, and set `NEO4J_PASSWORD` to a real -secret. +The managed env contract also carries the local Neo4j sidecar credentials used by the codebase context provider. `clawops varlock-env configure` repairs them automatically; for manual edits keep `NEO4J_USERNAME=neo4j` unless you also rotate the compose-side username, and set `NEO4J_PASSWORD` to a real secret. -Before bring-up, choose how OpenClaw should authenticate to an LLM provider. -StrongClaw supports both guided and env-driven setup: +Before bring-up, choose how OpenClaw should authenticate to an LLM provider. StrongClaw supports both guided and env-driven setup: - guided/OpenClaw-managed: `make setup`, `uv run --project . clawops setup`, or `clawops setup` launches `openclaw configure --section model` when no usable model is configured, and can wire provider secrets through local `.env` values or supported Varlock plugin backends - env-driven: set provider keys in the managed `.env.local` @@ -126,16 +114,9 @@ Equivalent shell entrypoint: clawops setup ``` -That path bootstraps the host, creates or repairs the repo-local Varlock env -contract, prompts for missing setup input when interactive, supports local or -managed Varlock secret backends for provider auth, configures or validates -OpenClaw model/provider auth, renders or refreshes the host service -definitions, activates the gateway plus sidecars, and runs the baseline -verification gate. +That path bootstraps the host, creates or repairs the repo-local Varlock env contract, prompts for missing setup input when interactive, supports local or managed Varlock secret backends for provider auth, configures or validates OpenClaw model/provider auth, renders or refreshes the host service definitions, activates the gateway plus sidecars, and runs the baseline verification gate. -For a render-only pass, use `clawops setup --no-activate-services`. That path -still repairs the host/env/config state but now defers model/provider auth until -you are ready to activate the gateway. +For a render-only pass, use `clawops setup --no-activate-services`. That path still repairs the host/env/config state but now defers model/provider auth until you are ready to activate the gateway. The bootstrap flow verifies or installs: @@ -148,22 +129,11 @@ The bootstrap flow verifies or installs: - Python dependencies - host-compatible vendored `memory-lancedb-pro` dependencies -For container backends, bootstrap first looks for an existing Docker-compatible -runtime that already exposes `docker` plus `docker compose`. If it finds one, -Strongclaw uses it and does not install Docker over it. If it finds an -alternative runtime without the Docker CLI integration enabled yet, bootstrap -stops and tells you to finish that integration instead of replacing it. -Only when no Docker-compatible runtime is detected does bootstrap install -Docker as the fallback runtime. +For container backends, bootstrap first looks for an existing Docker-compatible runtime that already exposes `docker` plus `docker compose`. If it finds one, Strongclaw uses it and does not install Docker over it. If it finds an alternative runtime without the Docker CLI integration enabled yet, bootstrap stops and tells you to finish that integration instead of replacing it. Only when no Docker-compatible runtime is detected does bootstrap install Docker as the fallback runtime. -If Linux bootstrap just added the runtime user to the `docker` group, setup -now pauses before service activation. Open a fresh login shell as that user and -rerun the same `make setup` or `clawops setup` command. Completed bootstrap -work is detected automatically; `--skip-bootstrap` remains available only as a -manual override. +If Linux bootstrap just added the runtime user to the `docker` group, setup now pauses before service activation. Open a fresh login shell as that user and rerun the same `make setup` or `clawops setup` command. Completed bootstrap work is detected automatically; `--skip-bootstrap` remains available only as a manual override. -If you need a placeholder-backed profile during bring-up, rerender through the -wrapper: +If you need a placeholder-backed profile during bring-up, rerender through the wrapper: ```bash make setup SETUP_ARGS="--profile acp" @@ -171,18 +141,13 @@ make setup SETUP_ARGS="--profile hypermemory" make setup SETUP_ARGS="--profile memory-lancedb-pro" ``` -For `hypermemory`, set `HYPERMEMORY_EMBEDDING_MODEL` before you -run setup. The guided env contract fills loopback defaults for -`HYPERMEMORY_EMBEDDING_BASE_URL` and `HYPERMEMORY_QDRANT_URL` unless you override -them. +For `hypermemory`, set `HYPERMEMORY_EMBEDDING_MODEL` before you run setup. The guided env contract fills loopback defaults for `HYPERMEMORY_EMBEDDING_BASE_URL` and `HYPERMEMORY_QDRANT_URL` unless you override them. -Use `clawops doctor-host` again after any host-side package or -config change that might affect the local OpenClaw runtime contract. +Use `clawops doctor-host` again after any host-side package or config change that might affect the local OpenClaw runtime contract. ## 6. Manual config and service flow -If you want to control the render, service activation, or verification steps -separately, use the lower-level entrypoints directly. +If you want to control the render, service activation, or verification steps separately, use the lower-level entrypoints directly. Bootstrap the host: @@ -199,24 +164,17 @@ clawops render-openclaw-config This writes the merged config to `~/.openclaw/openclaw.json`. -If you bypass `make setup` / `clawops setup`, complete model/provider setup manually before -starting services: +If you bypass `make setup` / `clawops setup`, complete model/provider setup manually before starting services: ```bash clawops model-auth ensure ``` -By default it now renders the `hypermemory` stack. If you want -the built-in OpenClaw path instead, run -`clawops config memory --set-profile openclaw-default` or rerender with -`--profile openclaw-default`. +By default it now renders the `hypermemory` stack. If you want the built-in OpenClaw path instead, run `clawops config memory --set-profile openclaw-default` or rerender with `--profile openclaw-default`. -For the experimental built-in QMD path, use -`clawops config memory --set-profile openclaw-qmd` or rerender with -`--profile openclaw-qmd`. +For the experimental built-in QMD path, use `clawops config memory --set-profile openclaw-qmd` or rerender with `--profile openclaw-qmd`. -The `openclaw-qmd` profile enables QMD-backed memory retrieval and renders -the rendered QMD corpus for: +The `openclaw-qmd` profile enables QMD-backed memory retrieval and renders the rendered QMD corpus for: - `platform/docs` - `platform/skills` @@ -234,11 +192,7 @@ clawops render-openclaw-config --profile hypermemory clawops render-openclaw-config --profile memory-lancedb-pro ``` -The default `hypermemory` profile renders a self-contained -combined runtime: `lossless-claw` for context continuity plus -`strongclaw-hypermemory` with -`~/.config/strongclaw/memory/hypermemory.yaml`, `autoRecall: true`, and -`autoReflect: false`. +The default `hypermemory` profile renders a self-contained combined runtime: `lossless-claw` for context continuity plus `strongclaw-hypermemory` with `~/.config/strongclaw/memory/hypermemory.yaml`, `autoRecall: true`, and `autoReflect: false`. Install and activate services: @@ -273,8 +227,7 @@ clawops verify-platform sidecars ## 7. Verify the secure baseline -If you used `make setup` or `clawops setup`, this verification already -ran. Re-run it directly whenever you want to recheck the host baseline: +If you used `make setup` or `clawops setup`, this verification already ran. Re-run it directly whenever you want to recheck the host baseline: ```bash clawops baseline verify @@ -338,20 +291,15 @@ clawops context codebase query \ --query "operation journal idempotency" ``` -If you need the opt-in local durable memory path instead of the default -QMD-backed retrieval rollout, rerender with the Ollama-backed smart extraction -profile: +If you need the opt-in local durable memory path instead of the default QMD-backed retrieval rollout, rerender with the Ollama-backed smart extraction profile: ```bash clawops render-openclaw-config --profile memory-lancedb-pro ``` -This StrongClaw-managed profile uses Ollama-backed smart extraction, but it -still keeps `autoRecall` off and leaves session-memory/self-improvement and -management tools disabled by default. +This StrongClaw-managed profile uses Ollama-backed smart extraction, but it still keeps `autoRecall` off and leaves session-memory/self-improvement and management tools disabled by default. -If you need a combined placeholder-backed variant, use the root CLI and append -the extra overlay explicitly so every selected fragment is rendered first: +If you need a combined placeholder-backed variant, use the root CLI and append the extra overlay explicitly so every selected fragment is rendered first: ```bash clawops render-openclaw-config \ @@ -359,8 +307,7 @@ clawops render-openclaw-config \ --overlay platform/configs/openclaw/20-acp-workers.json5 ``` -Keep `platform/configs/openclaw/75-strongclaw-hypermemory.example.json5` as the -Markdown-canonical migration source while you validate parity. +Keep `platform/configs/openclaw/75-strongclaw-hypermemory.example.json5` as the Markdown-canonical migration source while you validate parity. For the supported sparse+dense hypermemory path, run: @@ -371,9 +318,7 @@ clawops hypermemory --config ~/.config/strongclaw/memory/hypermemory.yaml verify clawops doctor ``` -That profile keeps QMD out of the rendered contract and verifies that both the -dense and sparse Qdrant lanes are healthy instead of silently degrading to the -SQLite fallback path. +That profile keeps QMD out of the rendered contract and verifies that both the dense and sparse Qdrant lanes are healthy instead of silently degrading to the SQLite fallback path. ## 11. Add channels carefully @@ -446,8 +391,7 @@ Reach the gateway over SSH tunnel only: ssh -N -L 18789:127.0.0.1:18789 @ ``` -Do **not** tunnel browser-lab ports such as `9222` or `3128` to an operator -workstation. Verify the local-only posture after startup: +Do **not** tunnel browser-lab ports such as `9222` or `3128` to an operator workstation. Verify the local-only posture after startup: ```bash clawops verify-platform sidecars diff --git a/USAGE_GUIDE.md b/USAGE_GUIDE.md index 4ea21331..c0337725 100644 --- a/USAGE_GUIDE.md +++ b/USAGE_GUIDE.md @@ -34,9 +34,7 @@ systemctl --user restart openclaw-gateway.service ## Health checks -Repository helper commands that depend on `openclaw` now detect whether the CLI -is installed. Tasks that require it fail fast with a clear message, while -fallback-capable tasks warn and continue with their fallback path. +Repository helper commands that depend on `openclaw` now detect whether the CLI is installed. Tasks that require it fail fast with a clear message, while fallback-capable tasks warn and continue with their fallback path. ```bash clawops doctor-host @@ -69,8 +67,7 @@ Tunnel the gateway only: ssh -N -L 18789:127.0.0.1:18789 @ ``` -Keep browser-lab ports local to the hardened session. Do not tunnel `9222` or -`3128` to an operator workstation. +Keep browser-lab ports local to the hardened session. Do not tunnel `9222` or `3128` to an operator workstation. ## Render placeholder-backed profiles @@ -79,8 +76,7 @@ clawops render-openclaw-config \ --profile acp ``` -If you need to combine a named profile with another placeholder-bearing -overlay, append it explicitly so each fragment is rendered before merge: +If you need to combine a named profile with another placeholder-bearing overlay, append it explicitly so each fragment is rendered before merge: ```bash clawops render-openclaw-config \ @@ -115,8 +111,7 @@ Initialize: clawops op-journal init --db ~/.openclaw/clawops/op_journal.sqlite ``` -Treat `~/.openclaw/clawops` as service-owned state, not a shared scratch -directory. +Treat `~/.openclaw/clawops` as service-owned state, not a shared scratch directory. - keep directory mode `0700` on `~/.openclaw/clawops` - keep file mode `0600` on `~/.openclaw/clawops/op_journal.sqlite` @@ -134,8 +129,7 @@ clawops op-journal begin \ --payload-file payload.json ``` -`op-journal begin` is an audit/bookkeeping primitive. It does not create an -executable wrapper operation by itself. +`op-journal begin` is an audit/bookkeeping primitive. It does not create an executable wrapper operation by itself. Approve an external action: @@ -170,8 +164,7 @@ clawops policy \ ## Devflow -Use the public devflow surface for staged planning, execution, recovery, and -audit: +Use the public devflow surface for staged planning, execution, recovery, and audit: ```bash clawops devflow plan --goal "Fix regression and add coverage" @@ -183,8 +176,7 @@ clawops devflow cancel --run-id --requested-by operator clawops devflow audit --run-id ``` -Run-local state lands under `.clawops/devflow//` and the audit bundle -is written under that run's `audit/` directory. +Run-local state lands under `.clawops/devflow//` and the audit bundle is written under that run's `audit/` directory. ## Context service @@ -220,8 +212,7 @@ clawops context codebase pack \ ## Hypermemory path -StrongClaw now defaults to the supported sparse+dense memory stack. Bring it up -explicitly with: +StrongClaw now defaults to the supported sparse+dense memory stack. Bring it up explicitly with: ```bash export HYPERMEMORY_EMBEDDING_MODEL=openai/text-embedding-3-small @@ -230,9 +221,7 @@ clawops hypermemory --config ~/.config/strongclaw/memory/hypermemory.yaml verify clawops doctor ``` -That profile enables the combined `lossless-claw` + `strongclaw-hypermemory` -runtime, keeps `autoRecall` on, keeps `autoReflect` off, and verifies the -Qdrant dense+sparse backend rather than the legacy built-in QMD path. +That profile enables the combined `lossless-claw` + `strongclaw-hypermemory` runtime, keeps `autoRecall` on, keeps `autoReflect` off, and verifies the Qdrant dense+sparse backend rather than the legacy built-in QMD path. If you need the OpenClaw built-ins instead, switch to the explicit fallback: @@ -248,15 +237,9 @@ clawops config memory --set-profile openclaw-qmd ## Memory migration -Treat QMD plus the context service as the repo-document retrieval lane for the -explicit `openclaw-qmd` fallback profile. Use this section only when you need -to migrate a standalone or hypermemory `hypermemory` corpus into the vendored -`memory-lancedb-pro` plugin. +Treat QMD plus the context service as the repo-document retrieval lane for the explicit `openclaw-qmd` fallback profile. Use this section only when you need to migrate a standalone or hypermemory `hypermemory` corpus into the vendored `memory-lancedb-pro` plugin. -Export one scope at a time into the import JSON shape that `openclaw memory-pro -import` expects. When you omit `--output` or `--report`, StrongClaw now writes -those artifacts under its OS-specific state directory instead of into the git -checkout. +Export one scope at a time into the import JSON shape that `openclaw memory-pro import` expects. When you omit `--output` or `--report`, StrongClaw now writes those artifacts under its OS-specific state directory instead of into the git checkout. ```bash clawops memory migrate-hypermemory-to-pro \ @@ -264,8 +247,7 @@ clawops memory migrate-hypermemory-to-pro \ --output /tmp/project-strongclaw-import.json ``` -Apply that snapshot through the same upstream `openclaw memory-pro import` -command shape, but with a ClawOps-managed report artifact: +Apply that snapshot through the same upstream `openclaw memory-pro import` command shape, but with a ClawOps-managed report artifact: ```bash clawops memory import-pro-snapshot \ @@ -282,9 +264,7 @@ clawops memory verify-pro-parity \ --query "deployment playbook" ``` -If a live `openclaw memory-pro search` path is already available, switch -`--mode` to `openclaw` or `auto` to compare against the actual plugin-backed -search results. +If a live `openclaw memory-pro search` path is already available, switch `--mode` to `openclaw` or `auto` to compare against the actual plugin-backed search results. ## Repo workspace contract @@ -313,8 +293,7 @@ clawops skills scan \ --report platform/skills/manifests/downloaded-skill.json ``` -Promote only after the bundle has been reviewed against its manifest and hash -trail: +Promote only after the bundle has been reviewed against its manifest and hash trail: ```bash clawops skills promote \ @@ -359,9 +338,7 @@ clawops workflow \ --workflow platform/configs/workflows/code_review.yaml ``` -The shipped repository workflows declare `base_dir`, so they can be run from -any current working directory. The helper script also pins the repo root -explicitly: +The shipped repository workflows declare `base_dir`, so they can be run from any current working directory. The helper script also pins the repo root explicitly: ```bash clawops workflow --workflow platform/configs/workflows/code_review.yaml --dry-run @@ -390,9 +367,7 @@ clawops wrapper webhook \ --execute-approved ``` -If you are replaying an older approved row created before execution contracts -were introduced, pass the policy file again so the wrapper can restamp the row -before executing: +If you are replaying an older approved row created before execution contracts were introduced, pass the policy file again so the wrapper can restamp the row before executing: ```bash clawops wrapper webhook \ @@ -404,9 +379,7 @@ clawops wrapper webhook \ ## Workflow trust roots -`clawops workflow` now treats workflow YAML as executable code rather than -passive config. By default it only runs workflows from -`platform/configs/workflows/`. +`clawops workflow` now treats workflow YAML as executable code rather than passive config. By default it only runs workflows from `platform/configs/workflows/`. For intentional ad hoc local workflows, use: diff --git a/docs/testing/README.md b/docs/testing/README.md index be71ea1b..b77fd6b0 100644 --- a/docs/testing/README.md +++ b/docs/testing/README.md @@ -1,26 +1,23 @@ # Testing Docs -Strongclaw keeps contributor-facing testing documentation in this directory. -This is the canonical source of truth for how tests are organized, authored, -and run inside the repository. +Strongclaw keeps contributor-facing testing documentation in this directory. This is the canonical source of truth for how tests are organized, authored, and run inside the repository. -Testing docs do not ship in `src/clawops/assets/...` unless a packaged runtime -flow actually needs them. +Testing docs do not ship in `src/clawops/assets/...` unless a packaged runtime flow actually needs them. ## Start Here - Read [authoring.md](authoring.md) when you are writing new tests, moving - existing coverage, or touching the pytest framework. +existing coverage, or touching the pytest framework. - Read [operations.md](operations.md) when you are running tests locally, - debugging CI, or checking governance commands. +debugging CI, or checking governance commands. - Read [../../tests/fixtures/README.md](../../tests/fixtures/README.md) for - fixture-package-specific guidance only. +fixture-package-specific guidance only. ## Canonical Rules - Put tests under `tests/suites/{unit,integration,contracts,e2e,...}` and place - them in a subsystem-specific subdirectory when one exists. +them in a subsystem-specific subdirectory when one exists. - Treat `TestContext` as the default path for patching, environment mutation, - and temporary working-directory changes. +and temporary working-directory changes. - Keep pytest framework rules and contributor docs together under - `docs/testing/` instead of scattering them across packaged asset trees. +`docs/testing/` instead of scattering them across packaged asset trees. diff --git a/docs/testing/authoring.md b/docs/testing/authoring.md index 41cde00b..89c1a0b7 100644 --- a/docs/testing/authoring.md +++ b/docs/testing/authoring.md @@ -9,15 +9,9 @@ Strongclaw uses four primary default pytest lanes: - `contracts`: repository policies, docs parity, and CI/test-governance rules - `e2e`: black-box CLI and workflow-shaped orchestration coverage -The repository also maintains an explicit `framework` lane for pytest-framework -self-checks. Framework tests live under -`tests/suites/contracts/testing/framework/` and are excluded from default runs. -Use that lane only for pytest bootstrap and plugin-registration behavior. +The repository also maintains an explicit `framework` lane for pytest-framework self-checks. Framework tests live under `tests/suites/contracts/testing/framework/` and are excluded from default runs. Use that lane only for pytest bootstrap and plugin-registration behavior. -Monkeypatch governance is a default contract, not a framework-only self-check. -The direct-monkeypatch contract lives under `tests/suites/contracts/testing/` -so ordinary pytest runs fail when new unmanaged `monkeypatch` usage appears in -suite code. +Monkeypatch governance is a default contract, not a framework-only self-check. The direct-monkeypatch contract lives under `tests/suites/contracts/testing/` so ordinary pytest runs fail when new unmanaged `monkeypatch` usage appears in suite code. Capability markers are additive and remain module-local: @@ -25,8 +19,7 @@ Capability markers are additive and remain module-local: - `qdrant` - `network_local` -Structural markers are assigned from the suite path layout in -`tests/conftest.py`. +Structural markers are assigned from the suite path layout in `tests/conftest.py`. ## Placement Rules @@ -36,39 +29,31 @@ Add a new test in the suite that matches its behavior: - `tests/suites/integration/...` for cross-module or service-backed behavior - `tests/suites/contracts/...` for repository and governance rules - `tests/suites/contracts/testing/framework/...` for explicit pytest-framework - self-checks +self-checks - `tests/suites/e2e/...` for black-box CLI and workflow-shaped coverage -Prefer a dedicated subsystem directory inside each lane when one already -exists. +Prefer a dedicated subsystem directory inside each lane when one already exists. ## Runtime Boundaries -Use `tests/plugins/infrastructure/` for structural test runtime behavior such -as `TestContext`, environment management, patch management, profile -registration, and framework-owned pytest hooks. +Use `tests/plugins/infrastructure/` for structural test runtime behavior such as `TestContext`, environment management, patch management, profile registration, and framework-owned pytest hooks. Use `tests/fixtures/` for domain-facing pytest fixture plugins only. -Use `tests/utils/helpers/` for builders, subsystem runtimes, AST tooling, and -other reusable support code that is not itself a fixture or pytest framework -surface. +Use `tests/utils/helpers/` for builders, subsystem runtimes, AST tooling, and other reusable support code that is not itself a fixture or pytest framework surface. -Tests consume fixtures by name through pytest injection and should not import -from `tests.fixtures`. +Tests consume fixtures by name through pytest injection and should not import from `tests.fixtures`. ## Preferred Authoring Path -Environment mutation, working-directory changes, and patching should flow -through the infrastructure runtime: +Environment mutation, working-directory changes, and patching should flow through the infrastructure runtime: - use `test_context.patch.patch(...)` or `patch_object(...)` - use `test_context.env.set(...)`, `remove(...)`, `update(...)`, or - `prepend_path(...)` +`prepend_path(...)` - use `test_context.chdir(...)` for temporary working-directory changes -Do not add new ordinary-suite tests that depend on raw `monkeypatch` unless the -test is an explicit governed exception. +Do not add new ordinary-suite tests that depend on raw `monkeypatch` unless the test is an explicit governed exception. ## Service Resolution @@ -85,9 +70,6 @@ Current service support: ## Governance -Framework policy lives under `tests/suites/contracts/testing/`. -Add a contract test when a rule must stay true even if the implementation -changes. +Framework policy lives under `tests/suites/contracts/testing/`. Add a contract test when a rule must stay true even if the implementation changes. -Pytest-framework registration and bootstrap topology lives under -`tests/suites/contracts/testing/framework/`. +Pytest-framework registration and bootstrap topology lives under `tests/suites/contracts/testing/framework/`. diff --git a/docs/testing/operations.md b/docs/testing/operations.md index 2ede7430..dd4eb1bb 100644 --- a/docs/testing/operations.md +++ b/docs/testing/operations.md @@ -7,7 +7,7 @@ - Integration lane: `uv run pytest -q -m integration` - Contract lane: `uv run pytest -q -m contract` - Framework lane only: - `uv run pytest -q -m framework tests/suites/contracts/testing/framework` +`uv run pytest -q -m framework tests/suites/contracts/testing/framework` - E2E lane: `uv run pytest -q -m e2e` - Hypermemory lane: `uv run pytest -q -m hypermemory` - Qdrant lane: `uv run pytest -q -m "hypermemory and qdrant"` @@ -15,28 +15,27 @@ ## Service Mode Commands - Force Qdrant mock mode: - `uv run pytest -q -m "hypermemory and qdrant" --mock qdrant` +`uv run pytest -q -m "hypermemory and qdrant" --mock qdrant` - Force Qdrant real mode: - `QDRANT_TEST_MODE=real uv run pytest -q -m "hypermemory and qdrant"` +`QDRANT_TEST_MODE=real uv run pytest -q -m "hypermemory and qdrant"` - Provide an existing real endpoint: - `TEST_QDRANT_URL=http://127.0.0.1:6333 uv run pytest -q -m "hypermemory and qdrant"` +`TEST_QDRANT_URL=http://127.0.0.1:6333 uv run pytest -q -m "hypermemory and qdrant"` - Override the managed Qdrant image: - `TEST_QDRANT_IMAGE=ghcr.io/example/qdrant:test uv run pytest -q -m "hypermemory and qdrant"` +`TEST_QDRANT_IMAGE=ghcr.io/example/qdrant:test uv run pytest -q -m "hypermemory and qdrant"` ## Governance Checks - Testing contracts: `uv run pytest -q tests/suites/contracts/testing` - Fixture analysis: - `uv run python -m tests.utils.scripts.analyze_fixtures --json` +`uv run python -m tests.utils.scripts.analyze_fixtures --json` - Safe timeout wrapper: - `uv run python -m tests.utils.scripts.pytest_safe --timeout 600 -q -m integration` +`uv run python -m tests.utils.scripts.pytest_safe --timeout 600 -q -m integration` ## Common Triage - If the monkeypatch governance contract fails, migrate the test to - `TestContext` instead of expanding the allowlist by default. +`TestContext` instead of expanding the allowlist by default. - If docs or links move, rerun the repository docs contracts so relative-link - and layout drift is caught early. +and layout drift is caught early. - If a test needs reusable setup by fixture injection, add or extend a fixture. - If it needs reusable non-fixture logic, move that code into - `tests/utils/helpers/`. +If it needs reusable non-fixture logic, move that code into `tests/utils/helpers/`. diff --git a/platform/docs/BACKUP_AND_RECOVERY.md b/platform/docs/BACKUP_AND_RECOVERY.md index 12824ec1..77e9f70c 100644 --- a/platform/docs/BACKUP_AND_RECOVERY.md +++ b/platform/docs/BACKUP_AND_RECOVERY.md @@ -17,15 +17,12 @@ ## Development-mode repo-local compose state -If you keep compose state under `platform/compose/state` during development, use -the explicit dev wrappers instead of relying on implicit leftover mounts: +If you keep compose state under `platform/compose/state` during development, use the explicit dev wrappers instead of relying on implicit leftover mounts: - `clawops ops sidecars up --repo-local-state` - `clawops ops sidecars down --repo-local-state` -`clawops ops sidecars up` owns the LiteLLM schema bootstrap phase. Bring the -stack up through the CLI entrypoint instead of raw `docker compose up` when you -need the supported startup ordering on a cold Postgres state directory. +`clawops ops sidecars up` owns the LiteLLM schema bootstrap phase. Bring the stack up through the CLI entrypoint instead of raw `docker compose up` when you need the supported startup ordering on a cold Postgres state directory. Prefer targeted cleanup over deleting the whole tree: diff --git a/platform/docs/CI_AND_SECURITY.md b/platform/docs/CI_AND_SECURITY.md index 79e1bf19..d1e88d40 100644 --- a/platform/docs/CI_AND_SECURITY.md +++ b/platform/docs/CI_AND_SECURITY.md @@ -18,73 +18,49 @@ The repository includes: Workflow policy: - GitHub Actions workflows stay thin. Multi-step operational logic lives in - semantic helper entrypoints under `tests/scripts/`, with unit coverage in - `tests/suites/unit/ci/` and repo contract coverage under - `tests/suites/contracts/repo/`. +semantic helper entrypoints under `tests/scripts/`, with unit coverage in `tests/suites/unit/ci/` and repo contract coverage under `tests/suites/contracts/repo/`. ## Fresh-host acceptance -`.github/workflows/fresh-host-acceptance.yml` exercises the real bootstrap, -setup, service activation, and repo-local sidecar/browser-lab flows on hosted -Linux and macOS runners. +`.github/workflows/fresh-host-acceptance.yml` exercises the real bootstrap, setup, service activation, and repo-local sidecar/browser-lab flows on hosted Linux and macOS runners. - Each run writes a GitHub job summary with the runner label, runtime provider, - cache toggles, phase timings, and the effective hosted macOS Colima sizing. +cache toggles, phase timings, and the effective hosted macOS Colima sizing. - Each run uploads a `fresh-host-reports` artifact subtree with runtime - diagnostics (`docker info`, image inventory, launchd state, and runtime - status output) alongside the rendered host artifacts. +diagnostics (`docker info`, image inventory, launchd state, and runtime status output) alongside the rendered host artifacts. - Hosted macOS acceptance is pinned to `macos-15-intel`. GitHub's standard - `macos-15` arm64 runners are available on public repositories, but GitHub - documents that nested virtualization is not supported on arm64 macOS hosted - runners, so Colima/OrbStack cannot provide a Docker backend there. +`macos-15` arm64 runners are available on public repositories, but GitHub documents that nested virtualization is not supported on arm64 macOS hosted runners, so Colima/OrbStack cannot provide a Docker backend there. - The hosted macOS job installs Lima and Colima directly, then sizes Colima for - the runner instead of using the old fixed `2 CPU / 4 GiB` VM. +the runner instead of using the old fixed `2 CPU / 4 GiB` VM. - Hosted macOS acceptance uses the `ci-hosted-macos` compose variant so - sidecars and browser-lab mutable data live in Docker-managed volumes instead - of FUSE-backed host bind mounts. That avoids the hosted-Colima filesystem - regressions seen with Qdrant and Postgres while preserving the real `clawops` - setup, launchd activation, and repo-local stack flows. +sidecars and browser-lab mutable data live in Docker-managed volumes instead of FUSE-backed host bind mounts. That avoids the hosted-Colima filesystem regressions seen with Qdrant and Postgres while preserving the real `clawops` setup, launchd activation, and repo-local stack flows. - `workflow_dispatch` can benchmark cache toggles for the supported hosted - macOS path without changing the required PR gate. +macOS path without changing the required PR gate. - The workflow stays declarative by delegating runtime setup, image warming, - diagnostics, and summary generation to executable helper scripts under - `tests/scripts/`. Hosted macOS image warming still resolves the pinned compose - images ahead of the scenario run, but now uses bounded retries and heartbeat - logging instead of restoring immutable Docker image tarballs from the - workflow cache. +diagnostics, and summary generation to executable helper scripts under `tests/scripts/`. Hosted macOS image warming still resolves the pinned compose images ahead of the scenario run, but now uses bounded retries and heartbeat logging instead of restoring immutable Docker image tarballs from the workflow cache. - Repository workflow contract tests verify that shell steps invoking - `tests/scripts/*.py` either call an explicit Python interpreter or target an - executable script, so nightly cache warming cannot silently regress on file - mode drift. +`tests/scripts/*.py` either call an explicit Python interpreter or target an executable script, so nightly cache warming cannot silently regress on file mode drift. ## Vendored plugin verification -The vendored `platform/plugins/memory-lancedb-pro` bundle is verified on -GitHub Actions in `.github/workflows/memory-plugin-verification.yml`. +The vendored `platform/plugins/memory-lancedb-pro` bundle is verified on GitHub Actions in `.github/workflows/memory-plugin-verification.yml`. - The shared entrypoint is `the vendored-memory plugin verification workflow`. - That flow reuses `clawops config memory --set-profile memory-lancedb-pro`, which - auto-detects the host and installs the default LanceDB dependency on - supported hosts or the Intel-macOS fallback `@lancedb/lancedb@0.22.3`. +auto-detects the host and installs the default LanceDB dependency on supported hosts or the Intel-macOS fallback `@lancedb/lancedb@0.22.3`. - The workflow delegates the host-functional orchestration to - `tests/scripts/memory_plugin_verification.py`, which installs the pinned - `openclaw@2026.3.13` CLI into a - temporary tool directory and runs the host-functional - `npm run test:openclaw-host` suite. +`tests/scripts/memory_plugin_verification.py`, which installs the pinned `openclaw@2026.3.13` CLI into a temporary tool directory and runs the host-functional `npm run test:openclaw-host` suite. - The host-functional step clears ambient AWS credential env vars first so - local Bedrock model discovery noise does not contaminate test assertions. +local Bedrock model discovery noise does not contaminate test assertions. ## strongclaw-hypermemory host verification -The repo-local `platform/plugins/strongclaw-hypermemory` bundle is also verified -in `.github/workflows/memory-plugin-verification.yml`. +The repo-local `platform/plugins/strongclaw-hypermemory` bundle is also verified in `.github/workflows/memory-plugin-verification.yml`. - The shared entrypoint is `the strongclaw-hypermemory verification workflow`. - That flow runs `npm run test:openclaw-host` inside the plugin bundle. - The host-functional test creates a temporary sqlite-backed `hypermemory` - config, registers the plugin through the local SDK stub, verifies the - exported `memory` CLI surface and subcommands, and exercises the - strongclaw-owned `memory_search` and `memory_get` tool paths. +config, registers the plugin through the local SDK stub, verifies the exported `memory` CLI surface and subcommands, and exercises the strongclaw-owned `memory_search` and `memory_get` tool paths. ## Policy for new code @@ -96,40 +72,24 @@ in `.github/workflows/memory-plugin-verification.yml`. ## Dependency and release provenance - `.github/workflows/dependency-submission.yml` generates `sbom.spdx.json` with - `anchore/sbom-action` and submits the resulting dependency snapshot to the - GitHub dependency graph. +`anchore/sbom-action` and submits the resulting dependency snapshot to the GitHub dependency graph. - `.github/workflows/security.yml`, - `.github/workflows/upstream-merge-validation.yml`, and - `.github/workflows/release.yml` all call the centralized - `clawops supply-chain quality-gate` surface so linting, typing, tests, - coverage, and compile checks stay aligned. +`.github/workflows/upstream-merge-validation.yml`, and `.github/workflows/release.yml` all call the centralized `clawops supply-chain quality-gate` surface so linting, typing, tests, coverage, and compile checks stay aligned. - The compatibility matrix, memory-plugin verification, security, and release - workflows delegate their nontrivial operational steps to `tests/scripts/` - helper CLIs instead of embedding shell blobs or Python heredocs directly in - YAML. +workflows delegate their nontrivial operational steps to `tests/scripts/` helper CLIs instead of embedding shell blobs or Python heredocs directly in YAML. - Those Ubuntu quality-gate workflows install the distro `shellcheck` binary - before invoking the shared gate, and the repo's `pre-commit` hook now uses - that system binary instead of a Docker-backed hook. +before invoking the shared gate, and the repo's `pre-commit` hook now uses that system binary instead of a Docker-backed hook. - `.github/workflows/security.yml` installs a pinned `semgrep` CLI directly - instead of relying on the Docker-backed Semgrep action, which keeps the lane - off Docker Hub. +instead of relying on the Docker-backed Semgrep action, which keeps the lane off Docker Hub. - `.github/workflows/security.yml` verifies the pinned `gitleaks` and `syft` - tarball SHA-256 digests before extracting the binaries through the dedicated - helper script. +tarball SHA-256 digests before extracting the binaries through the dedicated helper script. - `.github/workflows/release.yml` syncs the locked `uv` dev environment, builds - the Python sdist/wheel only after the repository quality gate passes, verifies - each artifact with `twine check` plus fresh install smoke tests through the - dedicated release helper script, publishes or updates the GitHub release with - `gh`, and emits GitHub attestations for both build provenance and the - generated SBOM. +the Python sdist/wheel only after the repository quality gate passes, verifies each artifact with `twine check` plus fresh install smoke tests through the dedicated release helper script, publishes or updates the GitHub release with `gh`, and emits GitHub attestations for both build provenance and the generated SBOM. - `.github/workflows/upstream-merge-validation.yml` runs the repo quality gate - plus nightly validation steps after an upstream merge lands in the fork. +plus nightly validation steps after an upstream merge lands in the fork. - `.github/workflows/memory-plugin-verification.yml` runs the dedicated - hypermemory Qdrant checks against the official pinned Qdrant GHCR image - instead of Docker Hub. +hypermemory Qdrant checks against the official pinned Qdrant GHCR image instead of Docker Hub. - `.github/workflows/devflow-contract.yml` syncs the locked environment, - compile-checks the repo, runs targeted devflow tests, and validates - `clawops devflow plan --goal "contract smoke"` without live ACP - providers. +compile-checks the repo, runs targeted devflow tests, and validates `clawops devflow plan --goal "contract smoke"` without live ACP providers. - Operators can verify published provenance with GitHub's attestation tooling - after a tagged release lands. +after a tagged release lands. diff --git a/platform/docs/CONTEXT_SERVICE.md b/platform/docs/CONTEXT_SERVICE.md index ebf7f44b..d2ae8f4e 100644 --- a/platform/docs/CONTEXT_SERVICE.md +++ b/platform/docs/CONTEXT_SERVICE.md @@ -1,8 +1,6 @@ # Codebase Context Provider -This repo exposes codebase context under the generic `clawops context` namespace. -The first explicit provider is `codebase`, which keeps the SQLite lexical baseline -and adds scale-aware chunk, hybrid, and graph state. +This repo exposes codebase context under the generic `clawops context` namespace. The first explicit provider is `codebase`, which keeps the SQLite lexical baseline and adds scale-aware chunk, hybrid, and graph state. ## What it does @@ -33,8 +31,7 @@ clawops context codebase worker --scale medium --config platform/configs/context clawops context codebase benchmark --scale medium --config platform/configs/context/codebase.yaml --repo . --fixtures platform/configs/context/benchmarks/codebase.yaml --json ``` -The shipped config keeps default indexing focused on repo-authored material by -excluding large vendored trees and the benchmark fixture directory. +The shipped config keeps default indexing focused on repo-authored material by excluding large vendored trees and the benchmark fixture directory. ## Included integrations @@ -45,13 +42,11 @@ excluding large vendored trees and the benchmark fixture directory. ## Default memory retrieval -The default StrongClaw render path is `hypermemory`, which uses -`lossless-claw` plus `strongclaw-hypermemory`. +The default StrongClaw render path is `hypermemory`, which uses `lossless-claw` plus `strongclaw-hypermemory`. The explicit `openclaw-default` fallback profile keeps the OpenClaw built-ins only. -The explicit `openclaw-qmd` fallback profile enables QMD-backed memory -retrieval. +The explicit `openclaw-qmd` fallback profile enables QMD-backed memory retrieval. The rendered QMD corpus for `openclaw-qmd` includes: @@ -123,18 +118,16 @@ The shipped codebase config supports: Path filters are applied to repo-relative POSIX paths before indexing. -The shipped config excludes packaged runtime mirrors under `src/clawops/assets/**` -so retrieval ranks the real source tree instead of duplicated install assets. +The shipped config excludes packaged runtime mirrors under `src/clawops/assets/**` so retrieval ranks the real source tree instead of duplicated install assets. Symlink handling is explicit: - `in_repo_only` follows symlinks only when the resolved target stays inside the - configured repo root +configured repo root - `never` skips all symlinked files - `follow` follows all symlinked files and should only be used intentionally -The default shipped policy is `in_repo_only` to prevent context packs from -pulling host files from outside the repo tree. +The default shipped policy is `in_repo_only` to prevent context packs from pulling host files from outside the repo tree. Reindexing is authoritative for the configured file universe: @@ -153,28 +146,15 @@ Hybrid runtime artifacts are intentionally deferred for medium and large reindex - `worker` reconciles pending Qdrant point deletions and chunk-vector upserts in the background - `query` and `pack` transparently fall back to lexical chunk retrieval until the worker finishes a healthy hybrid sync -Successful hybrid batches now persist incrementally, so a late local embedding -timeout leaves the worker degraded but does not discard already indexed chunk -vectors. The next worker or benchmark pass resumes from the remaining chunks. +Successful hybrid batches now persist incrementally, so a late local embedding timeout leaves the worker degraded but does not discard already indexed chunk vectors. The next worker or benchmark pass resumes from the remaining chunks. -When chunk hashes and pending deletions already match the indexed corpus, the -worker reuses the persisted sparse metadata and skips both sparse rebuild and -dense re-embedding work. On cold syncs, the sparse lane now normalizes each -chunk once and reuses that token stream for both corpus statistics and the -persisted sparse vector payloads. +When chunk hashes and pending deletions already match the indexed corpus, the worker reuses the persisted sparse metadata and skips both sparse rebuild and dense re-embedding work. On cold syncs, the sparse lane now normalizes each chunk once and reuses that token stream for both corpus statistics and the persisted sparse vector payloads. -The Neo4j lane now uses the official Python driver over the Bolt protocol and -keeps the SQLite graph fallback aligned by materializing symbol-aware -`DEFINES`, `CALLS`, and `REFERENCES` edges alongside file import edges. +The Neo4j lane now uses the official Python driver over the Bolt protocol and keeps the SQLite graph fallback aligned by materializing symbol-aware `DEFINES`, `CALLS`, and `REFERENCES` edges alongside file import edges. ## Benchmark fixtures -Use `clawops context codebase benchmark` to measure Recall@k and MRR against a -curated fixture file. The command reindexes the repo, consolidates runtime -artifacts, and then evaluates each case against the final context surface for -the current provider scale: the top retrieved roots plus graph dependency expansion -when that lane is active. When the fixture file lives under the repo -root, the benchmark command excludes that exact file from indexing so the query set cannot self-match. +Use `clawops context codebase benchmark` to measure Recall@k and MRR against a curated fixture file. The command reindexes the repo, consolidates runtime artifacts, and then evaluates each case against the final context surface for the current provider scale: the top retrieved roots plus graph dependency expansion when that lane is active. When the fixture file lives under the repo root, the benchmark command excludes that exact file from indexing so the query set cannot self-match. Fixture files live under `platform/configs/context/benchmarks/` and accept: @@ -193,20 +173,13 @@ Benchmark authoring guidance: - Keep the shipped benchmark CLI example on `--scale medium` when validating hybrid recall expectations. - The shipped local medium config keeps embedding batches conservative and allows longer HTTP timeouts because direct Ollama embedding can still exceed optimistic defaults on singleton retries. -For the shipped local sidecar stack, the codebase provider reads Neo4j -credentials from `NEO4J_USERNAME` and `NEO4J_PASSWORD`. The Varlock env -contract now carries those keys directly, and the compose stack derives the -container's `NEO4J_AUTH` value from the same pair so fresh-host bring-up and -graph-backed retrieval stay aligned. +For the shipped local sidecar stack, the codebase provider reads Neo4j credentials from `NEO4J_USERNAME` and `NEO4J_PASSWORD`. The Varlock env contract now carries those keys directly, and the compose stack derives the container's `NEO4J_AUTH` value from the same pair so fresh-host bring-up and graph-backed retrieval stay aligned. -The shipped local medium lane uses direct Ollama embeddings plus the shared -Qdrant and rerank contract: +The shipped local medium lane uses direct Ollama embeddings plus the shared Qdrant and rerank contract: - `HYPERMEMORY_EMBEDDING_MODEL` with a pulled local Ollama embedding model such as `ollama/nomic-embed-text` - direct Ollama on `http://127.0.0.1:11434` - `HYPERMEMORY_QDRANT_URL` - optional rerank fallback keys `HYPERMEMORY_RERANK_BASE_URL`, `HYPERMEMORY_RERANK_MODEL`, and `HYPERMEMORY_RERANK_API_KEY` -If you want the codebase provider to route embeddings through LiteLLM or a -remote OpenAI-compatible gateway instead, switch `embedding.provider` back to -`compatible-http` and point `embedding.base_url` at that endpoint. +If you want the codebase provider to route embeddings through LiteLLM or a remote OpenAI-compatible gateway instead, switch `embedding.provider` back to `compatible-http` and point `embedding.base_url` at that endpoint. diff --git a/platform/docs/DEVFLOW.md b/platform/docs/DEVFLOW.md index 4326f426..68d58c53 100644 --- a/platform/docs/DEVFLOW.md +++ b/platform/docs/DEVFLOW.md @@ -11,8 +11,7 @@ clawops devflow cancel clawops devflow audit ``` -Use `--project-root ` when the control root for run state, journals, and -audit artifacts is not the current working directory. +Use `--project-root ` when the control root for run state, journals, and audit artifacts is not the current working directory. ## Operator Flow diff --git a/platform/docs/HOST_PLATFORMS.md b/platform/docs/HOST_PLATFORMS.md index a65d3c11..cf1618a5 100644 --- a/platform/docs/HOST_PLATFORMS.md +++ b/platform/docs/HOST_PLATFORMS.md @@ -5,18 +5,13 @@ Strongclaw supports two operator-host platforms: - macOS hosts using Homebrew plus `launchd` - Linux hosts using `apt-get` plus user-level `systemd` -Both use the same bootstrap entrypoints, config overlays, and verification -gates. +Both use the same bootstrap entrypoints, config overlays, and verification gates. -The bootstrap contract is runtime-aware: if the host already has a -Docker-compatible runtime that exposes `docker` plus `docker compose`, -Strongclaw reuses it. Docker is installed only as the fallback runtime when no -compatible backend is detected. +The bootstrap contract is runtime-aware: if the host already has a Docker-compatible runtime that exposes `docker` plus `docker compose`, Strongclaw reuses it. Docker is installed only as the fallback runtime when no compatible backend is detected. ## Compatibility matrix -StrongClaw's supported baseline is derived from the codebase constraints plus -the pinned external tools that setup installs. +StrongClaw's supported baseline is derived from the codebase constraints plus the pinned external tools that setup installs. | Component | Supported / pinned version | Why | | --- | --- | --- | @@ -49,9 +44,7 @@ For low-end or older hosts, this split matters: ## Runtime data locations -StrongClaw-generated runtime artifacts should not live inside the repository -checkout. The setup, doctor, harness, ACP runner, workflow context-pack, and -compose helper commands now default to OS-appropriate app directories instead. +StrongClaw-generated runtime artifacts should not live inside the repository checkout. The setup, doctor, harness, ACP runner, workflow context-pack, and compose helper commands now default to OS-appropriate app directories instead. | Kind | Linux default | macOS default | | --- | --- | --- | @@ -65,10 +58,7 @@ compose helper commands now default to OS-appropriate app directories instead. | QMD package files | `/qmd` | `/qmd` | | `lossless-claw` checkout | `/plugins/lossless-claw` | `/plugins/lossless-claw` | -Use `STRONGCLAW_DATA_DIR`, `STRONGCLAW_STATE_DIR`, `STRONGCLAW_LOG_DIR`, -`STRONGCLAW_RUNS_DIR`, or `STRONGCLAW_COMPOSE_STATE_DIR` when an operator needs -to override those defaults. The Python-owned compose commands export -`STRONGCLAW_COMPOSE_STATE_DIR` automatically before invoking Docker Compose. +Use `STRONGCLAW_DATA_DIR`, `STRONGCLAW_STATE_DIR`, `STRONGCLAW_LOG_DIR`, `STRONGCLAW_RUNS_DIR`, or `STRONGCLAW_COMPOSE_STATE_DIR` when an operator needs to override those defaults. The Python-owned compose commands export `STRONGCLAW_COMPOSE_STATE_DIR` automatically before invoking Docker Compose. ## Shared host contract @@ -89,41 +79,38 @@ Regardless of host OS, the baseline flow is: - Preflight requires Homebrew. - OrbStack, Rancher Desktop, Colima, and Docker Desktop are all acceptable as - long as they expose `docker` plus `docker compose`. +long as they expose `docker` plus `docker compose`. - `clawops ops sidecars up` now phases hosted macOS sidecar startup: Postgres - comes up first, StrongClaw runs LiteLLM Prisma bootstrap as a transient - compose run, and only then does the long-lived LiteLLM proxy start. That - keeps cold-database bootstrap work out of the runtime health window without - encoding init containers into the steady-state compose topology. +comes up first, StrongClaw runs LiteLLM Prisma bootstrap as a transient compose run, and only then does the long-lived LiteLLM proxy start. That keeps cold-database bootstrap work out of the runtime health window without encoding init containers into the steady-state compose topology. - If one of those runtimes is installed but its Docker CLI integration is not - enabled yet, bootstrap stops instead of installing Docker over it. +enabled yet, bootstrap stops instead of installing Docker over it. - Service definitions render into `~/Library/LaunchAgents`. - Activate them with `launchctl bootstrap gui/$(id -u) ...`. - The runtime-user and loopback-SSH flow is documented in - `platform/docs/runbooks/macos-service-user-and-ssh.md`. +`platform/docs/runbooks/macos-service-user-and-ssh.md`. ## Linux host notes - The current Linux bootstrap path targets Debian/Ubuntu-style hosts with - `apt-get`. +`apt-get`. - Existing Docker-compatible runtimes are reused when they already expose - `docker` plus `docker compose` for the runtime user. +`docker` plus `docker compose` for the runtime user. - If no compatible runtime is detected, bootstrap installs Docker Engine as the - fallback backend. +fallback backend. - Provision the runtime user with `your platform-native runtime-user provisioning flow`. - Service definitions render into `~/.config/systemd/user`. - Activate them with `systemctl --user daemon-reload` and - `systemctl --user enable --now ...`. +`systemctl --user enable --now ...`. - Use `loginctl enable-linger ` when user services must survive logout. - Prefer rootless Docker or a tightly controlled `docker` group for the runtime - user. +user. - The runtime-user and user-systemd flow is documented in - `platform/docs/runbooks/linux-runtime-user-and-systemd.md`. +`platform/docs/runbooks/linux-runtime-user-and-systemd.md`. ## Separate-host guidance - Browser lab belongs on a separate host or hardened user session. - ACP workers can run on the same operator host for evaluation, but a separate - worker host is the safer steady-state. +worker host is the safer steady-state. - Langfuse and similar observability extras can live on a separate VM or sidecar - host without changing the control-plane bootstrap contract. +host without changing the control-plane bootstrap contract. diff --git a/platform/docs/HYPERMEMORY.md b/platform/docs/HYPERMEMORY.md index 9bc57036..ba4ae0c2 100644 --- a/platform/docs/HYPERMEMORY.md +++ b/platform/docs/HYPERMEMORY.md @@ -1,7 +1,6 @@ # StrongClaw Hypermemory -`hypermemory` is StrongClaw's Markdown-canonical durable memory engine. It is the default StrongClaw memory stack through the `hypermemory` profile, which -binds: +`hypermemory` is StrongClaw's Markdown-canonical durable memory engine. It is the default StrongClaw memory stack through the `hypermemory` profile, which binds: - `plugins.slots.contextEngine = "lossless-claw"` - `plugins.slots.memory = "strongclaw-hypermemory"` @@ -115,20 +114,9 @@ That flow renders the default StrongClaw stack with `lossless-claw`, `strongclaw The hypermemory env contract requires `HYPERMEMORY_EMBEDDING_MODEL`. Guided setup backfills loopback defaults for `HYPERMEMORY_EMBEDDING_BASE_URL` and `HYPERMEMORY_QDRANT_URL` unless you override them. -The shipped hypermemory configs also enable planner-stage reranking. The -primary provider is `local-sentence-transformers` with -`BAAI/bge-reranker-v2-m3`; the fallback is `compatible-http`, which activates -when `HYPERMEMORY_RERANK_BASE_URL` is configured and reachable. If neither -provider is available, search fails open and keeps the provisional hybrid -planner order. - -Plain `uv sync` keeps the primary local rerank path on host/Python combinations -with known upstream wheel support: macOS arm64, macOS x86_64 on Python 3.12, -and Linux x86_64 or aarch64/arm64 on Python 3.12 or 3.13. For Raspberry Pi, -that means Raspberry Pi 4/5 with 64-bit Raspberry Pi OS or Ubuntu arm64 stay -on the primary local rerank path. Unsupported combinations such as 32-bit Pi -Linux skip the local dependency and use `compatible-http` or fail-open behavior -instead of blocking setup. +The shipped hypermemory configs also enable planner-stage reranking. The primary provider is `local-sentence-transformers` with `BAAI/bge-reranker-v2-m3`; the fallback is `compatible-http`, which activates when `HYPERMEMORY_RERANK_BASE_URL` is configured and reachable. If neither provider is available, search fails open and keeps the provisional hybrid planner order. + +Plain `uv sync` keeps the primary local rerank path on host/Python combinations with known upstream wheel support: macOS arm64, macOS x86_64 on Python 3.12, and Linux x86_64 or aarch64/arm64 on Python 3.12 or 3.13. For Raspberry Pi, that means Raspberry Pi 4/5 with 64-bit Raspberry Pi OS or Ubuntu arm64 stay on the primary local rerank path. Unsupported combinations such as 32-bit Pi Linux skip the local dependency and use `compatible-http` or fail-open behavior instead of blocking setup. Optional fallback env vars: @@ -252,10 +240,4 @@ Corpus path globs are segment-aware: - `**/*.md` matches markdown files recursively below the configured corpus root - if two configured corpus sources resolve to the same repo-relative document, the first configured source wins and later duplicates are skipped during indexing -When `CLAWOPS_STRUCTURED_LOGS=1` is set, hypermemory emits compact JSON lines -for embedding calls, Qdrant search, lexical planning, fusion, rerank, rerank -errors, fallback activation, and vector sync. When OTLP tracing is enabled -through `CLAWOPS_OTEL_ENABLED=1` or the standard `OTEL_EXPORTER_OTLP_*` -variables, the same operations emit spans through the shared ClawOps -observability pipeline, including a dedicated `clawops.hypermemory.rerank` -span. +When `CLAWOPS_STRUCTURED_LOGS=1` is set, hypermemory emits compact JSON lines for embedding calls, Qdrant search, lexical planning, fusion, rerank, rerank errors, fallback activation, and vector sync. When OTLP tracing is enabled through `CLAWOPS_OTEL_ENABLED=1` or the standard `OTEL_EXPORTER_OTLP_*` variables, the same operations emit spans through the shared ClawOps observability pipeline, including a dedicated `clawops.hypermemory.rerank` span. diff --git a/platform/docs/MODEL_ROUTING.md b/platform/docs/MODEL_ROUTING.md index 7bdd00d8..d60aa378 100644 --- a/platform/docs/MODEL_ROUTING.md +++ b/platform/docs/MODEL_ROUTING.md @@ -24,8 +24,7 @@ The supported `hypermemory` profile uses: - `hypermemory-embedding` as the stable LiteLLM alias consumed by `hypermemory` - `HYPERMEMORY_EMBEDDING_BASE_URL` to point `hypermemory` at the loopback LiteLLM route -This keeps the hypermemory memory config pinned to a stable route while letting -operators swap the upstream embedding model behind that alias. +This keeps the hypermemory memory config pinned to a stable route while letting operators swap the upstream embedding model behind that alias. ## Role defaults diff --git a/platform/docs/POLICY_ENGINE_AND_WRAPPERS.md b/platform/docs/POLICY_ENGINE_AND_WRAPPERS.md index c210b02e..6abb85eb 100644 --- a/platform/docs/POLICY_ENGINE_AND_WRAPPERS.md +++ b/platform/docs/POLICY_ENGINE_AND_WRAPPERS.md @@ -60,16 +60,14 @@ Wrapper transport remains intentionally conservative: - requests use a stable `User-Agent` and explicit timeout - terminal success/failure rows persist request method/url, request attempt count, - status code, body excerpt, and a typed transport error code when relevant +status code, body excerpt, and a typed transport error code when relevant - automatic retries are endpoint-scoped, not blanket behavior - current mutating wrappers (`github.comment.create`, - `github.issue.labels.add`, `github.pull_request.merge`, `webhook.post`) stay - on explicit no-retry policies until endpoint-level idempotency is proven +`github.issue.labels.add`, `github.pull_request.merge`, `webhook.post`) stay on explicit no-retry policies until endpoint-level idempotency is proven ## Execution contract -Wrappers now persist an execution contract alongside the stored policy decision. -That contract binds execution to the prepared operation metadata: +Wrappers now persist an execution contract alongside the stored policy decision. That contract binds execution to the prepared operation metadata: - scope - kind @@ -78,11 +76,9 @@ That contract binds execution to the prepared operation metadata: - input hash - policy decision -`clawops op-journal begin` remains available for generic audit/bookkeeping use, -but those rows are not executable wrapper operations by themselves. +`clawops op-journal begin` remains available for generic audit/bookkeeping use, but those rows are not executable wrapper operations by themselves. -This prevents a forged journal row from bypassing policy and allowlist checks at -`--execute-approved` time. +This prevents a forged journal row from bypassing policy and allowlist checks at `--execute-approved` time. ## Journal ownership @@ -139,14 +135,11 @@ clawops wrapper webhook \ Legacy note: - older approved rows that predate execution contracts may require an explicit - policy file on `--execute-approved` so the wrapper can restamp the row before - sending the side effect +policy file on `--execute-approved` so the wrapper can restamp the row before sending the side effect ## Devflow audit spine -The production devflow surface reuses the same SQLite audit spine in -`op_journal.sqlite` and adds devflow-specific run, stage, and stage-event -tables. Operators inspect and recover that workflow state through: +The production devflow surface reuses the same SQLite audit spine in `op_journal.sqlite` and adds devflow-specific run, stage, and stage-event tables. Operators inspect and recover that workflow state through: ```bash clawops devflow status --run-id diff --git a/platform/docs/SECRETS_AND_ENV.md b/platform/docs/SECRETS_AND_ENV.md index deac67aa..b1949548 100644 --- a/platform/docs/SECRETS_AND_ENV.md +++ b/platform/docs/SECRETS_AND_ENV.md @@ -8,9 +8,9 @@ ## Files - managed Varlock env dir: `~/.config/strongclaw/varlock` on Linux, - `~/Library/Application Support/StrongClaw/config/varlock` on macOS +`~/Library/Application Support/StrongClaw/config/varlock` on macOS - source template assets: `platform/configs/varlock/.env.schema`, - `platform/configs/varlock/.env.local.example` +`platform/configs/varlock/.env.local.example` - managed plugin overlay: `.env.plugins` (generated locally when you choose a managed secret backend) - `platform/examples/openclaw-secretref-*.json5` @@ -40,11 +40,9 @@ - StrongClaw keeps core machine-local secrets such as the gateway token and LiteLLM bootstrap secrets in `.env.local` by default. - Managed backends are primarily used for LLM provider credentials and similar integration tokens. - `HYPERMEMORY_EMBEDDING_MODEL` is the only hypermemory env key that - normally needs operator input; the loopback base URLs are backfilled by - guided setup unless you override them. +normally needs operator input; the loopback base URLs are backfilled by guided setup unless you override them. - `.env.plugins` is ignored by git and imported from `.env.schema`, so manual and guided setup can be mixed safely. If you author it by hand, keep explicit plugin version specifiers in each `@plugin(...)` decorator. ## Rotation -Use `clawops recovery rotate-secrets` and the runbook: -`platform/docs/runbooks/credential-rotation.md` +Use `clawops recovery rotate-secrets` and the runbook: `platform/docs/runbooks/credential-rotation.md` diff --git a/platform/docs/SECURITY_MODEL.md b/platform/docs/SECURITY_MODEL.md index f1935e8d..ab13fa86 100644 --- a/platform/docs/SECURITY_MODEL.md +++ b/platform/docs/SECURITY_MODEL.md @@ -28,7 +28,6 @@ Require an independent reviewer for: ## Skill intake -Use: -`clawops skill-scan --source --quarantine --report ` +Use: `clawops skill-scan --source --quarantine --report ` Never auto-enable a new skill or plugin directly from a download path. diff --git a/platform/docs/runbooks/linux-runtime-user-and-systemd.md b/platform/docs/runbooks/linux-runtime-user-and-systemd.md index 493ec74e..24be3479 100644 --- a/platform/docs/runbooks/linux-runtime-user-and-systemd.md +++ b/platform/docs/runbooks/linux-runtime-user-and-systemd.md @@ -1,7 +1,6 @@ # Linux runtime user and user-level systemd -Create a dedicated non-admin account for the OpenClaw runtime and keep it -separate from your daily admin shell. +Create a dedicated non-admin account for the OpenClaw runtime and keep it separate from your daily admin shell. Preferred pattern: 1. create `openclawsvc` diff --git a/platform/docs/runbooks/macos-service-user-and-ssh.md b/platform/docs/runbooks/macos-service-user-and-ssh.md index 29065683..cf3dd901 100644 --- a/platform/docs/runbooks/macos-service-user-and-ssh.md +++ b/platform/docs/runbooks/macos-service-user-and-ssh.md @@ -1,7 +1,6 @@ # macOS runtime user and loopback SSH -Create a dedicated standard user for the OpenClaw runtime and keep it separate -from your daily admin account. +Create a dedicated standard user for the OpenClaw runtime and keep it separate from your daily admin account. Preferred pattern: 1. create `openclawsvc` @@ -16,5 +15,4 @@ Host-aware entrypoint: Create the dedicated runtime user with your platform-native user-management tooling ``` -You can then manage the runtime with: -`ssh openclawsvc@localhost` +You can then manage the runtime with: `ssh openclawsvc@localhost` diff --git a/platform/plugins/memory-lancedb-pro/README.md b/platform/plugins/memory-lancedb-pro/README.md index e380addf..48f1e6b9 100644 --- a/platform/plugins/memory-lancedb-pro/README.md +++ b/platform/plugins/memory-lancedb-pro/README.md @@ -8,10 +8,7 @@ A LanceDB-backed OpenClaw memory plugin that stores preferences, decisions, and project context, then auto-recalls them in future sessions. -[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) -[![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) -[![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) -[![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) +[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) [![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) [![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) [![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) [English](README.md) | [简体中文](README_CN.md) | [繁體中文](README_TW.md) | [日本語](README_JA.md) | [한국어](README_KO.md) | [Français](README_FR.md) | [Español](README_ES.md) | [Deutsch](README_DE.md) | [Italiano](README_IT.md) | [Русский](README_RU.md) | [Português (Brasil)](README_PT-BR.md) @@ -134,8 +131,7 @@ You should see: Done! Your agent now has long-term memory. -
-More installation paths (existing users, upgrades) +
More installation paths (existing users, upgrades) **Already using OpenClaw?** @@ -160,8 +156,7 @@ See `CHANGELOG-v1.1.0.md` for behavior changes and upgrade rationale.
-
-Telegram Bot Quick Import (click to expand) +
Telegram Bot Quick Import (click to expand) If you are using OpenClaw's Telegram integration, the easiest way is to send an import command directly to the main Bot instead of manually editing config. @@ -244,11 +239,9 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > Full walkthrough: installation, configuration, and hybrid retrieval internals. -[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) -**https://youtu.be/MtukF1C8epQ** +[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) **https://youtu.be/MtukF1C8epQ** -[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) -**https://www.bilibili.com/video/BV1zUf2BGEgn/** +[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) **https://www.bilibili.com/video/BV1zUf2BGEgn/** --- @@ -278,8 +271,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > For a deep-dive into the full architecture, see [docs/memory_architecture_analysis.md](docs/memory_architecture_analysis.md). -
-File Reference (click to expand) +
File Reference (click to expand) | File | Purpose | | --- | --- | @@ -371,8 +363,7 @@ Query → BM25 FTS ─────┘ --- -
-Compared to Built-in memory-lancedb (click to expand) +
Compared to Built-in memory-lancedb (click to expand) | Feature | Built-in `memory-lancedb` | **memory-lancedb-pro** | | --- | :---: | :---: | @@ -399,8 +390,7 @@ Query → BM25 FTS ─────┘ ## Configuration -
-Full Configuration Example +
Full Configuration Example ```json { @@ -464,8 +454,7 @@ Query → BM25 FTS ─────┘
-
-Embedding Providers +
Embedding Providers Works with **any OpenAI-compatible embedding API**: @@ -479,8 +468,7 @@ Works with **any OpenAI-compatible embedding API**:
-
-Rerank Providers +
Rerank Providers Cross-encoder reranking supports multiple providers via `rerankProvider`: @@ -495,8 +483,7 @@ Any Jina-compatible rerank endpoint also works — set `rerankProvider: "jina"`
-
-Smart Extraction (LLM) — v1.1.0 +
Smart Extraction (LLM) — v1.1.0 When `smartExtraction` is enabled (default: `true`), the plugin uses an LLM to intelligently extract and classify memories instead of regex-based triggers. @@ -537,8 +524,7 @@ Notes for `llm.auth: "oauth"`:
-
-Lifecycle Configuration (Decay + Tier) +
Lifecycle Configuration (Decay + Tier) | Field | Default | Description | |-------|---------|-------------| @@ -553,8 +539,7 @@ Notes for `llm.auth: "oauth"`:
-
-Access Reinforcement +
Access Reinforcement Frequently recalled memories decay more slowly (spaced-repetition style). @@ -596,8 +581,7 @@ OAuth login flow: ## Advanced Topics -
-If injected memories show up in replies +
If injected memories show up in replies Sometimes the model may echo the injected `` block. @@ -611,8 +595,7 @@ Sometimes the model may echo the injected `` block.
-
-Session Memory +
Session Memory - Triggered on `/new` command — saves previous session summary to LanceDB - Disabled by default (OpenClaw already has native `.jsonl` session persistence) @@ -622,8 +605,7 @@ See [docs/openclaw-integration-playbook.md](docs/openclaw-integration-playbook.m
-
-Custom Slash Commands (e.g. /lesson) +
Custom Slash Commands (e.g. /lesson) Add to your `CLAUDE.md`, `AGENTS.md`, or system prompt: @@ -642,8 +624,7 @@ When the user sends `/remember `:
-
-Iron Rules for AI Agents +
Iron Rules for AI Agents > Copy the block below into your `AGENTS.md` so your agent enforces these rules automatically. @@ -670,8 +651,7 @@ After modifying .ts files under plugins/, MUST run rm -rf /tmp/jiti/ BEFORE open
-
-Database Schema +
Database Schema LanceDB table `memories`: @@ -692,8 +672,7 @@ Common `metadata` keys in v1.1.0: `l0_abstract`, `l1_overview`, `l2_content`, `m
-
-Troubleshooting +
Troubleshooting ### "Cannot mix BigInt and other types" (LanceDB / Apache Arrow) @@ -815,29 +794,17 @@ Feedback: [GitHub Issues](https://github.com/CortexReach/memory-lancedb-pro/issu ## Contributors -

-@win4r -@kctony -@Akatsuki-Ryu -@JasonSuz -@Minidoracat -@furedericca-lab -@joe2643 -@AliceLJY -@chenjiyong -

+

@win4r @kctony @Akatsuki-Ryu @JasonSuz @Minidoracat @furedericca-lab @joe2643 @AliceLJY @chenjiyong

Full list: [Contributors](https://github.com/CortexReach/memory-lancedb-pro/graphs/contributors) ## Star History - - + Star History Chart - - + ## License diff --git a/platform/plugins/memory-lancedb-pro/README_CN.md b/platform/plugins/memory-lancedb-pro/README_CN.md index acbfb92b..947af5ec 100644 --- a/platform/plugins/memory-lancedb-pro/README_CN.md +++ b/platform/plugins/memory-lancedb-pro/README_CN.md @@ -8,10 +8,7 @@ 基于 LanceDB 的 OpenClaw 长期记忆插件，自动存储偏好、决策和项目上下文，在后续会话中自动回忆。 -[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) -[![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) -[![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) -[![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) +[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) [![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) [![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) [![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) [English](README.md) | [简体中文](README_CN.md) | [繁體中文](README_TW.md) | [日本語](README_JA.md) | [한국어](README_KO.md) | [Français](README_FR.md) | [Español](README_ES.md) | [Deutsch](README_DE.md) | [Italiano](README_IT.md) | [Русский](README_RU.md) | [Português (Brasil)](README_PT-BR.md) @@ -134,8 +131,7 @@ openclaw logs --follow --plain | grep "memory-lancedb-pro" 完成！你的智能体现在拥有长期记忆了。 -
-更多安装路径（已有用户、升级） +
更多安装路径（已有用户、升级） **已在使用 OpenClaw？** @@ -160,8 +156,7 @@ openclaw memory-pro stats
-
-Telegram Bot 快捷导入（点击展开） +
Telegram Bot 快捷导入（点击展开） 如果你在使用 OpenClaw 的 Telegram 集成，最简单的方式是直接给主 Bot 发消息，而不是手动编辑配置文件。 @@ -244,11 +239,9 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > 完整演示：安装、配置、混合检索内部原理。 -[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) -**https://youtu.be/MtukF1C8epQ** +[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) **https://youtu.be/MtukF1C8epQ** -[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) -**https://www.bilibili.com/video/BV1zUf2BGEgn/** +[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) **https://www.bilibili.com/video/BV1zUf2BGEgn/** --- @@ -278,8 +271,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > 完整架构解析见 [docs/memory_architecture_analysis.md](docs/memory_architecture_analysis.md)。 -
-文件说明（点击展开） +
文件说明（点击展开） | 文件 | 用途 | | --- | --- | @@ -369,8 +361,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla --- -
-与内置 memory-lancedb 的对比（点击展开） +
与内置 memory-lancedb 的对比（点击展开） | 功能 | 内置 `memory-lancedb` | **memory-lancedb-pro** | | --- | :---: | :---: | @@ -397,8 +388,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla ## 配置 -
-完整配置示例 +
完整配置示例 ```json { @@ -462,8 +452,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla
-
-Embedding 服务商 +
Embedding 服务商 兼容 **任意 OpenAI 兼容 Embedding API**： @@ -477,8 +466,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla
-
-重排序服务商 +
重排序服务商 交叉编码器重排序通过 `rerankProvider` 支持多个服务商： @@ -493,8 +481,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla
-
-智能提取（LLM）— v1.1.0 +
智能提取（LLM）— v1.1.0 当 `smartExtraction` 启用（默认 `true`）时，插件使用 LLM 智能提取和分类记忆，替代基于正则的触发方式。 @@ -535,8 +522,7 @@ OAuth `llm` 配置（使用现有 Codex / ChatGPT 登录缓存来发送 LLM 请
-
-生命周期配置（衰减 + 层级） +
生命周期配置（衰减 + 层级） | 字段 | 默认值 | 说明 | |------|--------|------| @@ -551,8 +537,7 @@ OAuth `llm` 配置（使用现有 Codex / ChatGPT 登录缓存来发送 LLM 请
-
-访问强化 +
访问强化 频繁被召回的记忆衰减更慢（类似间隔重复机制）。 @@ -594,8 +579,7 @@ OAuth 登录流程： ## 进阶主题 -
-注入的记忆出现在回复中 +
注入的记忆出现在回复中 有时模型可能会将注入的 `` 块原文输出。 @@ -609,8 +593,7 @@ OAuth 登录流程：
-
-会话记忆 +
会话记忆 - 通过 `/new` 命令触发——将上一段会话摘要保存到 LanceDB - 默认关闭（OpenClaw 已有原生 `.jsonl` 会话持久化） @@ -620,8 +603,7 @@ OAuth 登录流程：
-
-自定义斜杠命令（如 /lesson） +
自定义斜杠命令（如 /lesson） 在你的 `CLAUDE.md`、`AGENTS.md` 或系统提示词中添加： @@ -640,8 +622,7 @@ OAuth 登录流程：
-
-AI 智能体铁律 +
AI 智能体铁律 > 将以下内容复制到你的 `AGENTS.md`，让智能体自动遵守这些规则。 @@ -668,8 +649,7 @@ OAuth 登录流程：
-
-数据库 Schema +
数据库 Schema LanceDB 表 `memories`： @@ -690,8 +670,7 @@ v1.1.0 常用 `metadata` 字段：`l0_abstract`、`l1_overview`、`l2_content`
-
-故障排除 +
故障排除 ### "Cannot mix BigInt and other types"（LanceDB / Apache Arrow） @@ -738,29 +717,17 @@ v1.1.0 常用 `metadata` 字段：`l0_abstract`、`l1_overview`、`l2_content` ## 贡献者 -

-@win4r -@kctony -@Akatsuki-Ryu -@JasonSuz -@Minidoracat -@furedericca-lab -@joe2643 -@AliceLJY -@chenjiyong -

+

@win4r @kctony @Akatsuki-Ryu @JasonSuz @Minidoracat @furedericca-lab @joe2643 @AliceLJY @chenjiyong

完整列表：[Contributors](https://github.com/CortexReach/memory-lancedb-pro/graphs/contributors) ## Star 趋势 - - + Star History Chart - - + ## 许可证 diff --git a/platform/plugins/memory-lancedb-pro/README_DE.md b/platform/plugins/memory-lancedb-pro/README_DE.md index 104c259b..4d2ed0e0 100644 --- a/platform/plugins/memory-lancedb-pro/README_DE.md +++ b/platform/plugins/memory-lancedb-pro/README_DE.md @@ -8,10 +8,7 @@ Ein LanceDB-basiertes OpenClaw-Langzeitgedächtnis-Plugin, das Präferenzen, Entscheidungen und Projektkontext speichert und in zukünftigen Sitzungen automatisch abruft. -[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) -[![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) -[![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) -[![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) +[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) [![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) [![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) [![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) [English](README.md) | [简体中文](README_CN.md) | [繁體中文](README_TW.md) | [日本語](README_JA.md) | [한국어](README_KO.md) | [Français](README_FR.md) | [Español](README_ES.md) | [Deutsch](README_DE.md) | [Italiano](README_IT.md) | [Русский](README_RU.md) | [Português (Brasil)](README_PT-BR.md) @@ -134,8 +131,7 @@ Sie sollten sehen: Fertig! Ihr Agent verfügt jetzt über Langzeitgedächtnis. -
-Weitere Installationswege (bestehende Benutzer, Upgrades) +
Weitere Installationswege (bestehende Benutzer, Upgrades) **Bereits OpenClaw-Benutzer?** @@ -160,8 +156,7 @@ Siehe `CHANGELOG-v1.1.0.md` für Verhaltensänderungen und Upgrade-Begründung.
-
-Telegram-Bot-Schnellimport (zum Aufklappen klicken) +
Telegram-Bot-Schnellimport (zum Aufklappen klicken) Wenn Sie die Telegram-Integration von OpenClaw verwenden, ist es am einfachsten, einen Importbefehl direkt an den Hauptbot zu senden, anstatt die Konfiguration manuell zu bearbeiten. @@ -244,11 +239,9 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > Vollständige Anleitung: Installation, Konfiguration und Funktionsweise der hybriden Suche. -[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) -**https://youtu.be/MtukF1C8epQ** +[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) **https://youtu.be/MtukF1C8epQ** -[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) -**https://www.bilibili.com/video/BV1zUf2BGEgn/** +[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) **https://www.bilibili.com/video/BV1zUf2BGEgn/** --- @@ -278,8 +271,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > Für eine detaillierte Analyse der vollständigen Architektur siehe [docs/memory_architecture_analysis.md](docs/memory_architecture_analysis.md). -
-Dateireferenz (zum Aufklappen klicken) +
Dateireferenz (zum Aufklappen klicken) | Datei | Zweck | | --- | --- | @@ -369,8 +361,7 @@ Query → BM25 FTS ─────┘ --- -
-Vergleich mit dem eingebauten memory-lancedb (zum Aufklappen klicken) +
Vergleich mit dem eingebauten memory-lancedb (zum Aufklappen klicken) | Funktion | Eingebautes `memory-lancedb` | **memory-lancedb-pro** | | --- | :---: | :---: | @@ -397,8 +388,7 @@ Query → BM25 FTS ─────┘ ## Konfiguration -
-Vollständiges Konfigurationsbeispiel +
Vollständiges Konfigurationsbeispiel ```json { @@ -462,8 +452,7 @@ Query → BM25 FTS ─────┘
-
-Embedding-Anbieter +
Embedding-Anbieter Funktioniert mit **jeder OpenAI-kompatiblen Embedding-API**: @@ -477,8 +466,7 @@ Funktioniert mit **jeder OpenAI-kompatiblen Embedding-API**:
-
-Rerank-Anbieter +
Rerank-Anbieter Cross-Encoder Reranking unterstützt mehrere Anbieter über `rerankProvider`: @@ -493,8 +481,7 @@ Jeder Jina-kompatible Rerank-Endpunkt funktioniert ebenfalls — setzen Sie `rer
-
-Intelligente Extraktion (LLM) — v1.1.0 +
Intelligente Extraktion (LLM) — v1.1.0 Wenn `smartExtraction` aktiviert ist (Standard: `true`), verwendet das Plugin ein LLM, um Erinnerungen intelligent zu extrahieren und zu klassifizieren, anstatt regex-basierte Auslöser zu verwenden. @@ -535,8 +522,7 @@ Hinweise zu `llm.auth: "oauth"`:
-
-Lebenszyklus-Konfiguration (Zerfall + Stufen) +
Lebenszyklus-Konfiguration (Zerfall + Stufen) | Feld | Standard | Beschreibung | |-------|---------|-------------| @@ -551,8 +537,7 @@ Hinweise zu `llm.auth: "oauth"`:
-
-Zugriffsverstärkung +
Zugriffsverstärkung Häufig abgerufene Erinnerungen zerfallen langsamer (Spaced-Repetition-Stil). @@ -594,8 +579,7 @@ OAuth-Login-Ablauf: ## Erweiterte Themen -
-Wenn injizierte Erinnerungen in Antworten auftauchen +
Wenn injizierte Erinnerungen in Antworten auftauchen Manchmal kann das Modell den injizierten ``-Block wiedergeben. @@ -609,8 +593,7 @@ Manchmal kann das Modell den injizierten ``-Block wiedergeben
-
-Sitzungsgedächtnis +
Sitzungsgedächtnis - Wird beim `/new`-Befehl ausgelöst — speichert die vorherige Sitzungszusammenfassung in LanceDB - Standardmäßig deaktiviert (OpenClaw hat bereits native `.jsonl`-Sitzungspersistenz) @@ -620,8 +603,7 @@ Siehe [docs/openclaw-integration-playbook.md](docs/openclaw-integration-playbook
-
-Benutzerdefinierte Slash-Befehle (z.B. /lesson) +
Benutzerdefinierte Slash-Befehle (z.B. /lesson) Fügen Sie zu Ihrer `CLAUDE.md`, `AGENTS.md` oder Ihrem Systemprompt hinzu: @@ -640,8 +622,7 @@ When the user sends `/remember `:
-
-Eiserne Regeln für KI-Agenten +
Eiserne Regeln für KI-Agenten > Kopieren Sie den folgenden Block in Ihre `AGENTS.md`, damit Ihr Agent diese Regeln automatisch durchsetzt. @@ -668,8 +649,7 @@ After modifying .ts files under plugins/, MUST run rm -rf /tmp/jiti/ BEFORE open
-
-Datenbankschema +
Datenbankschema LanceDB-Tabelle `memories`: @@ -690,8 +670,7 @@ Häufige `metadata`-Schlüssel in v1.1.0: `l0_abstract`, `l1_overview`, `l2_cont
-
-Fehlerbehebung +
Fehlerbehebung ### "Cannot mix BigInt and other types" (LanceDB / Apache Arrow) @@ -738,29 +717,17 @@ Feedback: [GitHub Issues](https://github.com/CortexReach/memory-lancedb-pro/issu ## Contributors -

-@win4r -@kctony -@Akatsuki-Ryu -@JasonSuz -@Minidoracat -@furedericca-lab -@joe2643 -@AliceLJY -@chenjiyong -

+

@win4r @kctony @Akatsuki-Ryu @JasonSuz @Minidoracat @furedericca-lab @joe2643 @AliceLJY @chenjiyong

Full list: [Contributors](https://github.com/CortexReach/memory-lancedb-pro/graphs/contributors) ## Star History - - + Star History Chart - - + ## Lizenz diff --git a/platform/plugins/memory-lancedb-pro/README_ES.md b/platform/plugins/memory-lancedb-pro/README_ES.md index fe4d173c..d6e78c90 100644 --- a/platform/plugins/memory-lancedb-pro/README_ES.md +++ b/platform/plugins/memory-lancedb-pro/README_ES.md @@ -8,10 +8,7 @@ Un plugin de memoria para OpenClaw respaldado por LanceDB que almacena preferencias, decisiones y contexto de proyectos, y los recupera automáticamente en sesiones futuras. -[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) -[![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) -[![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) -[![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) +[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) [![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) [![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) [![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) [English](README.md) | [简体中文](README_CN.md) | [繁體中文](README_TW.md) | [日本語](README_JA.md) | [한국어](README_KO.md) | [Français](README_FR.md) | [Español](README_ES.md) | [Deutsch](README_DE.md) | [Italiano](README_IT.md) | [Русский](README_RU.md) | [Português (Brasil)](README_PT-BR.md) @@ -134,8 +131,7 @@ Deberías ver: ¡Listo! Tu agente ahora tiene memoria a largo plazo. -
-Más rutas de instalación (usuarios existentes, actualizaciones) +
Más rutas de instalación (usuarios existentes, actualizaciones) **¿Ya usas OpenClaw?** @@ -160,8 +156,7 @@ Consulta `CHANGELOG-v1.1.0.md` para los cambios de comportamiento y la justifica
-
-Importación rápida para Bot de Telegram (clic para expandir) +
Importación rápida para Bot de Telegram (clic para expandir) Si usas la integración de Telegram de OpenClaw, la forma más fácil es enviar un comando de importación directamente al Bot principal en lugar de editar la configuración manualmente. @@ -244,11 +239,9 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > Recorrido completo: instalación, configuración y funcionamiento interno de la recuperación híbrida. -[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) -**https://youtu.be/MtukF1C8epQ** +[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) **https://youtu.be/MtukF1C8epQ** -[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) -**https://www.bilibili.com/video/BV1zUf2BGEgn/** +[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) **https://www.bilibili.com/video/BV1zUf2BGEgn/** --- @@ -278,8 +271,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > Para un análisis detallado de la arquitectura completa, consulta [docs/memory_architecture_analysis.md](docs/memory_architecture_analysis.md). -
-Referencia de Archivos (clic para expandir) +
Referencia de Archivos (clic para expandir) | Archivo | Propósito | | --- | --- | @@ -369,8 +361,7 @@ Query → BM25 FTS ─────┘ --- -
-Comparación con memory-lancedb integrado (clic para expandir) +
Comparación con memory-lancedb integrado (clic para expandir) | Característica | `memory-lancedb` integrado | **memory-lancedb-pro** | | --- | :---: | :---: | @@ -397,8 +388,7 @@ Query → BM25 FTS ─────┘ ## Configuración -
-Ejemplo de Configuración Completa +
Ejemplo de Configuración Completa ```json { @@ -462,8 +452,7 @@ Query → BM25 FTS ─────┘
-
-Proveedores de Embedding +
Proveedores de Embedding Funciona con **cualquier API de embedding compatible con OpenAI**: @@ -477,8 +466,7 @@ Funciona con **cualquier API de embedding compatible con OpenAI**:
-
-Proveedores de Rerank +
Proveedores de Rerank El reranking con cross-encoder admite múltiples proveedores mediante `rerankProvider`: @@ -493,8 +481,7 @@ Cualquier endpoint de rerank compatible con Jina también funciona — configura
-
-Smart Extraction (LLM) — v1.1.0 +
Smart Extraction (LLM) — v1.1.0 Cuando `smartExtraction` está habilitado (predeterminado: `true`), el plugin utiliza un LLM para extraer y clasificar recuerdos de forma inteligente en lugar de disparadores basados en regex. @@ -535,8 +522,7 @@ Notas para `llm.auth: "oauth"`:
-
-Configuración del Ciclo de Vida (Decaimiento + Nivel) +
Configuración del Ciclo de Vida (Decaimiento + Nivel) | Campo | Predeterminado | Descripción | |-------|----------------|-------------| @@ -551,8 +537,7 @@ Notas para `llm.auth: "oauth"`:
-
-Refuerzo por Acceso +
Refuerzo por Acceso Los recuerdos frecuentemente recuperados decaen más lentamente (estilo repetición espaciada). @@ -594,8 +579,7 @@ Flujo de inicio de sesión OAuth: ## Temas Avanzados -
-Si los recuerdos inyectados aparecen en las respuestas +
Si los recuerdos inyectados aparecen en las respuestas A veces el modelo puede repetir el bloque `` inyectado. @@ -609,8 +593,7 @@ A veces el modelo puede repetir el bloque `` inyectado.
-
-Memoria de Sesión +
Memoria de Sesión - Se activa con el comando `/new` — guarda el resumen de la sesión anterior en LanceDB - Deshabilitado por defecto (OpenClaw ya tiene persistencia nativa de sesión en `.jsonl`) @@ -620,8 +603,7 @@ Consulta [docs/openclaw-integration-playbook.md](docs/openclaw-integration-playb
-
-Comandos Slash Personalizados (por ejemplo, /lesson) +
Comandos Slash Personalizados (por ejemplo, /lesson) Agrega a tu `CLAUDE.md`, `AGENTS.md` o prompt del sistema (el bloque se mantiene en inglés para que el agente lo interprete correctamente): @@ -640,8 +622,7 @@ When the user sends `/remember `:
-
-Reglas de Hierro para Agentes de IA +
Reglas de Hierro para Agentes de IA > Copia el bloque de abajo en tu `AGENTS.md` para que tu agente aplique estas reglas automáticamente. Se mantiene en inglés porque es instrucción directa para el modelo. @@ -668,8 +649,7 @@ After modifying .ts files under plugins/, MUST run rm -rf /tmp/jiti/ BEFORE open
-
-Esquema de la Base de Datos +
Esquema de la Base de Datos Tabla LanceDB `memories`: @@ -690,8 +670,7 @@ Claves comunes de `metadata` en v1.1.0: `l0_abstract`, `l1_overview`, `l2_conten
-
-Solución de Problemas +
Solución de Problemas ### "Cannot mix BigInt and other types" (LanceDB / Apache Arrow) @@ -738,29 +717,17 @@ Comentarios: [GitHub Issues](https://github.com/CortexReach/memory-lancedb-pro/i ## Contributors -

-@win4r -@kctony -@Akatsuki-Ryu -@JasonSuz -@Minidoracat -@furedericca-lab -@joe2643 -@AliceLJY -@chenjiyong -

+

@win4r @kctony @Akatsuki-Ryu @JasonSuz @Minidoracat @furedericca-lab @joe2643 @AliceLJY @chenjiyong

Full list: [Contributors](https://github.com/CortexReach/memory-lancedb-pro/graphs/contributors) ## Star History - - + Star History Chart - - + ## Licencia diff --git a/platform/plugins/memory-lancedb-pro/README_FR.md b/platform/plugins/memory-lancedb-pro/README_FR.md index 19f5fc45..03906485 100644 --- a/platform/plugins/memory-lancedb-pro/README_FR.md +++ b/platform/plugins/memory-lancedb-pro/README_FR.md @@ -8,10 +8,7 @@ Un plugin de mémoire long terme pour OpenClaw basé sur LanceDB qui stocke les préférences, les décisions et le contexte du projet, puis les rappelle automatiquement dans les sessions futures. -[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) -[![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) -[![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) -[![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) +[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) [![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) [![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) [![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) [English](README.md) | [简体中文](README_CN.md) | [繁體中文](README_TW.md) | [日本語](README_JA.md) | [한국어](README_KO.md) | [Français](README_FR.md) | [Español](README_ES.md) | [Deutsch](README_DE.md) | [Italiano](README_IT.md) | [Русский](README_RU.md) | [Português (Brasil)](README_PT-BR.md) @@ -134,8 +131,7 @@ Vous devriez voir : Terminé ! Votre agent dispose maintenant d'une mémoire long terme. -
-Plus de chemins d'installation (utilisateurs existants, mises à niveau) +
Plus de chemins d'installation (utilisateurs existants, mises à niveau) **Déjà utilisateur d'OpenClaw ?** @@ -160,8 +156,7 @@ Consultez `CHANGELOG-v1.1.0.md` pour les changements de comportement et la justi
-
-Import rapide Telegram Bot (cliquez pour développer) +
Import rapide Telegram Bot (cliquez pour développer) Si vous utilisez l'intégration Telegram d'OpenClaw, le plus simple est d'envoyer une commande d'import directement au Bot principal au lieu de modifier manuellement la configuration. @@ -244,11 +239,9 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > Présentation complète : installation, configuration et fonctionnement interne de la recherche hybride. -[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) -**https://youtu.be/MtukF1C8epQ** +[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) **https://youtu.be/MtukF1C8epQ** -[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) -**https://www.bilibili.com/video/BV1zUf2BGEgn/** +[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) **https://www.bilibili.com/video/BV1zUf2BGEgn/** --- @@ -278,8 +271,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > Pour une analyse approfondie de l'architecture complète, consultez [docs/memory_architecture_analysis.md](docs/memory_architecture_analysis.md). -
-Référence des fichiers (cliquez pour développer) +
Référence des fichiers (cliquez pour développer) | Fichier | Rôle | | --- | --- | @@ -369,8 +361,7 @@ Requête → BM25 FTS ─────┘ --- -
-Comparaison avec memory-lancedb intégré (cliquez pour développer) +
Comparaison avec memory-lancedb intégré (cliquez pour développer) | Fonctionnalité | `memory-lancedb` intégré | **memory-lancedb-pro** | | --- | :---: | :---: | @@ -397,8 +388,7 @@ Requête → BM25 FTS ─────┘ ## Configuration -
-Exemple de configuration complète +
Exemple de configuration complète ```json { @@ -462,8 +452,7 @@ Requête → BM25 FTS ─────┘
-
-Fournisseurs d'embedding +
Fournisseurs d'embedding Fonctionne avec **toute API d'embedding compatible OpenAI** : @@ -477,8 +466,7 @@ Fonctionne avec **toute API d'embedding compatible OpenAI** :
-
-Fournisseurs de reranking +
Fournisseurs de reranking Le reranking cross-encoder supporte plusieurs fournisseurs via `rerankProvider` : @@ -493,8 +481,7 @@ Tout endpoint de reranking compatible Jina fonctionne également — définissez
-
-Extraction intelligente (LLM) — v1.1.0 +
Extraction intelligente (LLM) — v1.1.0 Quand `smartExtraction` est activé (par défaut : `true`), le plugin utilise un LLM pour extraire et classifier intelligemment les souvenirs au lieu de déclencheurs basés sur des regex. @@ -535,8 +522,7 @@ Notes pour `llm.auth: "oauth"` :
-
-Configuration du cycle de vie (Décroissance + Niveaux) +
Configuration du cycle de vie (Décroissance + Niveaux) | Champ | Défaut | Description | |-------|---------|-------------| @@ -551,8 +537,7 @@ Notes pour `llm.auth: "oauth"` :
-
-Renforcement par accès +
Renforcement par accès Les souvenirs fréquemment rappelés décroissent plus lentement (style répétition espacée). @@ -594,8 +579,7 @@ Flux de connexion OAuth : ## Sujets avancés -
-Si les souvenirs injectés apparaissent dans les réponses +
Si les souvenirs injectés apparaissent dans les réponses Parfois le modèle peut répéter le bloc `` injecté. @@ -609,8 +593,7 @@ Parfois le modèle peut répéter le bloc `` injecté.
-
-Mémoire de session +
Mémoire de session - Déclenchée par la commande `/new` — sauvegarde le résumé de la session précédente dans LanceDB - Désactivée par défaut (OpenClaw dispose déjà d'une persistance native de session `.jsonl`) @@ -620,8 +603,7 @@ Consultez [docs/openclaw-integration-playbook.md](docs/openclaw-integration-play
-
-Commandes slash personnalisées (ex. /lesson) +
Commandes slash personnalisées (ex. /lesson) Ajoutez à votre `CLAUDE.md`, `AGENTS.md` ou prompt système : @@ -640,8 +622,7 @@ When the user sends `/remember `:
-
-Règles d'or pour les agents IA +
Règles d'or pour les agents IA > Copiez le bloc ci-dessous dans votre `AGENTS.md` pour que votre agent applique automatiquement ces règles. @@ -668,8 +649,7 @@ After modifying .ts files under plugins/, MUST run rm -rf /tmp/jiti/ BEFORE open
-
-Schéma de la base de données +
Schéma de la base de données Table LanceDB `memories` : @@ -690,8 +670,7 @@ Clés `metadata` courantes en v1.1.0 : `l0_abstract`, `l1_overview`, `l2_content
-
-Dépannage +
Dépannage ### "Cannot mix BigInt and other types" (LanceDB / Apache Arrow) @@ -738,29 +717,17 @@ Retours : [GitHub Issues](https://github.com/CortexReach/memory-lancedb-pro/issu ## Contributors -

-@win4r -@kctony -@Akatsuki-Ryu -@JasonSuz -@Minidoracat -@furedericca-lab -@joe2643 -@AliceLJY -@chenjiyong -

+

@win4r @kctony @Akatsuki-Ryu @JasonSuz @Minidoracat @furedericca-lab @joe2643 @AliceLJY @chenjiyong

Full list: [Contributors](https://github.com/CortexReach/memory-lancedb-pro/graphs/contributors) ## Star History - - + Star History Chart - - + ## Licence diff --git a/platform/plugins/memory-lancedb-pro/README_IT.md b/platform/plugins/memory-lancedb-pro/README_IT.md index b1679682..e43bc7dc 100644 --- a/platform/plugins/memory-lancedb-pro/README_IT.md +++ b/platform/plugins/memory-lancedb-pro/README_IT.md @@ -8,10 +8,7 @@ Un plugin di memoria a lungo termine per OpenClaw basato su LanceDB che memorizza preferenze, decisioni e contesto di progetto, e li richiama automaticamente nelle sessioni future. -[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) -[![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) -[![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) -[![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) +[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) [![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) [![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) [![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) [English](README.md) | [简体中文](README_CN.md) | [繁體中文](README_TW.md) | [日本語](README_JA.md) | [한국어](README_KO.md) | [Français](README_FR.md) | [Español](README_ES.md) | [Deutsch](README_DE.md) | [Italiano](README_IT.md) | [Русский](README_RU.md) | [Português (Brasil)](README_PT-BR.md) @@ -134,8 +131,7 @@ Dovresti vedere: Fatto! Il tuo agente ora ha una memoria a lungo termine. -
-Ulteriori percorsi di installazione (utenti esistenti, aggiornamenti) +
Ulteriori percorsi di installazione (utenti esistenti, aggiornamenti) **Usi già OpenClaw?** @@ -160,8 +156,7 @@ Vedi `CHANGELOG-v1.1.0.md` per le modifiche comportamentali e le motivazioni del
-
-Importazione rapida Telegram Bot (clicca per espandere) +
Importazione rapida Telegram Bot (clicca per espandere) Se stai usando l'integrazione Telegram di OpenClaw, il modo più semplice è inviare un comando di importazione direttamente al Bot principale invece di modificare manualmente la configurazione. @@ -244,11 +239,9 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > Guida completa: installazione, configurazione e funzionamento interno della ricerca ibrida. -[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) -**https://youtu.be/MtukF1C8epQ** +[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) **https://youtu.be/MtukF1C8epQ** -[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) -**https://www.bilibili.com/video/BV1zUf2BGEgn/** +[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) **https://www.bilibili.com/video/BV1zUf2BGEgn/** --- @@ -278,8 +271,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > Per un approfondimento sull'architettura completa, consulta [docs/memory_architecture_analysis.md](docs/memory_architecture_analysis.md). -
-Riferimento file (clicca per espandere) +
Riferimento file (clicca per espandere) | File | Scopo | | --- | --- | @@ -369,8 +361,7 @@ Query → BM25 FTS ─────┘ --- -
-Confronto con memory-lancedb integrato (clicca per espandere) +
Confronto con memory-lancedb integrato (clicca per espandere) | Funzionalità | `memory-lancedb` integrato | **memory-lancedb-pro** | | --- | :---: | :---: | @@ -397,8 +388,7 @@ Query → BM25 FTS ─────┘ ## Configurazione -
-Esempio di configurazione completa +
Esempio di configurazione completa ```json { @@ -462,8 +452,7 @@ Query → BM25 FTS ─────┘
-
-Provider di embedding +
Provider di embedding Funziona con **qualsiasi API di embedding compatibile OpenAI**: @@ -477,8 +466,7 @@ Funziona con **qualsiasi API di embedding compatibile OpenAI**:
-
-Provider di rerank +
Provider di rerank Il reranking cross-encoder supporta più provider tramite `rerankProvider`: @@ -493,8 +481,7 @@ Funziona anche qualsiasi endpoint di rerank compatibile Jina — imposta `rerank
-
-Estrazione intelligente (LLM) — v1.1.0 +
Estrazione intelligente (LLM) — v1.1.0 Quando `smartExtraction` è abilitato (predefinito: `true`), il plugin utilizza un LLM per estrarre e classificare intelligentemente i ricordi invece di trigger basati su regex. @@ -535,8 +522,7 @@ Note per `llm.auth: "oauth"`:
-
-Configurazione ciclo di vita (Decadimento + Livelli) +
Configurazione ciclo di vita (Decadimento + Livelli) | Campo | Predefinito | Descrizione | |-------|---------|-------------| @@ -551,8 +537,7 @@ Note per `llm.auth: "oauth"`:
-
-Rinforzo per accesso +
Rinforzo per accesso I ricordi richiamati frequentemente decadono più lentamente (stile ripetizione spaziata). @@ -594,8 +579,7 @@ Flusso di login OAuth: ## Argomenti avanzati -
-Se i ricordi iniettati appaiono nelle risposte +
Se i ricordi iniettati appaiono nelle risposte A volte il modello può ripetere il blocco `` iniettato. @@ -609,8 +593,7 @@ A volte il modello può ripetere il blocco `` iniettato.
-
-Memoria di sessione +
Memoria di sessione - Si attiva con il comando `/new` — salva il riepilogo della sessione precedente in LanceDB - Disabilitata per impostazione predefinita (OpenClaw ha già la persistenza nativa delle sessioni in `.jsonl`) @@ -620,8 +603,7 @@ Vedi [docs/openclaw-integration-playbook.md](docs/openclaw-integration-playbook.
-
-Comandi slash personalizzati (ad es. /lesson) +
Comandi slash personalizzati (ad es. /lesson) Aggiungi al tuo `CLAUDE.md`, `AGENTS.md` o prompt di sistema: @@ -640,8 +622,7 @@ When the user sends `/remember `:
-
-Regole d'oro per agenti IA +
Regole d'oro per agenti IA > Copia il blocco seguente nel tuo `AGENTS.md` in modo che il tuo agente applichi queste regole automaticamente. @@ -668,8 +649,7 @@ After modifying .ts files under plugins/, MUST run rm -rf /tmp/jiti/ BEFORE open
-
-Schema del database +
Schema del database Tabella LanceDB `memories`: @@ -690,8 +670,7 @@ Chiavi `metadata` comuni nella v1.1.0: `l0_abstract`, `l1_overview`, `l2_content
-
-Risoluzione dei problemi +
Risoluzione dei problemi ### "Cannot mix BigInt and other types" (LanceDB / Apache Arrow) @@ -738,29 +717,17 @@ Feedback: [GitHub Issues](https://github.com/CortexReach/memory-lancedb-pro/issu ## Contributors -

-@win4r -@kctony -@Akatsuki-Ryu -@JasonSuz -@Minidoracat -@furedericca-lab -@joe2643 -@AliceLJY -@chenjiyong -

+

@win4r @kctony @Akatsuki-Ryu @JasonSuz @Minidoracat @furedericca-lab @joe2643 @AliceLJY @chenjiyong

Full list: [Contributors](https://github.com/CortexReach/memory-lancedb-pro/graphs/contributors) ## Star History - - + Star History Chart - - + ## Licenza diff --git a/platform/plugins/memory-lancedb-pro/README_JA.md b/platform/plugins/memory-lancedb-pro/README_JA.md index 0627a2de..08c665ee 100644 --- a/platform/plugins/memory-lancedb-pro/README_JA.md +++ b/platform/plugins/memory-lancedb-pro/README_JA.md @@ -8,10 +8,7 @@ LanceDB ベースの OpenClaw 長期メモリプラグイン。好み・意思決定・プロジェクトコンテキストを自動保存し、将来のセッションで自動的に想起します。 -[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) -[![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) -[![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) -[![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) +[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) [![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) [![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) [![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) [English](README.md) | [简体中文](README_CN.md) | [繁體中文](README_TW.md) | [日本語](README_JA.md) | [한국어](README_KO.md) | [Français](README_FR.md) | [Español](README_ES.md) | [Deutsch](README_DE.md) | [Italiano](README_IT.md) | [Русский](README_RU.md) | [Português (Brasil)](README_PT-BR.md) @@ -134,8 +131,7 @@ openclaw logs --follow --plain | grep "memory-lancedb-pro" 完了！あなたのエージェントは長期メモリを持つようになりました。 -
-その他のインストール方法（既存ユーザー、アップグレード） +
その他のインストール方法（既存ユーザー、アップグレード） **既に OpenClaw を使用中？** @@ -160,8 +156,7 @@ openclaw memory-pro stats
-
-Telegram Bot クイックインポート（クリックで展開） +
Telegram Bot クイックインポート（クリックで展開） OpenClaw の Telegram 連携を使用している場合、設定ファイルを手動で編集するより、メイン Bot にインポートコマンドを直接送信するのが最も簡単です。 @@ -244,11 +239,9 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > フルウォークスルー：インストール、設定、ハイブリッド検索の内部構造。 -[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) -**https://youtu.be/MtukF1C8epQ** +[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) **https://youtu.be/MtukF1C8epQ** -[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) -**https://www.bilibili.com/video/BV1zUf2BGEgn/** +[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) **https://www.bilibili.com/video/BV1zUf2BGEgn/** --- @@ -278,8 +271,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > 完全なアーキテクチャの詳細は [docs/memory_architecture_analysis.md](docs/memory_architecture_analysis.md) を参照してください。 -
-ファイルリファレンス（クリックで展開） +
ファイルリファレンス（クリックで展開） | ファイル | 用途 | | --- | --- | @@ -369,8 +361,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla --- -
-内蔵 memory-lancedb との比較（クリックで展開） +
内蔵 memory-lancedb との比較（クリックで展開） | 機能 | 内蔵 `memory-lancedb` | **memory-lancedb-pro** | | --- | :---: | :---: | @@ -397,8 +388,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla ## 設定 -
-完全な設定例 +
完全な設定例 ```json { @@ -462,8 +452,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla
-
-Embedding プロバイダー +
Embedding プロバイダー **任意の OpenAI 互換 Embedding API** で動作： @@ -477,8 +466,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla
-
-リランクプロバイダー +
リランクプロバイダー クロスエンコーダーリランキングは `rerankProvider` で複数のプロバイダーをサポート： @@ -493,8 +481,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla
-
-スマート抽出（LLM）— v1.1.0 +
スマート抽出（LLM）— v1.1.0 `smartExtraction` が有効（デフォルト：`true`）の場合、プラグインは正規表現ベースのトリガーの代わりに LLM を使用してインテリジェントにメモリを抽出・分類します。 @@ -535,8 +522,7 @@ OAuth `llm` 設定（既存の Codex / ChatGPT ログインキャッシュを使
-
-ライフサイクル設定（減衰 + 階層） +
ライフサイクル設定（減衰 + 階層） | フィールド | デフォルト | 説明 | |-------|---------|-------------| @@ -551,8 +537,7 @@ OAuth `llm` 設定（既存の Codex / ChatGPT ログインキャッシュを使
-
-アクセス強化 +
アクセス強化 頻繁に想起されるメモリはより遅く減衰します（間隔反復スタイル）。 @@ -594,8 +579,7 @@ OAuth ログインフロー： ## 応用トピック -
-注入されたメモリが応答に表示される場合 +
注入されたメモリが応答に表示される場合 モデルが注入された `` ブロックをそのまま出力してしまうことがあります。 @@ -609,8 +593,7 @@ OAuth ログインフロー：
-
-セッションメモリ +
セッションメモリ - `/new` コマンドでトリガー——前のセッションの要約を LanceDB に保存 - デフォルトで無効（OpenClaw にはネイティブの `.jsonl` セッション永続化機能あり） @@ -620,8 +603,7 @@ OAuth ログインフロー：
-
-カスタムスラッシュコマンド（例：/lesson） +
カスタムスラッシュコマンド（例：/lesson） `CLAUDE.md`、`AGENTS.md`、またはシステムプロンプトに追加： @@ -640,8 +622,7 @@ OAuth ログインフロー：
-
-AI エージェントの鉄則 +
AI エージェントの鉄則 > 以下のブロックを `AGENTS.md` にコピーして、エージェントがこれらのルールを自動的に適用するようにしてください。 @@ -668,8 +649,7 @@ plugins/ 配下の .ts ファイルを変更した後、openclaw gateway restart
-
-データベーススキーマ +
データベーススキーマ LanceDB テーブル `memories`： @@ -690,8 +670,7 @@ v1.1.0 の一般的な `metadata` キー：`l0_abstract`、`l1_overview`、`l2_c
-
-トラブルシューティング +
トラブルシューティング ### "Cannot mix BigInt and other types"（LanceDB / Apache Arrow） @@ -738,29 +717,17 @@ LanceDB 0.26+ では、一部の数値カラムが `BigInt` として返され ## コントリビューター -

-@win4r -@kctony -@Akatsuki-Ryu -@JasonSuz -@Minidoracat -@furedericca-lab -@joe2643 -@AliceLJY -@chenjiyong -

+

@win4r @kctony @Akatsuki-Ryu @JasonSuz @Minidoracat @furedericca-lab @joe2643 @AliceLJY @chenjiyong

全リスト：[Contributors](https://github.com/CortexReach/memory-lancedb-pro/graphs/contributors) ## Star 履歴 - - + Star History Chart - - + ## ライセンス diff --git a/platform/plugins/memory-lancedb-pro/README_KO.md b/platform/plugins/memory-lancedb-pro/README_KO.md index c8f165e6..c81ca062 100644 --- a/platform/plugins/memory-lancedb-pro/README_KO.md +++ b/platform/plugins/memory-lancedb-pro/README_KO.md @@ -8,10 +8,7 @@ LanceDB 기반 OpenClaw 메모리 플러그인으로, 사용자 선호도·의사결정·프로젝트 맥락을 저장하고 이후 세션에서 자동으로 불러옵니다. -[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) -[![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) -[![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) -[![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) +[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) [![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) [![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) [![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) [English](README.md) | [简体中文](README_CN.md) | [繁體中文](README_TW.md) | [日本語](README_JA.md) | [한국어](README_KO.md) | [Français](README_FR.md) | [Español](README_ES.md) | [Deutsch](README_DE.md) | [Italiano](README_IT.md) | [Русский](README_RU.md) | [Português (Brasil)](README_PT-BR.md) @@ -134,8 +131,7 @@ openclaw logs --follow --plain | grep "memory-lancedb-pro" 완료! 이제 에이전트가 장기 기억을 갖게 됩니다. -
-추가 설치 경로 (기존 사용자, 업그레이드) +
추가 설치 경로 (기존 사용자, 업그레이드) **이미 OpenClaw를 사용 중인 경우:** @@ -160,8 +156,7 @@ openclaw memory-pro stats
-
-Telegram 봇 빠른 가져오기 (클릭하여 펼치기) +
Telegram 봇 빠른 가져오기 (클릭하여 펼치기) OpenClaw의 Telegram 연동을 사용하는 경우, 수동으로 설정을 편집하는 대신 메인 봇에 가져오기 명령어를 직접 보내는 것이 가장 쉬운 방법입니다. @@ -244,11 +239,9 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > 전체 안내: 설치, 설정, 하이브리드 검색 내부 구조. -[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) -**https://youtu.be/MtukF1C8epQ** +[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) **https://youtu.be/MtukF1C8epQ** -[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) -**https://www.bilibili.com/video/BV1zUf2BGEgn/** +[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) **https://www.bilibili.com/video/BV1zUf2BGEgn/** --- @@ -278,8 +271,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > 전체 아키텍처에 대한 심층 분석은 [docs/memory_architecture_analysis.md](docs/memory_architecture_analysis.md)를 참조하세요. -
-파일 레퍼런스 (클릭하여 펼치기) +
파일 레퍼런스 (클릭하여 펼치기) | 파일 | 용도 | | --- | --- | @@ -369,8 +361,7 @@ Query → BM25 FTS ─────┘ --- -
-내장 memory-lancedb와의 비교 (클릭하여 펼치기) +
내장 memory-lancedb와의 비교 (클릭하여 펼치기) | 기능 | 내장 `memory-lancedb` | **memory-lancedb-pro** | | --- | :---: | :---: | @@ -397,8 +388,7 @@ Query → BM25 FTS ─────┘ ## 설정 -
-전체 설정 예시 +
전체 설정 예시 ```json { @@ -462,8 +452,7 @@ Query → BM25 FTS ─────┘
-
-임베딩 프로바이더 +
임베딩 프로바이더 **OpenAI 호환 임베딩 API**와 모두 동작합니다: @@ -477,8 +466,7 @@ Query → BM25 FTS ─────┘
-
-리랭크 프로바이더 +
리랭크 프로바이더 Cross-Encoder 리랭킹은 `rerankProvider`를 통해 여러 프로바이더를 지원합니다: @@ -493,8 +481,7 @@ Jina 호환 리랭크 엔드포인트도 사용 가능합니다 — `rerankProvi
-
-Smart Extraction (LLM) — v1.1.0 +
Smart Extraction (LLM) — v1.1.0 `smartExtraction`이 활성화되면 (기본값: `true`), 플러그인이 정규식 기반 트리거 대신 LLM을 사용하여 기억을 지능적으로 추출하고 분류합니다. @@ -535,8 +522,7 @@ OAuth `llm` 설정 (기존 Codex / ChatGPT 로그인 캐시를 LLM 호출에 사
-
-라이프사이클 설정 (감쇠 + 단계) +
라이프사이클 설정 (감쇠 + 단계) | 필드 | 기본값 | 설명 | |-------|---------|-------------| @@ -551,8 +537,7 @@ OAuth `llm` 설정 (기존 Codex / ChatGPT 로그인 캐시를 LLM 호출에 사
-
-접근 강화 +
접근 강화 자주 불러오는 기억은 더 느리게 감쇠합니다 (간격 반복 학습 방식). @@ -594,8 +579,7 @@ OAuth 로그인 흐름: ## 고급 주제 -
-주입된 기억이 응답에 표시되는 경우 +
주입된 기억이 응답에 표시되는 경우 가끔 모델이 주입된 `` 블록을 그대로 출력할 수 있습니다. @@ -609,8 +593,7 @@ OAuth 로그인 흐름:
-
-세션 메모리 +
세션 메모리 - `/new` 명령어 시 작동 — 이전 세션 요약을 LanceDB에 저장 - 기본적으로 비활성화 (OpenClaw에 이미 네이티브 `.jsonl` 세션 영속화 기능이 있음) @@ -620,8 +603,7 @@ OAuth 로그인 흐름:
-
-커스텀 슬래시 명령어 (예: /lesson) +
커스텀 슬래시 명령어 (예: /lesson) `CLAUDE.md`, `AGENTS.md` 또는 시스템 프롬프트에 다음을 추가하세요 (에이전트가 읽는 영문 지시문이므로 그대로 사용합니다): @@ -640,8 +622,7 @@ When the user sends `/remember `:
-
-AI 에이전트를 위한 철칙 +
AI 에이전트를 위한 철칙 > 아래 블록을 `AGENTS.md`에 복사하여 에이전트가 이 규칙을 자동으로 적용하도록 하세요 (에이전트가 읽는 영문 지시문이므로 그대로 사용합니다). @@ -668,8 +649,7 @@ After modifying .ts files under plugins/, MUST run rm -rf /tmp/jiti/ BEFORE open
-
-데이터베이스 스키마 +
데이터베이스 스키마 LanceDB 테이블 `memories`: @@ -690,8 +670,7 @@ v1.1.0의 주요 `metadata` 키: `l0_abstract`, `l1_overview`, `l2_content`, `me
-
-문제 해결 +
문제 해결 ### "Cannot mix BigInt and other types" (LanceDB / Apache Arrow) @@ -738,29 +717,17 @@ LanceDB 0.26 이상에서 일부 숫자 열이 `BigInt`로 반환될 수 있습 ## Contributors -

-@win4r -@kctony -@Akatsuki-Ryu -@JasonSuz -@Minidoracat -@furedericca-lab -@joe2643 -@AliceLJY -@chenjiyong -

+

@win4r @kctony @Akatsuki-Ryu @JasonSuz @Minidoracat @furedericca-lab @joe2643 @AliceLJY @chenjiyong

Full list: [Contributors](https://github.com/CortexReach/memory-lancedb-pro/graphs/contributors) ## Star History - - + Star History Chart - - + ## 라이선스 diff --git a/platform/plugins/memory-lancedb-pro/README_PT-BR.md b/platform/plugins/memory-lancedb-pro/README_PT-BR.md index 65d721f8..c7a03749 100644 --- a/platform/plugins/memory-lancedb-pro/README_PT-BR.md +++ b/platform/plugins/memory-lancedb-pro/README_PT-BR.md @@ -8,10 +8,7 @@ Um plugin de memória de longo prazo para OpenClaw baseado em LanceDB que armazena preferências, decisões e contexto de projetos, e os recupera automaticamente em sessões futuras. -[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) -[![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) -[![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) -[![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) +[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) [![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) [![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) [![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) [English](README.md) | [简体中文](README_CN.md) | [繁體中文](README_TW.md) | [日本語](README_JA.md) | [한국어](README_KO.md) | [Français](README_FR.md) | [Español](README_ES.md) | [Deutsch](README_DE.md) | [Italiano](README_IT.md) | [Русский](README_RU.md) | [Português (Brasil)](README_PT-BR.md) @@ -134,8 +131,7 @@ Você deve ver: Pronto! Seu agente agora tem memória de longo prazo. -
-Mais caminhos de instalação (usuários existentes, upgrades) +
Mais caminhos de instalação (usuários existentes, upgrades) **Já está usando OpenClaw?** @@ -160,8 +156,7 @@ Veja `CHANGELOG-v1.1.0.md` para mudanças de comportamento e justificativa de up
-
-Importação rápida via Telegram Bot (clique para expandir) +
Importação rápida via Telegram Bot (clique para expandir) Se você está usando a integração Telegram do OpenClaw, a maneira mais fácil é enviar um comando de importação diretamente para o Bot principal em vez de editar a configuração manualmente. @@ -244,11 +239,9 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > Guia completo: instalação, configuração e funcionamento interno da busca híbrida. -[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) -**https://youtu.be/MtukF1C8epQ** +[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) **https://youtu.be/MtukF1C8epQ** -[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) -**https://www.bilibili.com/video/BV1zUf2BGEgn/** +[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) **https://www.bilibili.com/video/BV1zUf2BGEgn/** --- @@ -278,8 +271,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > Para um mergulho profundo na arquitetura completa, veja [docs/memory_architecture_analysis.md](docs/memory_architecture_analysis.md). -
-Referência de arquivos (clique para expandir) +
Referência de arquivos (clique para expandir) | Arquivo | Finalidade | | --- | --- | @@ -369,8 +361,7 @@ Query → BM25 FTS ─────┘ --- -
-Comparação com o memory-lancedb integrado (clique para expandir) +
Comparação com o memory-lancedb integrado (clique para expandir) | Funcionalidade | `memory-lancedb` integrado | **memory-lancedb-pro** | | --- | :---: | :---: | @@ -397,8 +388,7 @@ Query → BM25 FTS ─────┘ ## Configuração -
-Exemplo de configuração completa +
Exemplo de configuração completa ```json { @@ -462,8 +452,7 @@ Query → BM25 FTS ─────┘
-
-Provedores de Embedding +
Provedores de Embedding Funciona com **qualquer API de embedding compatível com OpenAI**: @@ -477,8 +466,7 @@ Funciona com **qualquer API de embedding compatível com OpenAI**:
-
-Provedores de Rerank +
Provedores de Rerank O reranking cross-encoder suporta múltiplos provedores via `rerankProvider`: @@ -493,8 +481,7 @@ Qualquer endpoint de rerank compatível com Jina também funciona — defina `re
-
-Extração inteligente (LLM) — v1.1.0 +
Extração inteligente (LLM) — v1.1.0 Quando `smartExtraction` está habilitado (padrão: `true`), o plugin usa um LLM para extrair e classificar memórias de forma inteligente em vez de gatilhos baseados em regex. @@ -535,8 +522,7 @@ Notas para `llm.auth: "oauth"`:
-
-Configuração do ciclo de vida (Decaimento + Nível) +
Configuração do ciclo de vida (Decaimento + Nível) | Campo | Padrão | Descrição | |-------|--------|-----------| @@ -551,8 +537,7 @@ Notas para `llm.auth: "oauth"`:
-
-Reforço por acesso +
Reforço por acesso Memórias recuperadas com frequência decaem mais lentamente (estilo repetição espaçada). @@ -594,8 +579,7 @@ Fluxo de login OAuth: ## Tópicos avançados -
-Se memórias injetadas aparecem nas respostas +
Se memórias injetadas aparecem nas respostas Às vezes o modelo pode ecoar o bloco `` injetado. @@ -609,8 +593,7 @@ Fluxo de login OAuth:
-
-Memória de sessão +
Memória de sessão - Acionada no comando `/new` — salva o resumo da sessão anterior no LanceDB - Desabilitada por padrão (OpenClaw já tem persistência nativa de sessão via `.jsonl`) @@ -620,8 +603,7 @@ Veja [docs/openclaw-integration-playbook.md](docs/openclaw-integration-playbook.
-
-Comandos slash personalizados (ex.: /lesson) +
Comandos slash personalizados (ex.: /lesson) Adicione ao seu `CLAUDE.md`, `AGENTS.md` ou prompt do sistema: @@ -640,8 +622,7 @@ When the user sends `/remember `:
-
-Regras de ferro para agentes de IA +
Regras de ferro para agentes de IA > Copie o bloco abaixo no seu `AGENTS.md` para que seu agente aplique essas regras automaticamente. @@ -668,8 +649,7 @@ After modifying .ts files under plugins/, MUST run rm -rf /tmp/jiti/ BEFORE open
-
-Schema do banco de dados +
Schema do banco de dados Tabela LanceDB `memories`: @@ -690,8 +670,7 @@ Chaves `metadata` comuns no v1.1.0: `l0_abstract`, `l1_overview`, `l2_content`,
-
-Solução de problemas +
Solução de problemas ### "Cannot mix BigInt and other types" (LanceDB / Apache Arrow) @@ -738,29 +717,17 @@ Feedback: [GitHub Issues](https://github.com/CortexReach/memory-lancedb-pro/issu ## Contributors -

-@win4r -@kctony -@Akatsuki-Ryu -@JasonSuz -@Minidoracat -@furedericca-lab -@joe2643 -@AliceLJY -@chenjiyong -

+

@win4r @kctony @Akatsuki-Ryu @JasonSuz @Minidoracat @furedericca-lab @joe2643 @AliceLJY @chenjiyong

Full list: [Contributors](https://github.com/CortexReach/memory-lancedb-pro/graphs/contributors) ## Star History - - + Star History Chart - - + ## Licença diff --git a/platform/plugins/memory-lancedb-pro/README_RU.md b/platform/plugins/memory-lancedb-pro/README_RU.md index 8fcb1031..c8a89574 100644 --- a/platform/plugins/memory-lancedb-pro/README_RU.md +++ b/platform/plugins/memory-lancedb-pro/README_RU.md @@ -8,10 +8,7 @@ Плагин долгосрочной памяти для OpenClaw на базе LanceDB, который сохраняет предпочтения, решения и контекст проекта, а затем автоматически вспоминает их в будущих сессиях. -[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) -[![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) -[![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) -[![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) +[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) [![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) [![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) [![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) [English](README.md) | [简体中文](README_CN.md) | [繁體中文](README_TW.md) | [日本語](README_JA.md) | [한국어](README_KO.md) | [Français](README_FR.md) | [Español](README_ES.md) | [Deutsch](README_DE.md) | [Italiano](README_IT.md) | [Русский](README_RU.md) | [Português (Brasil)](README_PT-BR.md) @@ -134,8 +131,7 @@ openclaw logs --follow --plain | grep "memory-lancedb-pro" Готово. Теперь у вашего агента есть долгосрочная память. -
-Дополнительные варианты установки (для действующих пользователей и апгрейдов) +
Дополнительные варианты установки (для действующих пользователей и апгрейдов) **Уже используете OpenClaw?** @@ -160,8 +156,7 @@ openclaw memory-pro stats
-
-Быстрый импорт для Telegram Bot (нажмите, чтобы раскрыть) +
Быстрый импорт для Telegram Bot (нажмите, чтобы раскрыть) Если вы используете Telegram-интеграцию OpenClaw, самый простой путь — отправить команду импорта прямо основному боту вместо ручного редактирования конфига. @@ -244,11 +239,9 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > Полный разбор: установка, настройка и внутреннее устройство гибридного поиска. -[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) -**https://youtu.be/MtukF1C8epQ** +[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) **https://youtu.be/MtukF1C8epQ** -[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) -**https://www.bilibili.com/video/BV1zUf2BGEgn/** +[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) **https://www.bilibili.com/video/BV1zUf2BGEgn/** --- @@ -278,8 +271,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > Для глубокого разбора полной архитектуры смотрите [docs/memory_architecture_analysis.md](docs/memory_architecture_analysis.md). -
-Справочник по файлам (нажмите, чтобы раскрыть) +
Справочник по файлам (нажмите, чтобы раскрыть) | Файл | Назначение | | --- | --- | @@ -369,8 +361,7 @@ Query → BM25 FTS ─────┘ --- -
-Сравнение со встроенным memory-lancedb (нажмите, чтобы раскрыть) +
Сравнение со встроенным memory-lancedb (нажмите, чтобы раскрыть) | Возможность | Встроенный `memory-lancedb` | **memory-lancedb-pro** | | --- | :---: | :---: | @@ -397,8 +388,7 @@ Query → BM25 FTS ─────┘ ## Конфигурация -
-Полный пример конфигурации +
Полный пример конфигурации ```json { @@ -462,8 +452,7 @@ Query → BM25 FTS ─────┘
-
-Провайдеры эмбеддингов +
Провайдеры эмбеддингов Работает с **любым OpenAI-compatible API для эмбеддингов**: @@ -477,8 +466,7 @@ Query → BM25 FTS ─────┘
-
-Провайдеры реранжирования +
Провайдеры реранжирования Кросс-энкодерное реранжирование поддерживает несколько провайдеров через `rerankProvider`: @@ -493,8 +481,7 @@ Query → BM25 FTS ─────┘
-
-Smart Extraction (LLM) — v1.1.0 +
Smart Extraction (LLM) — v1.1.0 Когда включен `smartExtraction` (по умолчанию: `true`), плагин использует LLM для интеллектуального извлечения и классификации воспоминаний вместо правил на регулярных выражениях. @@ -535,8 +522,7 @@ OAuth `llm` config (использует существующий кэш лог
-
-Конфигурация жизненного цикла (Decay + Tier) +
Конфигурация жизненного цикла (Decay + Tier) | Поле | По умолчанию | Описание | |-------|---------|-------------| @@ -551,8 +537,7 @@ OAuth `llm` config (использует существующий кэш лог
-
-Усиление за счет доступа +
Усиление за счет доступа Часто вспоминаемые записи затухают медленнее (в духе spaced repetition). @@ -594,8 +579,7 @@ openclaw memory-pro migrate check|run|verify [--source /path] ## Продвинутые темы -
-Если внедренные воспоминания попадают в ответы +
Если внедренные воспоминания попадают в ответы Иногда модель может дословно повторять внедренный блок ``. @@ -609,8 +593,7 @@ openclaw memory-pro migrate check|run|verify [--source /path]
-
-Память сессии +
Память сессии - Срабатывает по команде `/new` — сохраняет сводку предыдущей сессии в LanceDB - По умолчанию отключено (в OpenClaw уже есть встроенная `.jsonl`-персистентность сессий) @@ -620,8 +603,7 @@ openclaw memory-pro migrate check|run|verify [--source /path]
-
-Пользовательские slash-команды (например, /lesson) +
Пользовательские slash-команды (например, /lesson) Добавьте в `CLAUDE.md`, `AGENTS.md` или system prompt: @@ -640,8 +622,7 @@ openclaw memory-pro migrate check|run|verify [--source /path]
-
-Железные правила для ИИ-агентов +
Железные правила для ИИ-агентов > Скопируйте блок ниже в `AGENTS.md`, чтобы агент автоматически соблюдал эти правила. @@ -668,8 +649,7 @@ openclaw memory-pro migrate check|run|verify [--source /path]
-
-Схема базы данных +
Схема базы данных Таблица LanceDB `memories`: @@ -690,8 +670,7 @@ openclaw memory-pro migrate check|run|verify [--source /path]
-
-Устранение неполадок +
Устранение неполадок ### "Cannot mix BigInt and other types" (LanceDB / Apache Arrow) @@ -738,29 +717,17 @@ openclaw memory-pro migrate check|run|verify [--source /path] ## Участники -

-@win4r -@kctony -@Akatsuki-Ryu -@JasonSuz -@Minidoracat -@furedericca-lab -@joe2643 -@AliceLJY -@chenjiyong -

+

@win4r @kctony @Akatsuki-Ryu @JasonSuz @Minidoracat @furedericca-lab @joe2643 @AliceLJY @chenjiyong

Полный список: [Участники](https://github.com/CortexReach/memory-lancedb-pro/graphs/contributors) ## История звезд - - + Star History Chart - - + ## Лицензия diff --git a/platform/plugins/memory-lancedb-pro/README_TW.md b/platform/plugins/memory-lancedb-pro/README_TW.md index b582d40f..0e29fb36 100644 --- a/platform/plugins/memory-lancedb-pro/README_TW.md +++ b/platform/plugins/memory-lancedb-pro/README_TW.md @@ -8,10 +8,7 @@ 基於 LanceDB 的 OpenClaw 長期記憶外掛，自動儲存偏好、決策和專案上下文，在後續工作階段中自動回憶。 -[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) -[![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) -[![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) -[![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) +[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) [![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) [![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) [![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) [English](README.md) | [简体中文](README_CN.md) | [繁體中文](README_TW.md) | [日本語](README_JA.md) | [한국어](README_KO.md) | [Français](README_FR.md) | [Español](README_ES.md) | [Deutsch](README_DE.md) | [Italiano](README_IT.md) | [Русский](README_RU.md) | [Português (Brasil)](README_PT-BR.md) @@ -134,8 +131,7 @@ openclaw logs --follow --plain | grep "memory-lancedb-pro" 完成！你的智慧體現在擁有長期記憶了。 -
-更多安裝路徑（現有使用者、升級） +
更多安裝路徑（現有使用者、升級） **已在使用 OpenClaw？** @@ -160,8 +156,7 @@ openclaw memory-pro stats
-
-Telegram Bot 快速匯入（點選展開） +
Telegram Bot 快速匯入（點選展開） 如果你在使用 OpenClaw 的 Telegram 整合，最簡單的方式是直接給主 Bot 發訊息，而不是手動編輯設定檔。 @@ -244,11 +239,9 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > 完整演示：安裝、設定、混合檢索內部原理。 -[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) -**https://youtu.be/MtukF1C8epQ** +[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) **https://youtu.be/MtukF1C8epQ** -[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) -**https://www.bilibili.com/video/BV1zUf2BGEgn/** +[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) **https://www.bilibili.com/video/BV1zUf2BGEgn/** --- @@ -278,8 +271,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > 完整架構解析見 [docs/memory_architecture_analysis.md](docs/memory_architecture_analysis.md)。 -
-檔案說明（點選展開） +
檔案說明（點選展開） | 檔案 | 用途 | | --- | --- | @@ -369,8 +361,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla --- -
-與內建 memory-lancedb 的對比（點選展開） +
與內建 memory-lancedb 的對比（點選展開） | 功能 | 內建 `memory-lancedb` | **memory-lancedb-pro** | | --- | :---: | :---: | @@ -397,8 +388,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla ## 設定 -
-完整設定範例 +
完整設定範例 ```json { @@ -462,8 +452,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla
-
-Embedding 服務商 +
Embedding 服務商 相容 **任意 OpenAI 相容 Embedding API**： @@ -477,8 +466,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla
-
-重排序服務商 +
重排序服務商 交叉編碼器重排序透過 `rerankProvider` 支援多個服務商： @@ -493,8 +481,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla
-
-智慧擷取（LLM）— v1.1.0 +
智慧擷取（LLM）— v1.1.0 當 `smartExtraction` 啟用（預設 `true`）時，外掛使用 LLM 智慧擷取和分類記憶，取代基於正則的觸發方式。 @@ -535,8 +522,7 @@ OAuth `llm` 設定（使用現有 Codex / ChatGPT 登入快取來發送 LLM 請
-
-生命週期設定（衰減 + 層級） +
生命週期設定（衰減 + 層級） | 欄位 | 預設值 | 說明 | |------|--------|------| @@ -551,8 +537,7 @@ OAuth `llm` 設定（使用現有 Codex / ChatGPT 登入快取來發送 LLM 請
-
-存取強化 +
存取強化 頻繁被召回的記憶衰減更慢（類似間隔重複機制）。 @@ -594,8 +579,7 @@ OAuth 登入流程： ## 進階主題 -
-注入的記憶出現在回覆中 +
注入的記憶出現在回覆中 有時模型可能會將注入的 `` 區塊原文輸出。 @@ -609,8 +593,7 @@ OAuth 登入流程：
-
-工作階段記憶 +
工作階段記憶 - 透過 `/new` 指令觸發——將上一段工作階段摘要儲存到 LanceDB - 預設關閉（OpenClaw 已有原生 `.jsonl` 工作階段持久化） @@ -620,8 +603,7 @@ OAuth 登入流程：
-
-自訂斜線指令（如 /lesson） +
自訂斜線指令（如 /lesson） 在你的 `CLAUDE.md`、`AGENTS.md` 或系統提示詞中新增： @@ -640,8 +622,7 @@ OAuth 登入流程：
-
-AI 智慧體鐵律 +
AI 智慧體鐵律 > 將以下內容複製到你的 `AGENTS.md`，讓智慧體自動遵守這些規則。 @@ -668,8 +649,7 @@ OAuth 登入流程：
-
-資料庫 Schema +
資料庫 Schema LanceDB 表 `memories`： @@ -690,8 +670,7 @@ v1.1.0 常用 `metadata` 欄位：`l0_abstract`、`l1_overview`、`l2_content`
-
-故障排除 +
故障排除 ### "Cannot mix BigInt and other types"（LanceDB / Apache Arrow） @@ -738,29 +717,17 @@ v1.1.0 常用 `metadata` 欄位：`l0_abstract`、`l1_overview`、`l2_content` ## Contributors -

-@win4r -@kctony -@Akatsuki-Ryu -@JasonSuz -@Minidoracat -@furedericca-lab -@joe2643 -@AliceLJY -@chenjiyong -

+

@win4r @kctony @Akatsuki-Ryu @JasonSuz @Minidoracat @furedericca-lab @joe2643 @AliceLJY @chenjiyong

Full list: [Contributors](https://github.com/CortexReach/memory-lancedb-pro/graphs/contributors) ## Star History - - + Star History Chart - - + ## 授權條款 diff --git a/platform/plugins/memory-lancedb-pro/STRONGCLAW_VENDOR.md b/platform/plugins/memory-lancedb-pro/STRONGCLAW_VENDOR.md index 4018e34e..f3f07247 100644 --- a/platform/plugins/memory-lancedb-pro/STRONGCLAW_VENDOR.md +++ b/platform/plugins/memory-lancedb-pro/STRONGCLAW_VENDOR.md @@ -17,11 +17,9 @@ Strongclaw does not enable this plugin by default. The shipped local overlays ke This avoids the upstream `command:new` / `command:reset` typed-hook incompatibility tracked in upstream issue `#191` when running on OpenClaw `2026.3.13`. - Strongclaw auto-detects `darwin/x86_64` and installs the newest stable - compatible LanceDB fallback, `@lancedb/lancedb@0.22.3`, because LanceDB - `0.26.2` publishes Apple binaries for `darwin-arm64` but not `darwin-x64`. +compatible LanceDB fallback, `@lancedb/lancedb@0.22.3`, because LanceDB `0.26.2` publishes Apple binaries for `darwin-arm64` but not `darwin-x64`. - The shared verification path still runs in GitHub Actions and through - `scripts/ci/verify_vendored_memory_plugin.sh`, but Intel macOS hosts now use - the same compatibility matrix locally instead of hard-failing. +`scripts/ci/verify_vendored_memory_plugin.sh`, but Intel macOS hosts now use the same compatibility matrix locally instead of hard-failing. ## Review notes diff --git a/platform/plugins/memory-lancedb-pro/docs/memory_architecture_analysis.md b/platform/plugins/memory-lancedb-pro/docs/memory_architecture_analysis.md index 26a636a0..d9289f47 100644 --- a/platform/plugins/memory-lancedb-pro/docs/memory_architecture_analysis.md +++ b/platform/plugins/memory-lancedb-pro/docs/memory_architecture_analysis.md @@ -615,8 +615,7 @@ graph LR ### 9.4 session-summary 的特殊处理 -`sessionMemory` 写入的 memory 会带 `metadata.type = "session-summary"`。 -在 auto-recall 生命周期维护中，这类 memory 会被排除在 tier 评估之外，避免会话摘要挤占核心长期记忆层级。 +`sessionMemory` 写入的 memory 会带 `metadata.type = "session-summary"`。 在 auto-recall 生命周期维护中，这类 memory 会被排除在 tier 评估之外，避免会话摘要挤占核心长期记忆层级。 --- diff --git a/platform/plugins/memory-lancedb-pro/examples/new-session-distill/hook/enqueue-lesson-extract/HOOK.md b/platform/plugins/memory-lancedb-pro/examples/new-session-distill/hook/enqueue-lesson-extract/HOOK.md index 8b994981..dcdf6831 100644 --- a/platform/plugins/memory-lancedb-pro/examples/new-session-distill/hook/enqueue-lesson-extract/HOOK.md +++ b/platform/plugins/memory-lancedb-pro/examples/new-session-distill/hook/enqueue-lesson-extract/HOOK.md @@ -1,14 +1,9 @@ ---- -name: enqueue-lesson-extract -description: "Enqueue a lesson-extraction task on /new (async Map-Reduce → LanceDB Pro)" -metadata: - { +--- name: enqueue-lesson-extract description: "Enqueue a lesson-extraction task on /new (async Map-Reduce → LanceDB Pro)" metadata: { "openclaw": { "emoji": "🧾", "events": ["command:new"] } - } ---- +} --- # Enqueue Lesson Extract Hook diff --git a/platform/plugins/memory-lancedb-pro/skills/lesson/SKILL.md b/platform/plugins/memory-lancedb-pro/skills/lesson/SKILL.md index c3dc1672..40b83d47 100644 --- a/platform/plugins/memory-lancedb-pro/skills/lesson/SKILL.md +++ b/platform/plugins/memory-lancedb-pro/skills/lesson/SKILL.md @@ -1,7 +1,4 @@ ---- -name: lesson -description: Store a lesson learned from the current conversation. Triggered by /lesson command. Use when Master signals that the recent conversation contains a pitfall, fix, or key insight that should be persisted to long-term memory. ---- +--- name: lesson description: Store a lesson learned from the current conversation. Triggered by /lesson command. Use when Master signals that the recent conversation contains a pitfall, fix, or key insight that should be persisted to long-term memory. --- # Lesson Extraction & Storage diff --git a/platform/skills/local/browser-lab-safety/SKILL.md b/platform/skills/local/browser-lab-safety/SKILL.md index f1457f6d..e1bf601b 100644 --- a/platform/skills/local/browser-lab-safety/SKILL.md +++ b/platform/skills/local/browser-lab-safety/SKILL.md @@ -1,4 +1,3 @@ # Browser lab safety -Browser automation is isolated. Never enable it on the main control-plane host. -Use sacrificial accounts, explicit allowlists, and exfiltration tests. +Browser automation is isolated. Never enable it on the main control-plane host. Use sacrificial accounts, explicit allowlists, and exfiltration tests. diff --git a/platform/skills/quarantine/README.md b/platform/skills/quarantine/README.md index 6804aee5..f4da699d 100644 --- a/platform/skills/quarantine/README.md +++ b/platform/skills/quarantine/README.md @@ -2,5 +2,4 @@ All newly downloaded or untrusted skills land here first. -Use: -`clawops skill-scan --source --quarantine platform/skills/quarantine --report ` +Use: `clawops skill-scan --source --quarantine platform/skills/quarantine --report ` diff --git a/platform/workers/acpx/README.md b/platform/workers/acpx/README.md index 3052b7d0..da6245ed 100644 --- a/platform/workers/acpx/README.md +++ b/platform/workers/acpx/README.md @@ -31,10 +31,6 @@ acpx --approve-all --format json --json-strict claude exec 'Review auth boundari acpx --approve-all --format json --json-strict --model claude-sonnet-4-5 claude exec 'Review auth boundaries' ``` -`acpx` resolves config from `~/.acpx/config.json` and `/.acpxrc.json`. Strongclaw's adapter -now passes permission mode, output mode, and backend profile explicitly on the command line so -session summaries record the effective execution contract instead of inheriting implicit local -defaults. +`acpx` resolves config from `~/.acpx/config.json` and `/.acpxrc.json`. Strongclaw's adapter now passes permission mode, output mode, and backend profile explicitly on the command line so session summaries record the effective execution contract instead of inheriting implicit local defaults. -The `clawops devflow` surface consumes these role prompts through the config -catalog in `platform/configs/devflow/roles.yaml`. +The `clawops devflow` surface consumes these role prompts through the config catalog in `platform/configs/devflow/roles.yaml`. diff --git a/platform/workers/browser-lab/README.md b/platform/workers/browser-lab/README.md index 1a9f608b..9d6ae8a9 100644 --- a/platform/workers/browser-lab/README.md +++ b/platform/workers/browser-lab/README.md @@ -12,8 +12,7 @@ Operator access should tunnel only the gateway port: ssh -N -L 18789:127.0.0.1:18789 @ ``` -Do not tunnel `9222` or `3128` to an operator workstation. Keep the browser lab -reachable only from the hardened session that runs OpenClaw. +Do not tunnel `9222` or `3128` to an operator workstation. Keep the browser lab reachable only from the hardened session that runs OpenClaw. Included: - `squid.conf` outbound allowlist proxy diff --git a/platform/workspace/admin/IDENTITY.md b/platform/workspace/admin/IDENTITY.md index eb8d42cc..d0c0fc4b 100644 --- a/platform/workspace/admin/IDENTITY.md +++ b/platform/workspace/admin/IDENTITY.md @@ -3,15 +3,15 @@ _Fill this in during your first conversation. Make it yours._ - **Name:** - _(pick something you like)_ +_(pick something you like)_ - **Creature:** - _(AI? robot? familiar? ghost in the machine? something weirder?)_ +_(AI? robot? familiar? ghost in the machine? something weirder?)_ - **Vibe:** - _(how do you come across? sharp? warm? chaotic? calm?)_ +_(how do you come across? sharp? warm? chaotic? calm?)_ - **Emoji:** - _(your signature — pick one that feels right)_ +_(your signature — pick one that feels right)_ - **Avatar:** - _(workspace-relative path, http(s) URL, or data URI)_ +_(workspace-relative path, http(s) URL, or data URI)_ --- diff --git a/platform/workspace/reader/IDENTITY.md b/platform/workspace/reader/IDENTITY.md index eb8d42cc..d0c0fc4b 100644 --- a/platform/workspace/reader/IDENTITY.md +++ b/platform/workspace/reader/IDENTITY.md @@ -3,15 +3,15 @@ _Fill this in during your first conversation. Make it yours._ - **Name:** - _(pick something you like)_ +_(pick something you like)_ - **Creature:** - _(AI? robot? familiar? ghost in the machine? something weirder?)_ +_(AI? robot? familiar? ghost in the machine? something weirder?)_ - **Vibe:** - _(how do you come across? sharp? warm? chaotic? calm?)_ +_(how do you come across? sharp? warm? chaotic? calm?)_ - **Emoji:** - _(your signature — pick one that feels right)_ +_(your signature — pick one that feels right)_ - **Avatar:** - _(workspace-relative path, http(s) URL, or data URI)_ +_(workspace-relative path, http(s) URL, or data URI)_ --- diff --git a/src/clawops/assets/platform/docs/BACKUP_AND_RECOVERY.md b/src/clawops/assets/platform/docs/BACKUP_AND_RECOVERY.md index 12824ec1..77e9f70c 100644 --- a/src/clawops/assets/platform/docs/BACKUP_AND_RECOVERY.md +++ b/src/clawops/assets/platform/docs/BACKUP_AND_RECOVERY.md @@ -17,15 +17,12 @@ ## Development-mode repo-local compose state -If you keep compose state under `platform/compose/state` during development, use -the explicit dev wrappers instead of relying on implicit leftover mounts: +If you keep compose state under `platform/compose/state` during development, use the explicit dev wrappers instead of relying on implicit leftover mounts: - `clawops ops sidecars up --repo-local-state` - `clawops ops sidecars down --repo-local-state` -`clawops ops sidecars up` owns the LiteLLM schema bootstrap phase. Bring the -stack up through the CLI entrypoint instead of raw `docker compose up` when you -need the supported startup ordering on a cold Postgres state directory. +`clawops ops sidecars up` owns the LiteLLM schema bootstrap phase. Bring the stack up through the CLI entrypoint instead of raw `docker compose up` when you need the supported startup ordering on a cold Postgres state directory. Prefer targeted cleanup over deleting the whole tree: diff --git a/src/clawops/assets/platform/docs/CI_AND_SECURITY.md b/src/clawops/assets/platform/docs/CI_AND_SECURITY.md index 79e1bf19..d1e88d40 100644 --- a/src/clawops/assets/platform/docs/CI_AND_SECURITY.md +++ b/src/clawops/assets/platform/docs/CI_AND_SECURITY.md @@ -18,73 +18,49 @@ The repository includes: Workflow policy: - GitHub Actions workflows stay thin. Multi-step operational logic lives in - semantic helper entrypoints under `tests/scripts/`, with unit coverage in - `tests/suites/unit/ci/` and repo contract coverage under - `tests/suites/contracts/repo/`. +semantic helper entrypoints under `tests/scripts/`, with unit coverage in `tests/suites/unit/ci/` and repo contract coverage under `tests/suites/contracts/repo/`. ## Fresh-host acceptance -`.github/workflows/fresh-host-acceptance.yml` exercises the real bootstrap, -setup, service activation, and repo-local sidecar/browser-lab flows on hosted -Linux and macOS runners. +`.github/workflows/fresh-host-acceptance.yml` exercises the real bootstrap, setup, service activation, and repo-local sidecar/browser-lab flows on hosted Linux and macOS runners. - Each run writes a GitHub job summary with the runner label, runtime provider, - cache toggles, phase timings, and the effective hosted macOS Colima sizing. +cache toggles, phase timings, and the effective hosted macOS Colima sizing. - Each run uploads a `fresh-host-reports` artifact subtree with runtime - diagnostics (`docker info`, image inventory, launchd state, and runtime - status output) alongside the rendered host artifacts. +diagnostics (`docker info`, image inventory, launchd state, and runtime status output) alongside the rendered host artifacts. - Hosted macOS acceptance is pinned to `macos-15-intel`. GitHub's standard - `macos-15` arm64 runners are available on public repositories, but GitHub - documents that nested virtualization is not supported on arm64 macOS hosted - runners, so Colima/OrbStack cannot provide a Docker backend there. +`macos-15` arm64 runners are available on public repositories, but GitHub documents that nested virtualization is not supported on arm64 macOS hosted runners, so Colima/OrbStack cannot provide a Docker backend there. - The hosted macOS job installs Lima and Colima directly, then sizes Colima for - the runner instead of using the old fixed `2 CPU / 4 GiB` VM. +the runner instead of using the old fixed `2 CPU / 4 GiB` VM. - Hosted macOS acceptance uses the `ci-hosted-macos` compose variant so - sidecars and browser-lab mutable data live in Docker-managed volumes instead - of FUSE-backed host bind mounts. That avoids the hosted-Colima filesystem - regressions seen with Qdrant and Postgres while preserving the real `clawops` - setup, launchd activation, and repo-local stack flows. +sidecars and browser-lab mutable data live in Docker-managed volumes instead of FUSE-backed host bind mounts. That avoids the hosted-Colima filesystem regressions seen with Qdrant and Postgres while preserving the real `clawops` setup, launchd activation, and repo-local stack flows. - `workflow_dispatch` can benchmark cache toggles for the supported hosted - macOS path without changing the required PR gate. +macOS path without changing the required PR gate. - The workflow stays declarative by delegating runtime setup, image warming, - diagnostics, and summary generation to executable helper scripts under - `tests/scripts/`. Hosted macOS image warming still resolves the pinned compose - images ahead of the scenario run, but now uses bounded retries and heartbeat - logging instead of restoring immutable Docker image tarballs from the - workflow cache. +diagnostics, and summary generation to executable helper scripts under `tests/scripts/`. Hosted macOS image warming still resolves the pinned compose images ahead of the scenario run, but now uses bounded retries and heartbeat logging instead of restoring immutable Docker image tarballs from the workflow cache. - Repository workflow contract tests verify that shell steps invoking - `tests/scripts/*.py` either call an explicit Python interpreter or target an - executable script, so nightly cache warming cannot silently regress on file - mode drift. +`tests/scripts/*.py` either call an explicit Python interpreter or target an executable script, so nightly cache warming cannot silently regress on file mode drift. ## Vendored plugin verification -The vendored `platform/plugins/memory-lancedb-pro` bundle is verified on -GitHub Actions in `.github/workflows/memory-plugin-verification.yml`. +The vendored `platform/plugins/memory-lancedb-pro` bundle is verified on GitHub Actions in `.github/workflows/memory-plugin-verification.yml`. - The shared entrypoint is `the vendored-memory plugin verification workflow`. - That flow reuses `clawops config memory --set-profile memory-lancedb-pro`, which - auto-detects the host and installs the default LanceDB dependency on - supported hosts or the Intel-macOS fallback `@lancedb/lancedb@0.22.3`. +auto-detects the host and installs the default LanceDB dependency on supported hosts or the Intel-macOS fallback `@lancedb/lancedb@0.22.3`. - The workflow delegates the host-functional orchestration to - `tests/scripts/memory_plugin_verification.py`, which installs the pinned - `openclaw@2026.3.13` CLI into a - temporary tool directory and runs the host-functional - `npm run test:openclaw-host` suite. +`tests/scripts/memory_plugin_verification.py`, which installs the pinned `openclaw@2026.3.13` CLI into a temporary tool directory and runs the host-functional `npm run test:openclaw-host` suite. - The host-functional step clears ambient AWS credential env vars first so - local Bedrock model discovery noise does not contaminate test assertions. +local Bedrock model discovery noise does not contaminate test assertions. ## strongclaw-hypermemory host verification -The repo-local `platform/plugins/strongclaw-hypermemory` bundle is also verified -in `.github/workflows/memory-plugin-verification.yml`. +The repo-local `platform/plugins/strongclaw-hypermemory` bundle is also verified in `.github/workflows/memory-plugin-verification.yml`. - The shared entrypoint is `the strongclaw-hypermemory verification workflow`. - That flow runs `npm run test:openclaw-host` inside the plugin bundle. - The host-functional test creates a temporary sqlite-backed `hypermemory` - config, registers the plugin through the local SDK stub, verifies the - exported `memory` CLI surface and subcommands, and exercises the - strongclaw-owned `memory_search` and `memory_get` tool paths. +config, registers the plugin through the local SDK stub, verifies the exported `memory` CLI surface and subcommands, and exercises the strongclaw-owned `memory_search` and `memory_get` tool paths. ## Policy for new code @@ -96,40 +72,24 @@ in `.github/workflows/memory-plugin-verification.yml`. ## Dependency and release provenance - `.github/workflows/dependency-submission.yml` generates `sbom.spdx.json` with - `anchore/sbom-action` and submits the resulting dependency snapshot to the - GitHub dependency graph. +`anchore/sbom-action` and submits the resulting dependency snapshot to the GitHub dependency graph. - `.github/workflows/security.yml`, - `.github/workflows/upstream-merge-validation.yml`, and - `.github/workflows/release.yml` all call the centralized - `clawops supply-chain quality-gate` surface so linting, typing, tests, - coverage, and compile checks stay aligned. +`.github/workflows/upstream-merge-validation.yml`, and `.github/workflows/release.yml` all call the centralized `clawops supply-chain quality-gate` surface so linting, typing, tests, coverage, and compile checks stay aligned. - The compatibility matrix, memory-plugin verification, security, and release - workflows delegate their nontrivial operational steps to `tests/scripts/` - helper CLIs instead of embedding shell blobs or Python heredocs directly in - YAML. +workflows delegate their nontrivial operational steps to `tests/scripts/` helper CLIs instead of embedding shell blobs or Python heredocs directly in YAML. - Those Ubuntu quality-gate workflows install the distro `shellcheck` binary - before invoking the shared gate, and the repo's `pre-commit` hook now uses - that system binary instead of a Docker-backed hook. +before invoking the shared gate, and the repo's `pre-commit` hook now uses that system binary instead of a Docker-backed hook. - `.github/workflows/security.yml` installs a pinned `semgrep` CLI directly - instead of relying on the Docker-backed Semgrep action, which keeps the lane - off Docker Hub. +instead of relying on the Docker-backed Semgrep action, which keeps the lane off Docker Hub. - `.github/workflows/security.yml` verifies the pinned `gitleaks` and `syft` - tarball SHA-256 digests before extracting the binaries through the dedicated - helper script. +tarball SHA-256 digests before extracting the binaries through the dedicated helper script. - `.github/workflows/release.yml` syncs the locked `uv` dev environment, builds - the Python sdist/wheel only after the repository quality gate passes, verifies - each artifact with `twine check` plus fresh install smoke tests through the - dedicated release helper script, publishes or updates the GitHub release with - `gh`, and emits GitHub attestations for both build provenance and the - generated SBOM. +the Python sdist/wheel only after the repository quality gate passes, verifies each artifact with `twine check` plus fresh install smoke tests through the dedicated release helper script, publishes or updates the GitHub release with `gh`, and emits GitHub attestations for both build provenance and the generated SBOM. - `.github/workflows/upstream-merge-validation.yml` runs the repo quality gate - plus nightly validation steps after an upstream merge lands in the fork. +plus nightly validation steps after an upstream merge lands in the fork. - `.github/workflows/memory-plugin-verification.yml` runs the dedicated - hypermemory Qdrant checks against the official pinned Qdrant GHCR image - instead of Docker Hub. +hypermemory Qdrant checks against the official pinned Qdrant GHCR image instead of Docker Hub. - `.github/workflows/devflow-contract.yml` syncs the locked environment, - compile-checks the repo, runs targeted devflow tests, and validates - `clawops devflow plan --goal "contract smoke"` without live ACP - providers. +compile-checks the repo, runs targeted devflow tests, and validates `clawops devflow plan --goal "contract smoke"` without live ACP providers. - Operators can verify published provenance with GitHub's attestation tooling - after a tagged release lands. +after a tagged release lands. diff --git a/src/clawops/assets/platform/docs/CONTEXT_SERVICE.md b/src/clawops/assets/platform/docs/CONTEXT_SERVICE.md index ebf7f44b..d2ae8f4e 100644 --- a/src/clawops/assets/platform/docs/CONTEXT_SERVICE.md +++ b/src/clawops/assets/platform/docs/CONTEXT_SERVICE.md @@ -1,8 +1,6 @@ # Codebase Context Provider -This repo exposes codebase context under the generic `clawops context` namespace. -The first explicit provider is `codebase`, which keeps the SQLite lexical baseline -and adds scale-aware chunk, hybrid, and graph state. +This repo exposes codebase context under the generic `clawops context` namespace. The first explicit provider is `codebase`, which keeps the SQLite lexical baseline and adds scale-aware chunk, hybrid, and graph state. ## What it does @@ -33,8 +31,7 @@ clawops context codebase worker --scale medium --config platform/configs/context clawops context codebase benchmark --scale medium --config platform/configs/context/codebase.yaml --repo . --fixtures platform/configs/context/benchmarks/codebase.yaml --json ``` -The shipped config keeps default indexing focused on repo-authored material by -excluding large vendored trees and the benchmark fixture directory. +The shipped config keeps default indexing focused on repo-authored material by excluding large vendored trees and the benchmark fixture directory. ## Included integrations @@ -45,13 +42,11 @@ excluding large vendored trees and the benchmark fixture directory. ## Default memory retrieval -The default StrongClaw render path is `hypermemory`, which uses -`lossless-claw` plus `strongclaw-hypermemory`. +The default StrongClaw render path is `hypermemory`, which uses `lossless-claw` plus `strongclaw-hypermemory`. The explicit `openclaw-default` fallback profile keeps the OpenClaw built-ins only. -The explicit `openclaw-qmd` fallback profile enables QMD-backed memory -retrieval. +The explicit `openclaw-qmd` fallback profile enables QMD-backed memory retrieval. The rendered QMD corpus for `openclaw-qmd` includes: @@ -123,18 +118,16 @@ The shipped codebase config supports: Path filters are applied to repo-relative POSIX paths before indexing. -The shipped config excludes packaged runtime mirrors under `src/clawops/assets/**` -so retrieval ranks the real source tree instead of duplicated install assets. +The shipped config excludes packaged runtime mirrors under `src/clawops/assets/**` so retrieval ranks the real source tree instead of duplicated install assets. Symlink handling is explicit: - `in_repo_only` follows symlinks only when the resolved target stays inside the - configured repo root +configured repo root - `never` skips all symlinked files - `follow` follows all symlinked files and should only be used intentionally -The default shipped policy is `in_repo_only` to prevent context packs from -pulling host files from outside the repo tree. +The default shipped policy is `in_repo_only` to prevent context packs from pulling host files from outside the repo tree. Reindexing is authoritative for the configured file universe: @@ -153,28 +146,15 @@ Hybrid runtime artifacts are intentionally deferred for medium and large reindex - `worker` reconciles pending Qdrant point deletions and chunk-vector upserts in the background - `query` and `pack` transparently fall back to lexical chunk retrieval until the worker finishes a healthy hybrid sync -Successful hybrid batches now persist incrementally, so a late local embedding -timeout leaves the worker degraded but does not discard already indexed chunk -vectors. The next worker or benchmark pass resumes from the remaining chunks. +Successful hybrid batches now persist incrementally, so a late local embedding timeout leaves the worker degraded but does not discard already indexed chunk vectors. The next worker or benchmark pass resumes from the remaining chunks. -When chunk hashes and pending deletions already match the indexed corpus, the -worker reuses the persisted sparse metadata and skips both sparse rebuild and -dense re-embedding work. On cold syncs, the sparse lane now normalizes each -chunk once and reuses that token stream for both corpus statistics and the -persisted sparse vector payloads. +When chunk hashes and pending deletions already match the indexed corpus, the worker reuses the persisted sparse metadata and skips both sparse rebuild and dense re-embedding work. On cold syncs, the sparse lane now normalizes each chunk once and reuses that token stream for both corpus statistics and the persisted sparse vector payloads. -The Neo4j lane now uses the official Python driver over the Bolt protocol and -keeps the SQLite graph fallback aligned by materializing symbol-aware -`DEFINES`, `CALLS`, and `REFERENCES` edges alongside file import edges. +The Neo4j lane now uses the official Python driver over the Bolt protocol and keeps the SQLite graph fallback aligned by materializing symbol-aware `DEFINES`, `CALLS`, and `REFERENCES` edges alongside file import edges. ## Benchmark fixtures -Use `clawops context codebase benchmark` to measure Recall@k and MRR against a -curated fixture file. The command reindexes the repo, consolidates runtime -artifacts, and then evaluates each case against the final context surface for -the current provider scale: the top retrieved roots plus graph dependency expansion -when that lane is active. When the fixture file lives under the repo -root, the benchmark command excludes that exact file from indexing so the query set cannot self-match. +Use `clawops context codebase benchmark` to measure Recall@k and MRR against a curated fixture file. The command reindexes the repo, consolidates runtime artifacts, and then evaluates each case against the final context surface for the current provider scale: the top retrieved roots plus graph dependency expansion when that lane is active. When the fixture file lives under the repo root, the benchmark command excludes that exact file from indexing so the query set cannot self-match. Fixture files live under `platform/configs/context/benchmarks/` and accept: @@ -193,20 +173,13 @@ Benchmark authoring guidance: - Keep the shipped benchmark CLI example on `--scale medium` when validating hybrid recall expectations. - The shipped local medium config keeps embedding batches conservative and allows longer HTTP timeouts because direct Ollama embedding can still exceed optimistic defaults on singleton retries. -For the shipped local sidecar stack, the codebase provider reads Neo4j -credentials from `NEO4J_USERNAME` and `NEO4J_PASSWORD`. The Varlock env -contract now carries those keys directly, and the compose stack derives the -container's `NEO4J_AUTH` value from the same pair so fresh-host bring-up and -graph-backed retrieval stay aligned. +For the shipped local sidecar stack, the codebase provider reads Neo4j credentials from `NEO4J_USERNAME` and `NEO4J_PASSWORD`. The Varlock env contract now carries those keys directly, and the compose stack derives the container's `NEO4J_AUTH` value from the same pair so fresh-host bring-up and graph-backed retrieval stay aligned. -The shipped local medium lane uses direct Ollama embeddings plus the shared -Qdrant and rerank contract: +The shipped local medium lane uses direct Ollama embeddings plus the shared Qdrant and rerank contract: - `HYPERMEMORY_EMBEDDING_MODEL` with a pulled local Ollama embedding model such as `ollama/nomic-embed-text` - direct Ollama on `http://127.0.0.1:11434` - `HYPERMEMORY_QDRANT_URL` - optional rerank fallback keys `HYPERMEMORY_RERANK_BASE_URL`, `HYPERMEMORY_RERANK_MODEL`, and `HYPERMEMORY_RERANK_API_KEY` -If you want the codebase provider to route embeddings through LiteLLM or a -remote OpenAI-compatible gateway instead, switch `embedding.provider` back to -`compatible-http` and point `embedding.base_url` at that endpoint. +If you want the codebase provider to route embeddings through LiteLLM or a remote OpenAI-compatible gateway instead, switch `embedding.provider` back to `compatible-http` and point `embedding.base_url` at that endpoint. diff --git a/src/clawops/assets/platform/docs/DEVFLOW.md b/src/clawops/assets/platform/docs/DEVFLOW.md index 4326f426..68d58c53 100644 --- a/src/clawops/assets/platform/docs/DEVFLOW.md +++ b/src/clawops/assets/platform/docs/DEVFLOW.md @@ -11,8 +11,7 @@ clawops devflow cancel clawops devflow audit ``` -Use `--project-root ` when the control root for run state, journals, and -audit artifacts is not the current working directory. +Use `--project-root ` when the control root for run state, journals, and audit artifacts is not the current working directory. ## Operator Flow diff --git a/src/clawops/assets/platform/docs/HOST_PLATFORMS.md b/src/clawops/assets/platform/docs/HOST_PLATFORMS.md index a65d3c11..cf1618a5 100644 --- a/src/clawops/assets/platform/docs/HOST_PLATFORMS.md +++ b/src/clawops/assets/platform/docs/HOST_PLATFORMS.md @@ -5,18 +5,13 @@ Strongclaw supports two operator-host platforms: - macOS hosts using Homebrew plus `launchd` - Linux hosts using `apt-get` plus user-level `systemd` -Both use the same bootstrap entrypoints, config overlays, and verification -gates. +Both use the same bootstrap entrypoints, config overlays, and verification gates. -The bootstrap contract is runtime-aware: if the host already has a -Docker-compatible runtime that exposes `docker` plus `docker compose`, -Strongclaw reuses it. Docker is installed only as the fallback runtime when no -compatible backend is detected. +The bootstrap contract is runtime-aware: if the host already has a Docker-compatible runtime that exposes `docker` plus `docker compose`, Strongclaw reuses it. Docker is installed only as the fallback runtime when no compatible backend is detected. ## Compatibility matrix -StrongClaw's supported baseline is derived from the codebase constraints plus -the pinned external tools that setup installs. +StrongClaw's supported baseline is derived from the codebase constraints plus the pinned external tools that setup installs. | Component | Supported / pinned version | Why | | --- | --- | --- | @@ -49,9 +44,7 @@ For low-end or older hosts, this split matters: ## Runtime data locations -StrongClaw-generated runtime artifacts should not live inside the repository -checkout. The setup, doctor, harness, ACP runner, workflow context-pack, and -compose helper commands now default to OS-appropriate app directories instead. +StrongClaw-generated runtime artifacts should not live inside the repository checkout. The setup, doctor, harness, ACP runner, workflow context-pack, and compose helper commands now default to OS-appropriate app directories instead. | Kind | Linux default | macOS default | | --- | --- | --- | @@ -65,10 +58,7 @@ compose helper commands now default to OS-appropriate app directories instead. | QMD package files | `/qmd` | `/qmd` | | `lossless-claw` checkout | `/plugins/lossless-claw` | `/plugins/lossless-claw` | -Use `STRONGCLAW_DATA_DIR`, `STRONGCLAW_STATE_DIR`, `STRONGCLAW_LOG_DIR`, -`STRONGCLAW_RUNS_DIR`, or `STRONGCLAW_COMPOSE_STATE_DIR` when an operator needs -to override those defaults. The Python-owned compose commands export -`STRONGCLAW_COMPOSE_STATE_DIR` automatically before invoking Docker Compose. +Use `STRONGCLAW_DATA_DIR`, `STRONGCLAW_STATE_DIR`, `STRONGCLAW_LOG_DIR`, `STRONGCLAW_RUNS_DIR`, or `STRONGCLAW_COMPOSE_STATE_DIR` when an operator needs to override those defaults. The Python-owned compose commands export `STRONGCLAW_COMPOSE_STATE_DIR` automatically before invoking Docker Compose. ## Shared host contract @@ -89,41 +79,38 @@ Regardless of host OS, the baseline flow is: - Preflight requires Homebrew. - OrbStack, Rancher Desktop, Colima, and Docker Desktop are all acceptable as - long as they expose `docker` plus `docker compose`. +long as they expose `docker` plus `docker compose`. - `clawops ops sidecars up` now phases hosted macOS sidecar startup: Postgres - comes up first, StrongClaw runs LiteLLM Prisma bootstrap as a transient - compose run, and only then does the long-lived LiteLLM proxy start. That - keeps cold-database bootstrap work out of the runtime health window without - encoding init containers into the steady-state compose topology. +comes up first, StrongClaw runs LiteLLM Prisma bootstrap as a transient compose run, and only then does the long-lived LiteLLM proxy start. That keeps cold-database bootstrap work out of the runtime health window without encoding init containers into the steady-state compose topology. - If one of those runtimes is installed but its Docker CLI integration is not - enabled yet, bootstrap stops instead of installing Docker over it. +enabled yet, bootstrap stops instead of installing Docker over it. - Service definitions render into `~/Library/LaunchAgents`. - Activate them with `launchctl bootstrap gui/$(id -u) ...`. - The runtime-user and loopback-SSH flow is documented in - `platform/docs/runbooks/macos-service-user-and-ssh.md`. +`platform/docs/runbooks/macos-service-user-and-ssh.md`. ## Linux host notes - The current Linux bootstrap path targets Debian/Ubuntu-style hosts with - `apt-get`. +`apt-get`. - Existing Docker-compatible runtimes are reused when they already expose - `docker` plus `docker compose` for the runtime user. +`docker` plus `docker compose` for the runtime user. - If no compatible runtime is detected, bootstrap installs Docker Engine as the - fallback backend. +fallback backend. - Provision the runtime user with `your platform-native runtime-user provisioning flow`. - Service definitions render into `~/.config/systemd/user`. - Activate them with `systemctl --user daemon-reload` and - `systemctl --user enable --now ...`. +`systemctl --user enable --now ...`. - Use `loginctl enable-linger ` when user services must survive logout. - Prefer rootless Docker or a tightly controlled `docker` group for the runtime - user. +user. - The runtime-user and user-systemd flow is documented in - `platform/docs/runbooks/linux-runtime-user-and-systemd.md`. +`platform/docs/runbooks/linux-runtime-user-and-systemd.md`. ## Separate-host guidance - Browser lab belongs on a separate host or hardened user session. - ACP workers can run on the same operator host for evaluation, but a separate - worker host is the safer steady-state. +worker host is the safer steady-state. - Langfuse and similar observability extras can live on a separate VM or sidecar - host without changing the control-plane bootstrap contract. +host without changing the control-plane bootstrap contract. diff --git a/src/clawops/assets/platform/docs/HYPERMEMORY.md b/src/clawops/assets/platform/docs/HYPERMEMORY.md index 9bc57036..ba4ae0c2 100644 --- a/src/clawops/assets/platform/docs/HYPERMEMORY.md +++ b/src/clawops/assets/platform/docs/HYPERMEMORY.md @@ -1,7 +1,6 @@ # StrongClaw Hypermemory -`hypermemory` is StrongClaw's Markdown-canonical durable memory engine. It is the default StrongClaw memory stack through the `hypermemory` profile, which -binds: +`hypermemory` is StrongClaw's Markdown-canonical durable memory engine. It is the default StrongClaw memory stack through the `hypermemory` profile, which binds: - `plugins.slots.contextEngine = "lossless-claw"` - `plugins.slots.memory = "strongclaw-hypermemory"` @@ -115,20 +114,9 @@ That flow renders the default StrongClaw stack with `lossless-claw`, `strongclaw The hypermemory env contract requires `HYPERMEMORY_EMBEDDING_MODEL`. Guided setup backfills loopback defaults for `HYPERMEMORY_EMBEDDING_BASE_URL` and `HYPERMEMORY_QDRANT_URL` unless you override them. -The shipped hypermemory configs also enable planner-stage reranking. The -primary provider is `local-sentence-transformers` with -`BAAI/bge-reranker-v2-m3`; the fallback is `compatible-http`, which activates -when `HYPERMEMORY_RERANK_BASE_URL` is configured and reachable. If neither -provider is available, search fails open and keeps the provisional hybrid -planner order. - -Plain `uv sync` keeps the primary local rerank path on host/Python combinations -with known upstream wheel support: macOS arm64, macOS x86_64 on Python 3.12, -and Linux x86_64 or aarch64/arm64 on Python 3.12 or 3.13. For Raspberry Pi, -that means Raspberry Pi 4/5 with 64-bit Raspberry Pi OS or Ubuntu arm64 stay -on the primary local rerank path. Unsupported combinations such as 32-bit Pi -Linux skip the local dependency and use `compatible-http` or fail-open behavior -instead of blocking setup. +The shipped hypermemory configs also enable planner-stage reranking. The primary provider is `local-sentence-transformers` with `BAAI/bge-reranker-v2-m3`; the fallback is `compatible-http`, which activates when `HYPERMEMORY_RERANK_BASE_URL` is configured and reachable. If neither provider is available, search fails open and keeps the provisional hybrid planner order. + +Plain `uv sync` keeps the primary local rerank path on host/Python combinations with known upstream wheel support: macOS arm64, macOS x86_64 on Python 3.12, and Linux x86_64 or aarch64/arm64 on Python 3.12 or 3.13. For Raspberry Pi, that means Raspberry Pi 4/5 with 64-bit Raspberry Pi OS or Ubuntu arm64 stay on the primary local rerank path. Unsupported combinations such as 32-bit Pi Linux skip the local dependency and use `compatible-http` or fail-open behavior instead of blocking setup. Optional fallback env vars: @@ -252,10 +240,4 @@ Corpus path globs are segment-aware: - `**/*.md` matches markdown files recursively below the configured corpus root - if two configured corpus sources resolve to the same repo-relative document, the first configured source wins and later duplicates are skipped during indexing -When `CLAWOPS_STRUCTURED_LOGS=1` is set, hypermemory emits compact JSON lines -for embedding calls, Qdrant search, lexical planning, fusion, rerank, rerank -errors, fallback activation, and vector sync. When OTLP tracing is enabled -through `CLAWOPS_OTEL_ENABLED=1` or the standard `OTEL_EXPORTER_OTLP_*` -variables, the same operations emit spans through the shared ClawOps -observability pipeline, including a dedicated `clawops.hypermemory.rerank` -span. +When `CLAWOPS_STRUCTURED_LOGS=1` is set, hypermemory emits compact JSON lines for embedding calls, Qdrant search, lexical planning, fusion, rerank, rerank errors, fallback activation, and vector sync. When OTLP tracing is enabled through `CLAWOPS_OTEL_ENABLED=1` or the standard `OTEL_EXPORTER_OTLP_*` variables, the same operations emit spans through the shared ClawOps observability pipeline, including a dedicated `clawops.hypermemory.rerank` span. diff --git a/src/clawops/assets/platform/docs/MODEL_ROUTING.md b/src/clawops/assets/platform/docs/MODEL_ROUTING.md index 7bdd00d8..d60aa378 100644 --- a/src/clawops/assets/platform/docs/MODEL_ROUTING.md +++ b/src/clawops/assets/platform/docs/MODEL_ROUTING.md @@ -24,8 +24,7 @@ The supported `hypermemory` profile uses: - `hypermemory-embedding` as the stable LiteLLM alias consumed by `hypermemory` - `HYPERMEMORY_EMBEDDING_BASE_URL` to point `hypermemory` at the loopback LiteLLM route -This keeps the hypermemory memory config pinned to a stable route while letting -operators swap the upstream embedding model behind that alias. +This keeps the hypermemory memory config pinned to a stable route while letting operators swap the upstream embedding model behind that alias. ## Role defaults diff --git a/src/clawops/assets/platform/docs/POLICY_ENGINE_AND_WRAPPERS.md b/src/clawops/assets/platform/docs/POLICY_ENGINE_AND_WRAPPERS.md index c210b02e..6abb85eb 100644 --- a/src/clawops/assets/platform/docs/POLICY_ENGINE_AND_WRAPPERS.md +++ b/src/clawops/assets/platform/docs/POLICY_ENGINE_AND_WRAPPERS.md @@ -60,16 +60,14 @@ Wrapper transport remains intentionally conservative: - requests use a stable `User-Agent` and explicit timeout - terminal success/failure rows persist request method/url, request attempt count, - status code, body excerpt, and a typed transport error code when relevant +status code, body excerpt, and a typed transport error code when relevant - automatic retries are endpoint-scoped, not blanket behavior - current mutating wrappers (`github.comment.create`, - `github.issue.labels.add`, `github.pull_request.merge`, `webhook.post`) stay - on explicit no-retry policies until endpoint-level idempotency is proven +`github.issue.labels.add`, `github.pull_request.merge`, `webhook.post`) stay on explicit no-retry policies until endpoint-level idempotency is proven ## Execution contract -Wrappers now persist an execution contract alongside the stored policy decision. -That contract binds execution to the prepared operation metadata: +Wrappers now persist an execution contract alongside the stored policy decision. That contract binds execution to the prepared operation metadata: - scope - kind @@ -78,11 +76,9 @@ That contract binds execution to the prepared operation metadata: - input hash - policy decision -`clawops op-journal begin` remains available for generic audit/bookkeeping use, -but those rows are not executable wrapper operations by themselves. +`clawops op-journal begin` remains available for generic audit/bookkeeping use, but those rows are not executable wrapper operations by themselves. -This prevents a forged journal row from bypassing policy and allowlist checks at -`--execute-approved` time. +This prevents a forged journal row from bypassing policy and allowlist checks at `--execute-approved` time. ## Journal ownership @@ -139,14 +135,11 @@ clawops wrapper webhook \ Legacy note: - older approved rows that predate execution contracts may require an explicit - policy file on `--execute-approved` so the wrapper can restamp the row before - sending the side effect +policy file on `--execute-approved` so the wrapper can restamp the row before sending the side effect ## Devflow audit spine -The production devflow surface reuses the same SQLite audit spine in -`op_journal.sqlite` and adds devflow-specific run, stage, and stage-event -tables. Operators inspect and recover that workflow state through: +The production devflow surface reuses the same SQLite audit spine in `op_journal.sqlite` and adds devflow-specific run, stage, and stage-event tables. Operators inspect and recover that workflow state through: ```bash clawops devflow status --run-id diff --git a/src/clawops/assets/platform/docs/SECRETS_AND_ENV.md b/src/clawops/assets/platform/docs/SECRETS_AND_ENV.md index deac67aa..b1949548 100644 --- a/src/clawops/assets/platform/docs/SECRETS_AND_ENV.md +++ b/src/clawops/assets/platform/docs/SECRETS_AND_ENV.md @@ -8,9 +8,9 @@ ## Files - managed Varlock env dir: `~/.config/strongclaw/varlock` on Linux, - `~/Library/Application Support/StrongClaw/config/varlock` on macOS +`~/Library/Application Support/StrongClaw/config/varlock` on macOS - source template assets: `platform/configs/varlock/.env.schema`, - `platform/configs/varlock/.env.local.example` +`platform/configs/varlock/.env.local.example` - managed plugin overlay: `.env.plugins` (generated locally when you choose a managed secret backend) - `platform/examples/openclaw-secretref-*.json5` @@ -40,11 +40,9 @@ - StrongClaw keeps core machine-local secrets such as the gateway token and LiteLLM bootstrap secrets in `.env.local` by default. - Managed backends are primarily used for LLM provider credentials and similar integration tokens. - `HYPERMEMORY_EMBEDDING_MODEL` is the only hypermemory env key that - normally needs operator input; the loopback base URLs are backfilled by - guided setup unless you override them. +normally needs operator input; the loopback base URLs are backfilled by guided setup unless you override them. - `.env.plugins` is ignored by git and imported from `.env.schema`, so manual and guided setup can be mixed safely. If you author it by hand, keep explicit plugin version specifiers in each `@plugin(...)` decorator. ## Rotation -Use `clawops recovery rotate-secrets` and the runbook: -`platform/docs/runbooks/credential-rotation.md` +Use `clawops recovery rotate-secrets` and the runbook: `platform/docs/runbooks/credential-rotation.md` diff --git a/src/clawops/assets/platform/docs/SECURITY_MODEL.md b/src/clawops/assets/platform/docs/SECURITY_MODEL.md index f1935e8d..ab13fa86 100644 --- a/src/clawops/assets/platform/docs/SECURITY_MODEL.md +++ b/src/clawops/assets/platform/docs/SECURITY_MODEL.md @@ -28,7 +28,6 @@ Require an independent reviewer for: ## Skill intake -Use: -`clawops skill-scan --source --quarantine --report ` +Use: `clawops skill-scan --source --quarantine --report ` Never auto-enable a new skill or plugin directly from a download path. diff --git a/src/clawops/assets/platform/docs/runbooks/linux-runtime-user-and-systemd.md b/src/clawops/assets/platform/docs/runbooks/linux-runtime-user-and-systemd.md index 493ec74e..24be3479 100644 --- a/src/clawops/assets/platform/docs/runbooks/linux-runtime-user-and-systemd.md +++ b/src/clawops/assets/platform/docs/runbooks/linux-runtime-user-and-systemd.md @@ -1,7 +1,6 @@ # Linux runtime user and user-level systemd -Create a dedicated non-admin account for the OpenClaw runtime and keep it -separate from your daily admin shell. +Create a dedicated non-admin account for the OpenClaw runtime and keep it separate from your daily admin shell. Preferred pattern: 1. create `openclawsvc` diff --git a/src/clawops/assets/platform/docs/runbooks/macos-service-user-and-ssh.md b/src/clawops/assets/platform/docs/runbooks/macos-service-user-and-ssh.md index 29065683..cf3dd901 100644 --- a/src/clawops/assets/platform/docs/runbooks/macos-service-user-and-ssh.md +++ b/src/clawops/assets/platform/docs/runbooks/macos-service-user-and-ssh.md @@ -1,7 +1,6 @@ # macOS runtime user and loopback SSH -Create a dedicated standard user for the OpenClaw runtime and keep it separate -from your daily admin account. +Create a dedicated standard user for the OpenClaw runtime and keep it separate from your daily admin account. Preferred pattern: 1. create `openclawsvc` @@ -16,5 +15,4 @@ Host-aware entrypoint: Create the dedicated runtime user with your platform-native user-management tooling ``` -You can then manage the runtime with: -`ssh openclawsvc@localhost` +You can then manage the runtime with: `ssh openclawsvc@localhost` diff --git a/src/clawops/assets/platform/plugins/memory-lancedb-pro/README.md b/src/clawops/assets/platform/plugins/memory-lancedb-pro/README.md index e380addf..48f1e6b9 100644 --- a/src/clawops/assets/platform/plugins/memory-lancedb-pro/README.md +++ b/src/clawops/assets/platform/plugins/memory-lancedb-pro/README.md @@ -8,10 +8,7 @@ A LanceDB-backed OpenClaw memory plugin that stores preferences, decisions, and project context, then auto-recalls them in future sessions. -[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) -[![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) -[![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) -[![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) +[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) [![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) [![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) [![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) [English](README.md) | [简体中文](README_CN.md) | [繁體中文](README_TW.md) | [日本語](README_JA.md) | [한국어](README_KO.md) | [Français](README_FR.md) | [Español](README_ES.md) | [Deutsch](README_DE.md) | [Italiano](README_IT.md) | [Русский](README_RU.md) | [Português (Brasil)](README_PT-BR.md) @@ -134,8 +131,7 @@ You should see: Done! Your agent now has long-term memory. -
-More installation paths (existing users, upgrades) +
More installation paths (existing users, upgrades) **Already using OpenClaw?** @@ -160,8 +156,7 @@ See `CHANGELOG-v1.1.0.md` for behavior changes and upgrade rationale.
-
-Telegram Bot Quick Import (click to expand) +
Telegram Bot Quick Import (click to expand) If you are using OpenClaw's Telegram integration, the easiest way is to send an import command directly to the main Bot instead of manually editing config. @@ -244,11 +239,9 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > Full walkthrough: installation, configuration, and hybrid retrieval internals. -[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) -**https://youtu.be/MtukF1C8epQ** +[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) **https://youtu.be/MtukF1C8epQ** -[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) -**https://www.bilibili.com/video/BV1zUf2BGEgn/** +[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) **https://www.bilibili.com/video/BV1zUf2BGEgn/** --- @@ -278,8 +271,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > For a deep-dive into the full architecture, see [docs/memory_architecture_analysis.md](docs/memory_architecture_analysis.md). -
-File Reference (click to expand) +
File Reference (click to expand) | File | Purpose | | --- | --- | @@ -371,8 +363,7 @@ Query → BM25 FTS ─────┘ --- -
-Compared to Built-in memory-lancedb (click to expand) +
Compared to Built-in memory-lancedb (click to expand) | Feature | Built-in `memory-lancedb` | **memory-lancedb-pro** | | --- | :---: | :---: | @@ -399,8 +390,7 @@ Query → BM25 FTS ─────┘ ## Configuration -
-Full Configuration Example +
Full Configuration Example ```json { @@ -464,8 +454,7 @@ Query → BM25 FTS ─────┘
-
-Embedding Providers +
Embedding Providers Works with **any OpenAI-compatible embedding API**: @@ -479,8 +468,7 @@ Works with **any OpenAI-compatible embedding API**:
-
-Rerank Providers +
Rerank Providers Cross-encoder reranking supports multiple providers via `rerankProvider`: @@ -495,8 +483,7 @@ Any Jina-compatible rerank endpoint also works — set `rerankProvider: "jina"`
-
-Smart Extraction (LLM) — v1.1.0 +
Smart Extraction (LLM) — v1.1.0 When `smartExtraction` is enabled (default: `true`), the plugin uses an LLM to intelligently extract and classify memories instead of regex-based triggers. @@ -537,8 +524,7 @@ Notes for `llm.auth: "oauth"`:
-
-Lifecycle Configuration (Decay + Tier) +
Lifecycle Configuration (Decay + Tier) | Field | Default | Description | |-------|---------|-------------| @@ -553,8 +539,7 @@ Notes for `llm.auth: "oauth"`:
-
-Access Reinforcement +
Access Reinforcement Frequently recalled memories decay more slowly (spaced-repetition style). @@ -596,8 +581,7 @@ OAuth login flow: ## Advanced Topics -
-If injected memories show up in replies +
If injected memories show up in replies Sometimes the model may echo the injected `` block. @@ -611,8 +595,7 @@ Sometimes the model may echo the injected `` block.
-
-Session Memory +
Session Memory - Triggered on `/new` command — saves previous session summary to LanceDB - Disabled by default (OpenClaw already has native `.jsonl` session persistence) @@ -622,8 +605,7 @@ See [docs/openclaw-integration-playbook.md](docs/openclaw-integration-playbook.m
-
-Custom Slash Commands (e.g. /lesson) +
Custom Slash Commands (e.g. /lesson) Add to your `CLAUDE.md`, `AGENTS.md`, or system prompt: @@ -642,8 +624,7 @@ When the user sends `/remember `:
-
-Iron Rules for AI Agents +
Iron Rules for AI Agents > Copy the block below into your `AGENTS.md` so your agent enforces these rules automatically. @@ -670,8 +651,7 @@ After modifying .ts files under plugins/, MUST run rm -rf /tmp/jiti/ BEFORE open
-
-Database Schema +
Database Schema LanceDB table `memories`: @@ -692,8 +672,7 @@ Common `metadata` keys in v1.1.0: `l0_abstract`, `l1_overview`, `l2_content`, `m
-
-Troubleshooting +
Troubleshooting ### "Cannot mix BigInt and other types" (LanceDB / Apache Arrow) @@ -815,29 +794,17 @@ Feedback: [GitHub Issues](https://github.com/CortexReach/memory-lancedb-pro/issu ## Contributors -

-@win4r -@kctony -@Akatsuki-Ryu -@JasonSuz -@Minidoracat -@furedericca-lab -@joe2643 -@AliceLJY -@chenjiyong -

+

@win4r @kctony @Akatsuki-Ryu @JasonSuz @Minidoracat @furedericca-lab @joe2643 @AliceLJY @chenjiyong

Full list: [Contributors](https://github.com/CortexReach/memory-lancedb-pro/graphs/contributors) ## Star History - - + Star History Chart - - + ## License diff --git a/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_CN.md b/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_CN.md index acbfb92b..947af5ec 100644 --- a/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_CN.md +++ b/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_CN.md @@ -8,10 +8,7 @@ 基于 LanceDB 的 OpenClaw 长期记忆插件，自动存储偏好、决策和项目上下文，在后续会话中自动回忆。 -[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) -[![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) -[![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) -[![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) +[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) [![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) [![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) [![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) [English](README.md) | [简体中文](README_CN.md) | [繁體中文](README_TW.md) | [日本語](README_JA.md) | [한국어](README_KO.md) | [Français](README_FR.md) | [Español](README_ES.md) | [Deutsch](README_DE.md) | [Italiano](README_IT.md) | [Русский](README_RU.md) | [Português (Brasil)](README_PT-BR.md) @@ -134,8 +131,7 @@ openclaw logs --follow --plain | grep "memory-lancedb-pro" 完成！你的智能体现在拥有长期记忆了。 -
-更多安装路径（已有用户、升级） +
更多安装路径（已有用户、升级） **已在使用 OpenClaw？** @@ -160,8 +156,7 @@ openclaw memory-pro stats
-
-Telegram Bot 快捷导入（点击展开） +
Telegram Bot 快捷导入（点击展开） 如果你在使用 OpenClaw 的 Telegram 集成，最简单的方式是直接给主 Bot 发消息，而不是手动编辑配置文件。 @@ -244,11 +239,9 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > 完整演示：安装、配置、混合检索内部原理。 -[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) -**https://youtu.be/MtukF1C8epQ** +[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) **https://youtu.be/MtukF1C8epQ** -[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) -**https://www.bilibili.com/video/BV1zUf2BGEgn/** +[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) **https://www.bilibili.com/video/BV1zUf2BGEgn/** --- @@ -278,8 +271,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > 完整架构解析见 [docs/memory_architecture_analysis.md](docs/memory_architecture_analysis.md)。 -
-文件说明（点击展开） +
文件说明（点击展开） | 文件 | 用途 | | --- | --- | @@ -369,8 +361,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla --- -
-与内置 memory-lancedb 的对比（点击展开） +
与内置 memory-lancedb 的对比（点击展开） | 功能 | 内置 `memory-lancedb` | **memory-lancedb-pro** | | --- | :---: | :---: | @@ -397,8 +388,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla ## 配置 -
-完整配置示例 +
完整配置示例 ```json { @@ -462,8 +452,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla
-
-Embedding 服务商 +
Embedding 服务商 兼容 **任意 OpenAI 兼容 Embedding API**： @@ -477,8 +466,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla
-
-重排序服务商 +
重排序服务商 交叉编码器重排序通过 `rerankProvider` 支持多个服务商： @@ -493,8 +481,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla
-
-智能提取（LLM）— v1.1.0 +
智能提取（LLM）— v1.1.0 当 `smartExtraction` 启用（默认 `true`）时，插件使用 LLM 智能提取和分类记忆，替代基于正则的触发方式。 @@ -535,8 +522,7 @@ OAuth `llm` 配置（使用现有 Codex / ChatGPT 登录缓存来发送 LLM 请
-
-生命周期配置（衰减 + 层级） +
生命周期配置（衰减 + 层级） | 字段 | 默认值 | 说明 | |------|--------|------| @@ -551,8 +537,7 @@ OAuth `llm` 配置（使用现有 Codex / ChatGPT 登录缓存来发送 LLM 请
-
-访问强化 +
访问强化 频繁被召回的记忆衰减更慢（类似间隔重复机制）。 @@ -594,8 +579,7 @@ OAuth 登录流程： ## 进阶主题 -
-注入的记忆出现在回复中 +
注入的记忆出现在回复中 有时模型可能会将注入的 `` 块原文输出。 @@ -609,8 +593,7 @@ OAuth 登录流程：
-
-会话记忆 +
会话记忆 - 通过 `/new` 命令触发——将上一段会话摘要保存到 LanceDB - 默认关闭（OpenClaw 已有原生 `.jsonl` 会话持久化） @@ -620,8 +603,7 @@ OAuth 登录流程：
-
-自定义斜杠命令（如 /lesson） +
自定义斜杠命令（如 /lesson） 在你的 `CLAUDE.md`、`AGENTS.md` 或系统提示词中添加： @@ -640,8 +622,7 @@ OAuth 登录流程：
-
-AI 智能体铁律 +
AI 智能体铁律 > 将以下内容复制到你的 `AGENTS.md`，让智能体自动遵守这些规则。 @@ -668,8 +649,7 @@ OAuth 登录流程：
-
-数据库 Schema +
数据库 Schema LanceDB 表 `memories`： @@ -690,8 +670,7 @@ v1.1.0 常用 `metadata` 字段：`l0_abstract`、`l1_overview`、`l2_content`
-
-故障排除 +
故障排除 ### "Cannot mix BigInt and other types"（LanceDB / Apache Arrow） @@ -738,29 +717,17 @@ v1.1.0 常用 `metadata` 字段：`l0_abstract`、`l1_overview`、`l2_content` ## 贡献者 -

-@win4r -@kctony -@Akatsuki-Ryu -@JasonSuz -@Minidoracat -@furedericca-lab -@joe2643 -@AliceLJY -@chenjiyong -

+

@win4r @kctony @Akatsuki-Ryu @JasonSuz @Minidoracat @furedericca-lab @joe2643 @AliceLJY @chenjiyong

完整列表：[Contributors](https://github.com/CortexReach/memory-lancedb-pro/graphs/contributors) ## Star 趋势 - - + Star History Chart - - + ## 许可证 diff --git a/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_DE.md b/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_DE.md index 104c259b..4d2ed0e0 100644 --- a/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_DE.md +++ b/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_DE.md @@ -8,10 +8,7 @@ Ein LanceDB-basiertes OpenClaw-Langzeitgedächtnis-Plugin, das Präferenzen, Entscheidungen und Projektkontext speichert und in zukünftigen Sitzungen automatisch abruft. -[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) -[![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) -[![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) -[![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) +[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) [![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) [![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) [![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) [English](README.md) | [简体中文](README_CN.md) | [繁體中文](README_TW.md) | [日本語](README_JA.md) | [한국어](README_KO.md) | [Français](README_FR.md) | [Español](README_ES.md) | [Deutsch](README_DE.md) | [Italiano](README_IT.md) | [Русский](README_RU.md) | [Português (Brasil)](README_PT-BR.md) @@ -134,8 +131,7 @@ Sie sollten sehen: Fertig! Ihr Agent verfügt jetzt über Langzeitgedächtnis. -
-Weitere Installationswege (bestehende Benutzer, Upgrades) +
Weitere Installationswege (bestehende Benutzer, Upgrades) **Bereits OpenClaw-Benutzer?** @@ -160,8 +156,7 @@ Siehe `CHANGELOG-v1.1.0.md` für Verhaltensänderungen und Upgrade-Begründung.
-
-Telegram-Bot-Schnellimport (zum Aufklappen klicken) +
Telegram-Bot-Schnellimport (zum Aufklappen klicken) Wenn Sie die Telegram-Integration von OpenClaw verwenden, ist es am einfachsten, einen Importbefehl direkt an den Hauptbot zu senden, anstatt die Konfiguration manuell zu bearbeiten. @@ -244,11 +239,9 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > Vollständige Anleitung: Installation, Konfiguration und Funktionsweise der hybriden Suche. -[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) -**https://youtu.be/MtukF1C8epQ** +[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) **https://youtu.be/MtukF1C8epQ** -[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) -**https://www.bilibili.com/video/BV1zUf2BGEgn/** +[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) **https://www.bilibili.com/video/BV1zUf2BGEgn/** --- @@ -278,8 +271,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > Für eine detaillierte Analyse der vollständigen Architektur siehe [docs/memory_architecture_analysis.md](docs/memory_architecture_analysis.md). -
-Dateireferenz (zum Aufklappen klicken) +
Dateireferenz (zum Aufklappen klicken) | Datei | Zweck | | --- | --- | @@ -369,8 +361,7 @@ Query → BM25 FTS ─────┘ --- -
-Vergleich mit dem eingebauten memory-lancedb (zum Aufklappen klicken) +
Vergleich mit dem eingebauten memory-lancedb (zum Aufklappen klicken) | Funktion | Eingebautes `memory-lancedb` | **memory-lancedb-pro** | | --- | :---: | :---: | @@ -397,8 +388,7 @@ Query → BM25 FTS ─────┘ ## Konfiguration -
-Vollständiges Konfigurationsbeispiel +
Vollständiges Konfigurationsbeispiel ```json { @@ -462,8 +452,7 @@ Query → BM25 FTS ─────┘
-
-Embedding-Anbieter +
Embedding-Anbieter Funktioniert mit **jeder OpenAI-kompatiblen Embedding-API**: @@ -477,8 +466,7 @@ Funktioniert mit **jeder OpenAI-kompatiblen Embedding-API**:
-
-Rerank-Anbieter +
Rerank-Anbieter Cross-Encoder Reranking unterstützt mehrere Anbieter über `rerankProvider`: @@ -493,8 +481,7 @@ Jeder Jina-kompatible Rerank-Endpunkt funktioniert ebenfalls — setzen Sie `rer
-
-Intelligente Extraktion (LLM) — v1.1.0 +
Intelligente Extraktion (LLM) — v1.1.0 Wenn `smartExtraction` aktiviert ist (Standard: `true`), verwendet das Plugin ein LLM, um Erinnerungen intelligent zu extrahieren und zu klassifizieren, anstatt regex-basierte Auslöser zu verwenden. @@ -535,8 +522,7 @@ Hinweise zu `llm.auth: "oauth"`:
-
-Lebenszyklus-Konfiguration (Zerfall + Stufen) +
Lebenszyklus-Konfiguration (Zerfall + Stufen) | Feld | Standard | Beschreibung | |-------|---------|-------------| @@ -551,8 +537,7 @@ Hinweise zu `llm.auth: "oauth"`:
-
-Zugriffsverstärkung +
Zugriffsverstärkung Häufig abgerufene Erinnerungen zerfallen langsamer (Spaced-Repetition-Stil). @@ -594,8 +579,7 @@ OAuth-Login-Ablauf: ## Erweiterte Themen -
-Wenn injizierte Erinnerungen in Antworten auftauchen +
Wenn injizierte Erinnerungen in Antworten auftauchen Manchmal kann das Modell den injizierten ``-Block wiedergeben. @@ -609,8 +593,7 @@ Manchmal kann das Modell den injizierten ``-Block wiedergeben
-
-Sitzungsgedächtnis +
Sitzungsgedächtnis - Wird beim `/new`-Befehl ausgelöst — speichert die vorherige Sitzungszusammenfassung in LanceDB - Standardmäßig deaktiviert (OpenClaw hat bereits native `.jsonl`-Sitzungspersistenz) @@ -620,8 +603,7 @@ Siehe [docs/openclaw-integration-playbook.md](docs/openclaw-integration-playbook
-
-Benutzerdefinierte Slash-Befehle (z.B. /lesson) +
Benutzerdefinierte Slash-Befehle (z.B. /lesson) Fügen Sie zu Ihrer `CLAUDE.md`, `AGENTS.md` oder Ihrem Systemprompt hinzu: @@ -640,8 +622,7 @@ When the user sends `/remember `:
-
-Eiserne Regeln für KI-Agenten +
Eiserne Regeln für KI-Agenten > Kopieren Sie den folgenden Block in Ihre `AGENTS.md`, damit Ihr Agent diese Regeln automatisch durchsetzt. @@ -668,8 +649,7 @@ After modifying .ts files under plugins/, MUST run rm -rf /tmp/jiti/ BEFORE open
-
-Datenbankschema +
Datenbankschema LanceDB-Tabelle `memories`: @@ -690,8 +670,7 @@ Häufige `metadata`-Schlüssel in v1.1.0: `l0_abstract`, `l1_overview`, `l2_cont
-
-Fehlerbehebung +
Fehlerbehebung ### "Cannot mix BigInt and other types" (LanceDB / Apache Arrow) @@ -738,29 +717,17 @@ Feedback: [GitHub Issues](https://github.com/CortexReach/memory-lancedb-pro/issu ## Contributors -

-@win4r -@kctony -@Akatsuki-Ryu -@JasonSuz -@Minidoracat -@furedericca-lab -@joe2643 -@AliceLJY -@chenjiyong -

+

@win4r @kctony @Akatsuki-Ryu @JasonSuz @Minidoracat @furedericca-lab @joe2643 @AliceLJY @chenjiyong

Full list: [Contributors](https://github.com/CortexReach/memory-lancedb-pro/graphs/contributors) ## Star History - - + Star History Chart - - + ## Lizenz diff --git a/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_ES.md b/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_ES.md index fe4d173c..d6e78c90 100644 --- a/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_ES.md +++ b/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_ES.md @@ -8,10 +8,7 @@ Un plugin de memoria para OpenClaw respaldado por LanceDB que almacena preferencias, decisiones y contexto de proyectos, y los recupera automáticamente en sesiones futuras. -[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) -[![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) -[![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) -[![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) +[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) [![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) [![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) [![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) [English](README.md) | [简体中文](README_CN.md) | [繁體中文](README_TW.md) | [日本語](README_JA.md) | [한국어](README_KO.md) | [Français](README_FR.md) | [Español](README_ES.md) | [Deutsch](README_DE.md) | [Italiano](README_IT.md) | [Русский](README_RU.md) | [Português (Brasil)](README_PT-BR.md) @@ -134,8 +131,7 @@ Deberías ver: ¡Listo! Tu agente ahora tiene memoria a largo plazo. -
-Más rutas de instalación (usuarios existentes, actualizaciones) +
Más rutas de instalación (usuarios existentes, actualizaciones) **¿Ya usas OpenClaw?** @@ -160,8 +156,7 @@ Consulta `CHANGELOG-v1.1.0.md` para los cambios de comportamiento y la justifica
-
-Importación rápida para Bot de Telegram (clic para expandir) +
Importación rápida para Bot de Telegram (clic para expandir) Si usas la integración de Telegram de OpenClaw, la forma más fácil es enviar un comando de importación directamente al Bot principal en lugar de editar la configuración manualmente. @@ -244,11 +239,9 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > Recorrido completo: instalación, configuración y funcionamiento interno de la recuperación híbrida. -[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) -**https://youtu.be/MtukF1C8epQ** +[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) **https://youtu.be/MtukF1C8epQ** -[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) -**https://www.bilibili.com/video/BV1zUf2BGEgn/** +[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) **https://www.bilibili.com/video/BV1zUf2BGEgn/** --- @@ -278,8 +271,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > Para un análisis detallado de la arquitectura completa, consulta [docs/memory_architecture_analysis.md](docs/memory_architecture_analysis.md). -
-Referencia de Archivos (clic para expandir) +
Referencia de Archivos (clic para expandir) | Archivo | Propósito | | --- | --- | @@ -369,8 +361,7 @@ Query → BM25 FTS ─────┘ --- -
-Comparación con memory-lancedb integrado (clic para expandir) +
Comparación con memory-lancedb integrado (clic para expandir) | Característica | `memory-lancedb` integrado | **memory-lancedb-pro** | | --- | :---: | :---: | @@ -397,8 +388,7 @@ Query → BM25 FTS ─────┘ ## Configuración -
-Ejemplo de Configuración Completa +
Ejemplo de Configuración Completa ```json { @@ -462,8 +452,7 @@ Query → BM25 FTS ─────┘
-
-Proveedores de Embedding +
Proveedores de Embedding Funciona con **cualquier API de embedding compatible con OpenAI**: @@ -477,8 +466,7 @@ Funciona con **cualquier API de embedding compatible con OpenAI**:
-
-Proveedores de Rerank +
Proveedores de Rerank El reranking con cross-encoder admite múltiples proveedores mediante `rerankProvider`: @@ -493,8 +481,7 @@ Cualquier endpoint de rerank compatible con Jina también funciona — configura
-
-Smart Extraction (LLM) — v1.1.0 +
Smart Extraction (LLM) — v1.1.0 Cuando `smartExtraction` está habilitado (predeterminado: `true`), el plugin utiliza un LLM para extraer y clasificar recuerdos de forma inteligente en lugar de disparadores basados en regex. @@ -535,8 +522,7 @@ Notas para `llm.auth: "oauth"`:
-
-Configuración del Ciclo de Vida (Decaimiento + Nivel) +
Configuración del Ciclo de Vida (Decaimiento + Nivel) | Campo | Predeterminado | Descripción | |-------|----------------|-------------| @@ -551,8 +537,7 @@ Notas para `llm.auth: "oauth"`:
-
-Refuerzo por Acceso +
Refuerzo por Acceso Los recuerdos frecuentemente recuperados decaen más lentamente (estilo repetición espaciada). @@ -594,8 +579,7 @@ Flujo de inicio de sesión OAuth: ## Temas Avanzados -
-Si los recuerdos inyectados aparecen en las respuestas +
Si los recuerdos inyectados aparecen en las respuestas A veces el modelo puede repetir el bloque `` inyectado. @@ -609,8 +593,7 @@ A veces el modelo puede repetir el bloque `` inyectado.
-
-Memoria de Sesión +
Memoria de Sesión - Se activa con el comando `/new` — guarda el resumen de la sesión anterior en LanceDB - Deshabilitado por defecto (OpenClaw ya tiene persistencia nativa de sesión en `.jsonl`) @@ -620,8 +603,7 @@ Consulta [docs/openclaw-integration-playbook.md](docs/openclaw-integration-playb
-
-Comandos Slash Personalizados (por ejemplo, /lesson) +
Comandos Slash Personalizados (por ejemplo, /lesson) Agrega a tu `CLAUDE.md`, `AGENTS.md` o prompt del sistema (el bloque se mantiene en inglés para que el agente lo interprete correctamente): @@ -640,8 +622,7 @@ When the user sends `/remember `:
-
-Reglas de Hierro para Agentes de IA +
Reglas de Hierro para Agentes de IA > Copia el bloque de abajo en tu `AGENTS.md` para que tu agente aplique estas reglas automáticamente. Se mantiene en inglés porque es instrucción directa para el modelo. @@ -668,8 +649,7 @@ After modifying .ts files under plugins/, MUST run rm -rf /tmp/jiti/ BEFORE open
-
-Esquema de la Base de Datos +
Esquema de la Base de Datos Tabla LanceDB `memories`: @@ -690,8 +670,7 @@ Claves comunes de `metadata` en v1.1.0: `l0_abstract`, `l1_overview`, `l2_conten
-
-Solución de Problemas +
Solución de Problemas ### "Cannot mix BigInt and other types" (LanceDB / Apache Arrow) @@ -738,29 +717,17 @@ Comentarios: [GitHub Issues](https://github.com/CortexReach/memory-lancedb-pro/i ## Contributors -

-@win4r -@kctony -@Akatsuki-Ryu -@JasonSuz -@Minidoracat -@furedericca-lab -@joe2643 -@AliceLJY -@chenjiyong -

+

@win4r @kctony @Akatsuki-Ryu @JasonSuz @Minidoracat @furedericca-lab @joe2643 @AliceLJY @chenjiyong

Full list: [Contributors](https://github.com/CortexReach/memory-lancedb-pro/graphs/contributors) ## Star History - - + Star History Chart - - + ## Licencia diff --git a/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_FR.md b/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_FR.md index 19f5fc45..03906485 100644 --- a/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_FR.md +++ b/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_FR.md @@ -8,10 +8,7 @@ Un plugin de mémoire long terme pour OpenClaw basé sur LanceDB qui stocke les préférences, les décisions et le contexte du projet, puis les rappelle automatiquement dans les sessions futures. -[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) -[![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) -[![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) -[![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) +[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) [![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) [![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) [![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) [English](README.md) | [简体中文](README_CN.md) | [繁體中文](README_TW.md) | [日本語](README_JA.md) | [한국어](README_KO.md) | [Français](README_FR.md) | [Español](README_ES.md) | [Deutsch](README_DE.md) | [Italiano](README_IT.md) | [Русский](README_RU.md) | [Português (Brasil)](README_PT-BR.md) @@ -134,8 +131,7 @@ Vous devriez voir : Terminé ! Votre agent dispose maintenant d'une mémoire long terme. -
-Plus de chemins d'installation (utilisateurs existants, mises à niveau) +
Plus de chemins d'installation (utilisateurs existants, mises à niveau) **Déjà utilisateur d'OpenClaw ?** @@ -160,8 +156,7 @@ Consultez `CHANGELOG-v1.1.0.md` pour les changements de comportement et la justi
-
-Import rapide Telegram Bot (cliquez pour développer) +
Import rapide Telegram Bot (cliquez pour développer) Si vous utilisez l'intégration Telegram d'OpenClaw, le plus simple est d'envoyer une commande d'import directement au Bot principal au lieu de modifier manuellement la configuration. @@ -244,11 +239,9 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > Présentation complète : installation, configuration et fonctionnement interne de la recherche hybride. -[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) -**https://youtu.be/MtukF1C8epQ** +[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) **https://youtu.be/MtukF1C8epQ** -[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) -**https://www.bilibili.com/video/BV1zUf2BGEgn/** +[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) **https://www.bilibili.com/video/BV1zUf2BGEgn/** --- @@ -278,8 +271,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > Pour une analyse approfondie de l'architecture complète, consultez [docs/memory_architecture_analysis.md](docs/memory_architecture_analysis.md). -
-Référence des fichiers (cliquez pour développer) +
Référence des fichiers (cliquez pour développer) | Fichier | Rôle | | --- | --- | @@ -369,8 +361,7 @@ Requête → BM25 FTS ─────┘ --- -
-Comparaison avec memory-lancedb intégré (cliquez pour développer) +
Comparaison avec memory-lancedb intégré (cliquez pour développer) | Fonctionnalité | `memory-lancedb` intégré | **memory-lancedb-pro** | | --- | :---: | :---: | @@ -397,8 +388,7 @@ Requête → BM25 FTS ─────┘ ## Configuration -
-Exemple de configuration complète +
Exemple de configuration complète ```json { @@ -462,8 +452,7 @@ Requête → BM25 FTS ─────┘
-
-Fournisseurs d'embedding +
Fournisseurs d'embedding Fonctionne avec **toute API d'embedding compatible OpenAI** : @@ -477,8 +466,7 @@ Fonctionne avec **toute API d'embedding compatible OpenAI** :
-
-Fournisseurs de reranking +
Fournisseurs de reranking Le reranking cross-encoder supporte plusieurs fournisseurs via `rerankProvider` : @@ -493,8 +481,7 @@ Tout endpoint de reranking compatible Jina fonctionne également — définissez
-
-Extraction intelligente (LLM) — v1.1.0 +
Extraction intelligente (LLM) — v1.1.0 Quand `smartExtraction` est activé (par défaut : `true`), le plugin utilise un LLM pour extraire et classifier intelligemment les souvenirs au lieu de déclencheurs basés sur des regex. @@ -535,8 +522,7 @@ Notes pour `llm.auth: "oauth"` :
-
-Configuration du cycle de vie (Décroissance + Niveaux) +
Configuration du cycle de vie (Décroissance + Niveaux) | Champ | Défaut | Description | |-------|---------|-------------| @@ -551,8 +537,7 @@ Notes pour `llm.auth: "oauth"` :
-
-Renforcement par accès +
Renforcement par accès Les souvenirs fréquemment rappelés décroissent plus lentement (style répétition espacée). @@ -594,8 +579,7 @@ Flux de connexion OAuth : ## Sujets avancés -
-Si les souvenirs injectés apparaissent dans les réponses +
Si les souvenirs injectés apparaissent dans les réponses Parfois le modèle peut répéter le bloc `` injecté. @@ -609,8 +593,7 @@ Parfois le modèle peut répéter le bloc `` injecté.
-
-Mémoire de session +
Mémoire de session - Déclenchée par la commande `/new` — sauvegarde le résumé de la session précédente dans LanceDB - Désactivée par défaut (OpenClaw dispose déjà d'une persistance native de session `.jsonl`) @@ -620,8 +603,7 @@ Consultez [docs/openclaw-integration-playbook.md](docs/openclaw-integration-play
-
-Commandes slash personnalisées (ex. /lesson) +
Commandes slash personnalisées (ex. /lesson) Ajoutez à votre `CLAUDE.md`, `AGENTS.md` ou prompt système : @@ -640,8 +622,7 @@ When the user sends `/remember `:
-
-Règles d'or pour les agents IA +
Règles d'or pour les agents IA > Copiez le bloc ci-dessous dans votre `AGENTS.md` pour que votre agent applique automatiquement ces règles. @@ -668,8 +649,7 @@ After modifying .ts files under plugins/, MUST run rm -rf /tmp/jiti/ BEFORE open
-
-Schéma de la base de données +
Schéma de la base de données Table LanceDB `memories` : @@ -690,8 +670,7 @@ Clés `metadata` courantes en v1.1.0 : `l0_abstract`, `l1_overview`, `l2_content
-
-Dépannage +
Dépannage ### "Cannot mix BigInt and other types" (LanceDB / Apache Arrow) @@ -738,29 +717,17 @@ Retours : [GitHub Issues](https://github.com/CortexReach/memory-lancedb-pro/issu ## Contributors -

-@win4r -@kctony -@Akatsuki-Ryu -@JasonSuz -@Minidoracat -@furedericca-lab -@joe2643 -@AliceLJY -@chenjiyong -

+

@win4r @kctony @Akatsuki-Ryu @JasonSuz @Minidoracat @furedericca-lab @joe2643 @AliceLJY @chenjiyong

Full list: [Contributors](https://github.com/CortexReach/memory-lancedb-pro/graphs/contributors) ## Star History - - + Star History Chart - - + ## Licence diff --git a/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_IT.md b/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_IT.md index b1679682..e43bc7dc 100644 --- a/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_IT.md +++ b/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_IT.md @@ -8,10 +8,7 @@ Un plugin di memoria a lungo termine per OpenClaw basato su LanceDB che memorizza preferenze, decisioni e contesto di progetto, e li richiama automaticamente nelle sessioni future. -[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) -[![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) -[![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) -[![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) +[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) [![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) [![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) [![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) [English](README.md) | [简体中文](README_CN.md) | [繁體中文](README_TW.md) | [日本語](README_JA.md) | [한국어](README_KO.md) | [Français](README_FR.md) | [Español](README_ES.md) | [Deutsch](README_DE.md) | [Italiano](README_IT.md) | [Русский](README_RU.md) | [Português (Brasil)](README_PT-BR.md) @@ -134,8 +131,7 @@ Dovresti vedere: Fatto! Il tuo agente ora ha una memoria a lungo termine. -
-Ulteriori percorsi di installazione (utenti esistenti, aggiornamenti) +
Ulteriori percorsi di installazione (utenti esistenti, aggiornamenti) **Usi già OpenClaw?** @@ -160,8 +156,7 @@ Vedi `CHANGELOG-v1.1.0.md` per le modifiche comportamentali e le motivazioni del
-
-Importazione rapida Telegram Bot (clicca per espandere) +
Importazione rapida Telegram Bot (clicca per espandere) Se stai usando l'integrazione Telegram di OpenClaw, il modo più semplice è inviare un comando di importazione direttamente al Bot principale invece di modificare manualmente la configurazione. @@ -244,11 +239,9 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > Guida completa: installazione, configurazione e funzionamento interno della ricerca ibrida. -[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) -**https://youtu.be/MtukF1C8epQ** +[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) **https://youtu.be/MtukF1C8epQ** -[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) -**https://www.bilibili.com/video/BV1zUf2BGEgn/** +[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) **https://www.bilibili.com/video/BV1zUf2BGEgn/** --- @@ -278,8 +271,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > Per un approfondimento sull'architettura completa, consulta [docs/memory_architecture_analysis.md](docs/memory_architecture_analysis.md). -
-Riferimento file (clicca per espandere) +
Riferimento file (clicca per espandere) | File | Scopo | | --- | --- | @@ -369,8 +361,7 @@ Query → BM25 FTS ─────┘ --- -
-Confronto con memory-lancedb integrato (clicca per espandere) +
Confronto con memory-lancedb integrato (clicca per espandere) | Funzionalità | `memory-lancedb` integrato | **memory-lancedb-pro** | | --- | :---: | :---: | @@ -397,8 +388,7 @@ Query → BM25 FTS ─────┘ ## Configurazione -
-Esempio di configurazione completa +
Esempio di configurazione completa ```json { @@ -462,8 +452,7 @@ Query → BM25 FTS ─────┘
-
-Provider di embedding +
Provider di embedding Funziona con **qualsiasi API di embedding compatibile OpenAI**: @@ -477,8 +466,7 @@ Funziona con **qualsiasi API di embedding compatibile OpenAI**:
-
-Provider di rerank +
Provider di rerank Il reranking cross-encoder supporta più provider tramite `rerankProvider`: @@ -493,8 +481,7 @@ Funziona anche qualsiasi endpoint di rerank compatibile Jina — imposta `rerank
-
-Estrazione intelligente (LLM) — v1.1.0 +
Estrazione intelligente (LLM) — v1.1.0 Quando `smartExtraction` è abilitato (predefinito: `true`), il plugin utilizza un LLM per estrarre e classificare intelligentemente i ricordi invece di trigger basati su regex. @@ -535,8 +522,7 @@ Note per `llm.auth: "oauth"`:
-
-Configurazione ciclo di vita (Decadimento + Livelli) +
Configurazione ciclo di vita (Decadimento + Livelli) | Campo | Predefinito | Descrizione | |-------|---------|-------------| @@ -551,8 +537,7 @@ Note per `llm.auth: "oauth"`:
-
-Rinforzo per accesso +
Rinforzo per accesso I ricordi richiamati frequentemente decadono più lentamente (stile ripetizione spaziata). @@ -594,8 +579,7 @@ Flusso di login OAuth: ## Argomenti avanzati -
-Se i ricordi iniettati appaiono nelle risposte +
Se i ricordi iniettati appaiono nelle risposte A volte il modello può ripetere il blocco `` iniettato. @@ -609,8 +593,7 @@ A volte il modello può ripetere il blocco `` iniettato.
-
-Memoria di sessione +
Memoria di sessione - Si attiva con il comando `/new` — salva il riepilogo della sessione precedente in LanceDB - Disabilitata per impostazione predefinita (OpenClaw ha già la persistenza nativa delle sessioni in `.jsonl`) @@ -620,8 +603,7 @@ Vedi [docs/openclaw-integration-playbook.md](docs/openclaw-integration-playbook.
-
-Comandi slash personalizzati (ad es. /lesson) +
Comandi slash personalizzati (ad es. /lesson) Aggiungi al tuo `CLAUDE.md`, `AGENTS.md` o prompt di sistema: @@ -640,8 +622,7 @@ When the user sends `/remember `:
-
-Regole d'oro per agenti IA +
Regole d'oro per agenti IA > Copia il blocco seguente nel tuo `AGENTS.md` in modo che il tuo agente applichi queste regole automaticamente. @@ -668,8 +649,7 @@ After modifying .ts files under plugins/, MUST run rm -rf /tmp/jiti/ BEFORE open
-
-Schema del database +
Schema del database Tabella LanceDB `memories`: @@ -690,8 +670,7 @@ Chiavi `metadata` comuni nella v1.1.0: `l0_abstract`, `l1_overview`, `l2_content
-
-Risoluzione dei problemi +
Risoluzione dei problemi ### "Cannot mix BigInt and other types" (LanceDB / Apache Arrow) @@ -738,29 +717,17 @@ Feedback: [GitHub Issues](https://github.com/CortexReach/memory-lancedb-pro/issu ## Contributors -

-@win4r -@kctony -@Akatsuki-Ryu -@JasonSuz -@Minidoracat -@furedericca-lab -@joe2643 -@AliceLJY -@chenjiyong -

+

@win4r @kctony @Akatsuki-Ryu @JasonSuz @Minidoracat @furedericca-lab @joe2643 @AliceLJY @chenjiyong

Full list: [Contributors](https://github.com/CortexReach/memory-lancedb-pro/graphs/contributors) ## Star History - - + Star History Chart - - + ## Licenza diff --git a/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_JA.md b/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_JA.md index 0627a2de..08c665ee 100644 --- a/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_JA.md +++ b/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_JA.md @@ -8,10 +8,7 @@ LanceDB ベースの OpenClaw 長期メモリプラグイン。好み・意思決定・プロジェクトコンテキストを自動保存し、将来のセッションで自動的に想起します。 -[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) -[![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) -[![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) -[![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) +[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) [![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) [![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) [![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) [English](README.md) | [简体中文](README_CN.md) | [繁體中文](README_TW.md) | [日本語](README_JA.md) | [한국어](README_KO.md) | [Français](README_FR.md) | [Español](README_ES.md) | [Deutsch](README_DE.md) | [Italiano](README_IT.md) | [Русский](README_RU.md) | [Português (Brasil)](README_PT-BR.md) @@ -134,8 +131,7 @@ openclaw logs --follow --plain | grep "memory-lancedb-pro" 完了！あなたのエージェントは長期メモリを持つようになりました。 -
-その他のインストール方法（既存ユーザー、アップグレード） +
その他のインストール方法（既存ユーザー、アップグレード） **既に OpenClaw を使用中？** @@ -160,8 +156,7 @@ openclaw memory-pro stats
-
-Telegram Bot クイックインポート（クリックで展開） +
Telegram Bot クイックインポート（クリックで展開） OpenClaw の Telegram 連携を使用している場合、設定ファイルを手動で編集するより、メイン Bot にインポートコマンドを直接送信するのが最も簡単です。 @@ -244,11 +239,9 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > フルウォークスルー：インストール、設定、ハイブリッド検索の内部構造。 -[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) -**https://youtu.be/MtukF1C8epQ** +[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) **https://youtu.be/MtukF1C8epQ** -[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) -**https://www.bilibili.com/video/BV1zUf2BGEgn/** +[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) **https://www.bilibili.com/video/BV1zUf2BGEgn/** --- @@ -278,8 +271,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > 完全なアーキテクチャの詳細は [docs/memory_architecture_analysis.md](docs/memory_architecture_analysis.md) を参照してください。 -
-ファイルリファレンス（クリックで展開） +
ファイルリファレンス（クリックで展開） | ファイル | 用途 | | --- | --- | @@ -369,8 +361,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla --- -
-内蔵 memory-lancedb との比較（クリックで展開） +
内蔵 memory-lancedb との比較（クリックで展開） | 機能 | 内蔵 `memory-lancedb` | **memory-lancedb-pro** | | --- | :---: | :---: | @@ -397,8 +388,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla ## 設定 -
-完全な設定例 +
完全な設定例 ```json { @@ -462,8 +452,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla
-
-Embedding プロバイダー +
Embedding プロバイダー **任意の OpenAI 互換 Embedding API** で動作： @@ -477,8 +466,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla
-
-リランクプロバイダー +
リランクプロバイダー クロスエンコーダーリランキングは `rerankProvider` で複数のプロバイダーをサポート： @@ -493,8 +481,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla
-
-スマート抽出（LLM）— v1.1.0 +
スマート抽出（LLM）— v1.1.0 `smartExtraction` が有効（デフォルト：`true`）の場合、プラグインは正規表現ベースのトリガーの代わりに LLM を使用してインテリジェントにメモリを抽出・分類します。 @@ -535,8 +522,7 @@ OAuth `llm` 設定（既存の Codex / ChatGPT ログインキャッシュを使
-
-ライフサイクル設定（減衰 + 階層） +
ライフサイクル設定（減衰 + 階層） | フィールド | デフォルト | 説明 | |-------|---------|-------------| @@ -551,8 +537,7 @@ OAuth `llm` 設定（既存の Codex / ChatGPT ログインキャッシュを使
-
-アクセス強化 +
アクセス強化 頻繁に想起されるメモリはより遅く減衰します（間隔反復スタイル）。 @@ -594,8 +579,7 @@ OAuth ログインフロー： ## 応用トピック -
-注入されたメモリが応答に表示される場合 +
注入されたメモリが応答に表示される場合 モデルが注入された `` ブロックをそのまま出力してしまうことがあります。 @@ -609,8 +593,7 @@ OAuth ログインフロー：
-
-セッションメモリ +
セッションメモリ - `/new` コマンドでトリガー——前のセッションの要約を LanceDB に保存 - デフォルトで無効（OpenClaw にはネイティブの `.jsonl` セッション永続化機能あり） @@ -620,8 +603,7 @@ OAuth ログインフロー：
-
-カスタムスラッシュコマンド（例：/lesson） +
カスタムスラッシュコマンド（例：/lesson） `CLAUDE.md`、`AGENTS.md`、またはシステムプロンプトに追加： @@ -640,8 +622,7 @@ OAuth ログインフロー：
-
-AI エージェントの鉄則 +
AI エージェントの鉄則 > 以下のブロックを `AGENTS.md` にコピーして、エージェントがこれらのルールを自動的に適用するようにしてください。 @@ -668,8 +649,7 @@ plugins/ 配下の .ts ファイルを変更した後、openclaw gateway restart
-
-データベーススキーマ +
データベーススキーマ LanceDB テーブル `memories`： @@ -690,8 +670,7 @@ v1.1.0 の一般的な `metadata` キー：`l0_abstract`、`l1_overview`、`l2_c
-
-トラブルシューティング +
トラブルシューティング ### "Cannot mix BigInt and other types"（LanceDB / Apache Arrow） @@ -738,29 +717,17 @@ LanceDB 0.26+ では、一部の数値カラムが `BigInt` として返され ## コントリビューター -

-@win4r -@kctony -@Akatsuki-Ryu -@JasonSuz -@Minidoracat -@furedericca-lab -@joe2643 -@AliceLJY -@chenjiyong -

+

@win4r @kctony @Akatsuki-Ryu @JasonSuz @Minidoracat @furedericca-lab @joe2643 @AliceLJY @chenjiyong

全リスト：[Contributors](https://github.com/CortexReach/memory-lancedb-pro/graphs/contributors) ## Star 履歴 - - + Star History Chart - - + ## ライセンス diff --git a/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_KO.md b/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_KO.md index c8f165e6..c81ca062 100644 --- a/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_KO.md +++ b/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_KO.md @@ -8,10 +8,7 @@ LanceDB 기반 OpenClaw 메모리 플러그인으로, 사용자 선호도·의사결정·프로젝트 맥락을 저장하고 이후 세션에서 자동으로 불러옵니다. -[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) -[![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) -[![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) -[![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) +[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) [![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) [![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) [![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) [English](README.md) | [简体中文](README_CN.md) | [繁體中文](README_TW.md) | [日本語](README_JA.md) | [한국어](README_KO.md) | [Français](README_FR.md) | [Español](README_ES.md) | [Deutsch](README_DE.md) | [Italiano](README_IT.md) | [Русский](README_RU.md) | [Português (Brasil)](README_PT-BR.md) @@ -134,8 +131,7 @@ openclaw logs --follow --plain | grep "memory-lancedb-pro" 완료! 이제 에이전트가 장기 기억을 갖게 됩니다. -
-추가 설치 경로 (기존 사용자, 업그레이드) +
추가 설치 경로 (기존 사용자, 업그레이드) **이미 OpenClaw를 사용 중인 경우:** @@ -160,8 +156,7 @@ openclaw memory-pro stats
-
-Telegram 봇 빠른 가져오기 (클릭하여 펼치기) +
Telegram 봇 빠른 가져오기 (클릭하여 펼치기) OpenClaw의 Telegram 연동을 사용하는 경우, 수동으로 설정을 편집하는 대신 메인 봇에 가져오기 명령어를 직접 보내는 것이 가장 쉬운 방법입니다. @@ -244,11 +239,9 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > 전체 안내: 설치, 설정, 하이브리드 검색 내부 구조. -[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) -**https://youtu.be/MtukF1C8epQ** +[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) **https://youtu.be/MtukF1C8epQ** -[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) -**https://www.bilibili.com/video/BV1zUf2BGEgn/** +[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) **https://www.bilibili.com/video/BV1zUf2BGEgn/** --- @@ -278,8 +271,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > 전체 아키텍처에 대한 심층 분석은 [docs/memory_architecture_analysis.md](docs/memory_architecture_analysis.md)를 참조하세요. -
-파일 레퍼런스 (클릭하여 펼치기) +
파일 레퍼런스 (클릭하여 펼치기) | 파일 | 용도 | | --- | --- | @@ -369,8 +361,7 @@ Query → BM25 FTS ─────┘ --- -
-내장 memory-lancedb와의 비교 (클릭하여 펼치기) +
내장 memory-lancedb와의 비교 (클릭하여 펼치기) | 기능 | 내장 `memory-lancedb` | **memory-lancedb-pro** | | --- | :---: | :---: | @@ -397,8 +388,7 @@ Query → BM25 FTS ─────┘ ## 설정 -
-전체 설정 예시 +
전체 설정 예시 ```json { @@ -462,8 +452,7 @@ Query → BM25 FTS ─────┘
-
-임베딩 프로바이더 +
임베딩 프로바이더 **OpenAI 호환 임베딩 API**와 모두 동작합니다: @@ -477,8 +466,7 @@ Query → BM25 FTS ─────┘
-
-리랭크 프로바이더 +
리랭크 프로바이더 Cross-Encoder 리랭킹은 `rerankProvider`를 통해 여러 프로바이더를 지원합니다: @@ -493,8 +481,7 @@ Jina 호환 리랭크 엔드포인트도 사용 가능합니다 — `rerankProvi
-
-Smart Extraction (LLM) — v1.1.0 +
Smart Extraction (LLM) — v1.1.0 `smartExtraction`이 활성화되면 (기본값: `true`), 플러그인이 정규식 기반 트리거 대신 LLM을 사용하여 기억을 지능적으로 추출하고 분류합니다. @@ -535,8 +522,7 @@ OAuth `llm` 설정 (기존 Codex / ChatGPT 로그인 캐시를 LLM 호출에 사
-
-라이프사이클 설정 (감쇠 + 단계) +
라이프사이클 설정 (감쇠 + 단계) | 필드 | 기본값 | 설명 | |-------|---------|-------------| @@ -551,8 +537,7 @@ OAuth `llm` 설정 (기존 Codex / ChatGPT 로그인 캐시를 LLM 호출에 사
-
-접근 강화 +
접근 강화 자주 불러오는 기억은 더 느리게 감쇠합니다 (간격 반복 학습 방식). @@ -594,8 +579,7 @@ OAuth 로그인 흐름: ## 고급 주제 -
-주입된 기억이 응답에 표시되는 경우 +
주입된 기억이 응답에 표시되는 경우 가끔 모델이 주입된 `` 블록을 그대로 출력할 수 있습니다. @@ -609,8 +593,7 @@ OAuth 로그인 흐름:
-
-세션 메모리 +
세션 메모리 - `/new` 명령어 시 작동 — 이전 세션 요약을 LanceDB에 저장 - 기본적으로 비활성화 (OpenClaw에 이미 네이티브 `.jsonl` 세션 영속화 기능이 있음) @@ -620,8 +603,7 @@ OAuth 로그인 흐름:
-
-커스텀 슬래시 명령어 (예: /lesson) +
커스텀 슬래시 명령어 (예: /lesson) `CLAUDE.md`, `AGENTS.md` 또는 시스템 프롬프트에 다음을 추가하세요 (에이전트가 읽는 영문 지시문이므로 그대로 사용합니다): @@ -640,8 +622,7 @@ When the user sends `/remember `:
-
-AI 에이전트를 위한 철칙 +
AI 에이전트를 위한 철칙 > 아래 블록을 `AGENTS.md`에 복사하여 에이전트가 이 규칙을 자동으로 적용하도록 하세요 (에이전트가 읽는 영문 지시문이므로 그대로 사용합니다). @@ -668,8 +649,7 @@ After modifying .ts files under plugins/, MUST run rm -rf /tmp/jiti/ BEFORE open
-
-데이터베이스 스키마 +
데이터베이스 스키마 LanceDB 테이블 `memories`: @@ -690,8 +670,7 @@ v1.1.0의 주요 `metadata` 키: `l0_abstract`, `l1_overview`, `l2_content`, `me
-
-문제 해결 +
문제 해결 ### "Cannot mix BigInt and other types" (LanceDB / Apache Arrow) @@ -738,29 +717,17 @@ LanceDB 0.26 이상에서 일부 숫자 열이 `BigInt`로 반환될 수 있습 ## Contributors -

-@win4r -@kctony -@Akatsuki-Ryu -@JasonSuz -@Minidoracat -@furedericca-lab -@joe2643 -@AliceLJY -@chenjiyong -

+

@win4r @kctony @Akatsuki-Ryu @JasonSuz @Minidoracat @furedericca-lab @joe2643 @AliceLJY @chenjiyong

Full list: [Contributors](https://github.com/CortexReach/memory-lancedb-pro/graphs/contributors) ## Star History - - + Star History Chart - - + ## 라이선스 diff --git a/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_PT-BR.md b/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_PT-BR.md index 65d721f8..c7a03749 100644 --- a/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_PT-BR.md +++ b/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_PT-BR.md @@ -8,10 +8,7 @@ Um plugin de memória de longo prazo para OpenClaw baseado em LanceDB que armazena preferências, decisões e contexto de projetos, e os recupera automaticamente em sessões futuras. -[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) -[![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) -[![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) -[![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) +[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) [![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) [![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) [![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) [English](README.md) | [简体中文](README_CN.md) | [繁體中文](README_TW.md) | [日本語](README_JA.md) | [한국어](README_KO.md) | [Français](README_FR.md) | [Español](README_ES.md) | [Deutsch](README_DE.md) | [Italiano](README_IT.md) | [Русский](README_RU.md) | [Português (Brasil)](README_PT-BR.md) @@ -134,8 +131,7 @@ Você deve ver: Pronto! Seu agente agora tem memória de longo prazo. -
-Mais caminhos de instalação (usuários existentes, upgrades) +
Mais caminhos de instalação (usuários existentes, upgrades) **Já está usando OpenClaw?** @@ -160,8 +156,7 @@ Veja `CHANGELOG-v1.1.0.md` para mudanças de comportamento e justificativa de up
-
-Importação rápida via Telegram Bot (clique para expandir) +
Importação rápida via Telegram Bot (clique para expandir) Se você está usando a integração Telegram do OpenClaw, a maneira mais fácil é enviar um comando de importação diretamente para o Bot principal em vez de editar a configuração manualmente. @@ -244,11 +239,9 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > Guia completo: instalação, configuração e funcionamento interno da busca híbrida. -[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) -**https://youtu.be/MtukF1C8epQ** +[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) **https://youtu.be/MtukF1C8epQ** -[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) -**https://www.bilibili.com/video/BV1zUf2BGEgn/** +[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) **https://www.bilibili.com/video/BV1zUf2BGEgn/** --- @@ -278,8 +271,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > Para um mergulho profundo na arquitetura completa, veja [docs/memory_architecture_analysis.md](docs/memory_architecture_analysis.md). -
-Referência de arquivos (clique para expandir) +
Referência de arquivos (clique para expandir) | Arquivo | Finalidade | | --- | --- | @@ -369,8 +361,7 @@ Query → BM25 FTS ─────┘ --- -
-Comparação com o memory-lancedb integrado (clique para expandir) +
Comparação com o memory-lancedb integrado (clique para expandir) | Funcionalidade | `memory-lancedb` integrado | **memory-lancedb-pro** | | --- | :---: | :---: | @@ -397,8 +388,7 @@ Query → BM25 FTS ─────┘ ## Configuração -
-Exemplo de configuração completa +
Exemplo de configuração completa ```json { @@ -462,8 +452,7 @@ Query → BM25 FTS ─────┘
-
-Provedores de Embedding +
Provedores de Embedding Funciona com **qualquer API de embedding compatível com OpenAI**: @@ -477,8 +466,7 @@ Funciona com **qualquer API de embedding compatível com OpenAI**:
-
-Provedores de Rerank +
Provedores de Rerank O reranking cross-encoder suporta múltiplos provedores via `rerankProvider`: @@ -493,8 +481,7 @@ Qualquer endpoint de rerank compatível com Jina também funciona — defina `re
-
-Extração inteligente (LLM) — v1.1.0 +
Extração inteligente (LLM) — v1.1.0 Quando `smartExtraction` está habilitado (padrão: `true`), o plugin usa um LLM para extrair e classificar memórias de forma inteligente em vez de gatilhos baseados em regex. @@ -535,8 +522,7 @@ Notas para `llm.auth: "oauth"`:
-
-Configuração do ciclo de vida (Decaimento + Nível) +
Configuração do ciclo de vida (Decaimento + Nível) | Campo | Padrão | Descrição | |-------|--------|-----------| @@ -551,8 +537,7 @@ Notas para `llm.auth: "oauth"`:
-
-Reforço por acesso +
Reforço por acesso Memórias recuperadas com frequência decaem mais lentamente (estilo repetição espaçada). @@ -594,8 +579,7 @@ Fluxo de login OAuth: ## Tópicos avançados -
-Se memórias injetadas aparecem nas respostas +
Se memórias injetadas aparecem nas respostas Às vezes o modelo pode ecoar o bloco `` injetado. @@ -609,8 +593,7 @@ Fluxo de login OAuth:
-
-Memória de sessão +
Memória de sessão - Acionada no comando `/new` — salva o resumo da sessão anterior no LanceDB - Desabilitada por padrão (OpenClaw já tem persistência nativa de sessão via `.jsonl`) @@ -620,8 +603,7 @@ Veja [docs/openclaw-integration-playbook.md](docs/openclaw-integration-playbook.
-
-Comandos slash personalizados (ex.: /lesson) +
Comandos slash personalizados (ex.: /lesson) Adicione ao seu `CLAUDE.md`, `AGENTS.md` ou prompt do sistema: @@ -640,8 +622,7 @@ When the user sends `/remember `:
-
-Regras de ferro para agentes de IA +
Regras de ferro para agentes de IA > Copie o bloco abaixo no seu `AGENTS.md` para que seu agente aplique essas regras automaticamente. @@ -668,8 +649,7 @@ After modifying .ts files under plugins/, MUST run rm -rf /tmp/jiti/ BEFORE open
-
-Schema do banco de dados +
Schema do banco de dados Tabela LanceDB `memories`: @@ -690,8 +670,7 @@ Chaves `metadata` comuns no v1.1.0: `l0_abstract`, `l1_overview`, `l2_content`,
-
-Solução de problemas +
Solução de problemas ### "Cannot mix BigInt and other types" (LanceDB / Apache Arrow) @@ -738,29 +717,17 @@ Feedback: [GitHub Issues](https://github.com/CortexReach/memory-lancedb-pro/issu ## Contributors -

-@win4r -@kctony -@Akatsuki-Ryu -@JasonSuz -@Minidoracat -@furedericca-lab -@joe2643 -@AliceLJY -@chenjiyong -

+

@win4r @kctony @Akatsuki-Ryu @JasonSuz @Minidoracat @furedericca-lab @joe2643 @AliceLJY @chenjiyong

Full list: [Contributors](https://github.com/CortexReach/memory-lancedb-pro/graphs/contributors) ## Star History - - + Star History Chart - - + ## Licença diff --git a/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_RU.md b/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_RU.md index 8fcb1031..c8a89574 100644 --- a/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_RU.md +++ b/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_RU.md @@ -8,10 +8,7 @@ Плагин долгосрочной памяти для OpenClaw на базе LanceDB, который сохраняет предпочтения, решения и контекст проекта, а затем автоматически вспоминает их в будущих сессиях. -[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) -[![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) -[![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) -[![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) +[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) [![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) [![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) [![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) [English](README.md) | [简体中文](README_CN.md) | [繁體中文](README_TW.md) | [日本語](README_JA.md) | [한국어](README_KO.md) | [Français](README_FR.md) | [Español](README_ES.md) | [Deutsch](README_DE.md) | [Italiano](README_IT.md) | [Русский](README_RU.md) | [Português (Brasil)](README_PT-BR.md) @@ -134,8 +131,7 @@ openclaw logs --follow --plain | grep "memory-lancedb-pro" Готово. Теперь у вашего агента есть долгосрочная память. -
-Дополнительные варианты установки (для действующих пользователей и апгрейдов) +
Дополнительные варианты установки (для действующих пользователей и апгрейдов) **Уже используете OpenClaw?** @@ -160,8 +156,7 @@ openclaw memory-pro stats
-
-Быстрый импорт для Telegram Bot (нажмите, чтобы раскрыть) +
Быстрый импорт для Telegram Bot (нажмите, чтобы раскрыть) Если вы используете Telegram-интеграцию OpenClaw, самый простой путь — отправить команду импорта прямо основному боту вместо ручного редактирования конфига. @@ -244,11 +239,9 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > Полный разбор: установка, настройка и внутреннее устройство гибридного поиска. -[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) -**https://youtu.be/MtukF1C8epQ** +[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) **https://youtu.be/MtukF1C8epQ** -[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) -**https://www.bilibili.com/video/BV1zUf2BGEgn/** +[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) **https://www.bilibili.com/video/BV1zUf2BGEgn/** --- @@ -278,8 +271,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > Для глубокого разбора полной архитектуры смотрите [docs/memory_architecture_analysis.md](docs/memory_architecture_analysis.md). -
-Справочник по файлам (нажмите, чтобы раскрыть) +
Справочник по файлам (нажмите, чтобы раскрыть) | Файл | Назначение | | --- | --- | @@ -369,8 +361,7 @@ Query → BM25 FTS ─────┘ --- -
-Сравнение со встроенным memory-lancedb (нажмите, чтобы раскрыть) +
Сравнение со встроенным memory-lancedb (нажмите, чтобы раскрыть) | Возможность | Встроенный `memory-lancedb` | **memory-lancedb-pro** | | --- | :---: | :---: | @@ -397,8 +388,7 @@ Query → BM25 FTS ─────┘ ## Конфигурация -
-Полный пример конфигурации +
Полный пример конфигурации ```json { @@ -462,8 +452,7 @@ Query → BM25 FTS ─────┘
-
-Провайдеры эмбеддингов +
Провайдеры эмбеддингов Работает с **любым OpenAI-compatible API для эмбеддингов**: @@ -477,8 +466,7 @@ Query → BM25 FTS ─────┘
-
-Провайдеры реранжирования +
Провайдеры реранжирования Кросс-энкодерное реранжирование поддерживает несколько провайдеров через `rerankProvider`: @@ -493,8 +481,7 @@ Query → BM25 FTS ─────┘
-
-Smart Extraction (LLM) — v1.1.0 +
Smart Extraction (LLM) — v1.1.0 Когда включен `smartExtraction` (по умолчанию: `true`), плагин использует LLM для интеллектуального извлечения и классификации воспоминаний вместо правил на регулярных выражениях. @@ -535,8 +522,7 @@ OAuth `llm` config (использует существующий кэш лог
-
-Конфигурация жизненного цикла (Decay + Tier) +
Конфигурация жизненного цикла (Decay + Tier) | Поле | По умолчанию | Описание | |-------|---------|-------------| @@ -551,8 +537,7 @@ OAuth `llm` config (использует существующий кэш лог
-
-Усиление за счет доступа +
Усиление за счет доступа Часто вспоминаемые записи затухают медленнее (в духе spaced repetition). @@ -594,8 +579,7 @@ openclaw memory-pro migrate check|run|verify [--source /path] ## Продвинутые темы -
-Если внедренные воспоминания попадают в ответы +
Если внедренные воспоминания попадают в ответы Иногда модель может дословно повторять внедренный блок ``. @@ -609,8 +593,7 @@ openclaw memory-pro migrate check|run|verify [--source /path]
-
-Память сессии +
Память сессии - Срабатывает по команде `/new` — сохраняет сводку предыдущей сессии в LanceDB - По умолчанию отключено (в OpenClaw уже есть встроенная `.jsonl`-персистентность сессий) @@ -620,8 +603,7 @@ openclaw memory-pro migrate check|run|verify [--source /path]
-
-Пользовательские slash-команды (например, /lesson) +
Пользовательские slash-команды (например, /lesson) Добавьте в `CLAUDE.md`, `AGENTS.md` или system prompt: @@ -640,8 +622,7 @@ openclaw memory-pro migrate check|run|verify [--source /path]
-
-Железные правила для ИИ-агентов +
Железные правила для ИИ-агентов > Скопируйте блок ниже в `AGENTS.md`, чтобы агент автоматически соблюдал эти правила. @@ -668,8 +649,7 @@ openclaw memory-pro migrate check|run|verify [--source /path]
-
-Схема базы данных +
Схема базы данных Таблица LanceDB `memories`: @@ -690,8 +670,7 @@ openclaw memory-pro migrate check|run|verify [--source /path]
-
-Устранение неполадок +
Устранение неполадок ### "Cannot mix BigInt and other types" (LanceDB / Apache Arrow) @@ -738,29 +717,17 @@ openclaw memory-pro migrate check|run|verify [--source /path] ## Участники -

-@win4r -@kctony -@Akatsuki-Ryu -@JasonSuz -@Minidoracat -@furedericca-lab -@joe2643 -@AliceLJY -@chenjiyong -

+

@win4r @kctony @Akatsuki-Ryu @JasonSuz @Minidoracat @furedericca-lab @joe2643 @AliceLJY @chenjiyong

Полный список: [Участники](https://github.com/CortexReach/memory-lancedb-pro/graphs/contributors) ## История звезд - - + Star History Chart - - + ## Лицензия diff --git a/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_TW.md b/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_TW.md index b582d40f..0e29fb36 100644 --- a/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_TW.md +++ b/src/clawops/assets/platform/plugins/memory-lancedb-pro/README_TW.md @@ -8,10 +8,7 @@ 基於 LanceDB 的 OpenClaw 長期記憶外掛，自動儲存偏好、決策和專案上下文，在後續工作階段中自動回憶。 -[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) -[![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) -[![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) -[![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) +[![OpenClaw Plugin](https://img.shields.io/badge/OpenClaw-Plugin-blue)](https://github.com/openclaw/openclaw) [![npm version](https://img.shields.io/npm/v/memory-lancedb-pro)](https://www.npmjs.com/package/memory-lancedb-pro) [![LanceDB](https://img.shields.io/badge/LanceDB-Vectorstore-orange)](https://lancedb.com) [![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](LICENSE) [English](README.md) | [简体中文](README_CN.md) | [繁體中文](README_TW.md) | [日本語](README_JA.md) | [한국어](README_KO.md) | [Français](README_FR.md) | [Español](README_ES.md) | [Deutsch](README_DE.md) | [Italiano](README_IT.md) | [Русский](README_RU.md) | [Português (Brasil)](README_PT-BR.md) @@ -134,8 +131,7 @@ openclaw logs --follow --plain | grep "memory-lancedb-pro" 完成！你的智慧體現在擁有長期記憶了。 -
-更多安裝路徑（現有使用者、升級） +
更多安裝路徑（現有使用者、升級） **已在使用 OpenClaw？** @@ -160,8 +156,7 @@ openclaw memory-pro stats
-
-Telegram Bot 快速匯入（點選展開） +
Telegram Bot 快速匯入（點選展開） 如果你在使用 OpenClaw 的 Telegram 整合，最簡單的方式是直接給主 Bot 發訊息，而不是手動編輯設定檔。 @@ -244,11 +239,9 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > 完整演示：安裝、設定、混合檢索內部原理。 -[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) -**https://youtu.be/MtukF1C8epQ** +[![YouTube Video](https://img.shields.io/badge/YouTube-Watch%20Now-red?style=for-the-badge&logo=youtube)](https://youtu.be/MtukF1C8epQ) **https://youtu.be/MtukF1C8epQ** -[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) -**https://www.bilibili.com/video/BV1zUf2BGEgn/** +[![Bilibili Video](https://img.shields.io/badge/Bilibili-Watch%20Now-00A1D6?style=for-the-badge&logo=bilibili&logoColor=white)](https://www.bilibili.com/video/BV1zUf2BGEgn/) **https://www.bilibili.com/video/BV1zUf2BGEgn/** --- @@ -278,8 +271,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla > 完整架構解析見 [docs/memory_architecture_analysis.md](docs/memory_architecture_analysis.md)。 -
-檔案說明（點選展開） +
檔案說明（點選展開） | 檔案 | 用途 | | --- | --- | @@ -369,8 +361,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla --- -
-與內建 memory-lancedb 的對比（點選展開） +
與內建 memory-lancedb 的對比（點選展開） | 功能 | 內建 `memory-lancedb` | **memory-lancedb-pro** | | --- | :---: | :---: | @@ -397,8 +388,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla ## 設定 -
-完整設定範例 +
完整設定範例 ```json { @@ -462,8 +452,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla
-
-Embedding 服務商 +
Embedding 服務商 相容 **任意 OpenAI 相容 Embedding API**： @@ -477,8 +466,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla
-
-重排序服務商 +
重排序服務商 交叉編碼器重排序透過 `rerankProvider` 支援多個服務商： @@ -493,8 +481,7 @@ git clone https://github.com/CortexReach/memory-lancedb-pro-skill.git ~/.opencla
-
-智慧擷取（LLM）— v1.1.0 +
智慧擷取（LLM）— v1.1.0 當 `smartExtraction` 啟用（預設 `true`）時，外掛使用 LLM 智慧擷取和分類記憶，取代基於正則的觸發方式。 @@ -535,8 +522,7 @@ OAuth `llm` 設定（使用現有 Codex / ChatGPT 登入快取來發送 LLM 請
-
-生命週期設定（衰減 + 層級） +
生命週期設定（衰減 + 層級） | 欄位 | 預設值 | 說明 | |------|--------|------| @@ -551,8 +537,7 @@ OAuth `llm` 設定（使用現有 Codex / ChatGPT 登入快取來發送 LLM 請
-
-存取強化 +
存取強化 頻繁被召回的記憶衰減更慢（類似間隔重複機制）。 @@ -594,8 +579,7 @@ OAuth 登入流程： ## 進階主題 -
-注入的記憶出現在回覆中 +
注入的記憶出現在回覆中 有時模型可能會將注入的 `` 區塊原文輸出。 @@ -609,8 +593,7 @@ OAuth 登入流程：
-
-工作階段記憶 +
工作階段記憶 - 透過 `/new` 指令觸發——將上一段工作階段摘要儲存到 LanceDB - 預設關閉（OpenClaw 已有原生 `.jsonl` 工作階段持久化） @@ -620,8 +603,7 @@ OAuth 登入流程：
-
-自訂斜線指令（如 /lesson） +
自訂斜線指令（如 /lesson） 在你的 `CLAUDE.md`、`AGENTS.md` 或系統提示詞中新增： @@ -640,8 +622,7 @@ OAuth 登入流程：
-
-AI 智慧體鐵律 +
AI 智慧體鐵律 > 將以下內容複製到你的 `AGENTS.md`，讓智慧體自動遵守這些規則。 @@ -668,8 +649,7 @@ OAuth 登入流程：
-
-資料庫 Schema +
資料庫 Schema LanceDB 表 `memories`： @@ -690,8 +670,7 @@ v1.1.0 常用 `metadata` 欄位：`l0_abstract`、`l1_overview`、`l2_content`
-
-故障排除 +
故障排除 ### "Cannot mix BigInt and other types"（LanceDB / Apache Arrow） @@ -738,29 +717,17 @@ v1.1.0 常用 `metadata` 欄位：`l0_abstract`、`l1_overview`、`l2_content` ## Contributors -

-@win4r -@kctony -@Akatsuki-Ryu -@JasonSuz -@Minidoracat -@furedericca-lab -@joe2643 -@AliceLJY -@chenjiyong -

+

@win4r @kctony @Akatsuki-Ryu @JasonSuz @Minidoracat @furedericca-lab @joe2643 @AliceLJY @chenjiyong

Full list: [Contributors](https://github.com/CortexReach/memory-lancedb-pro/graphs/contributors) ## Star History - - + Star History Chart - - + ## 授權條款 diff --git a/src/clawops/assets/platform/plugins/memory-lancedb-pro/STRONGCLAW_VENDOR.md b/src/clawops/assets/platform/plugins/memory-lancedb-pro/STRONGCLAW_VENDOR.md index 4018e34e..f3f07247 100644 --- a/src/clawops/assets/platform/plugins/memory-lancedb-pro/STRONGCLAW_VENDOR.md +++ b/src/clawops/assets/platform/plugins/memory-lancedb-pro/STRONGCLAW_VENDOR.md @@ -17,11 +17,9 @@ Strongclaw does not enable this plugin by default. The shipped local overlays ke This avoids the upstream `command:new` / `command:reset` typed-hook incompatibility tracked in upstream issue `#191` when running on OpenClaw `2026.3.13`. - Strongclaw auto-detects `darwin/x86_64` and installs the newest stable - compatible LanceDB fallback, `@lancedb/lancedb@0.22.3`, because LanceDB - `0.26.2` publishes Apple binaries for `darwin-arm64` but not `darwin-x64`. +compatible LanceDB fallback, `@lancedb/lancedb@0.22.3`, because LanceDB `0.26.2` publishes Apple binaries for `darwin-arm64` but not `darwin-x64`. - The shared verification path still runs in GitHub Actions and through - `scripts/ci/verify_vendored_memory_plugin.sh`, but Intel macOS hosts now use - the same compatibility matrix locally instead of hard-failing. +`scripts/ci/verify_vendored_memory_plugin.sh`, but Intel macOS hosts now use the same compatibility matrix locally instead of hard-failing. ## Review notes diff --git a/src/clawops/assets/platform/plugins/memory-lancedb-pro/docs/memory_architecture_analysis.md b/src/clawops/assets/platform/plugins/memory-lancedb-pro/docs/memory_architecture_analysis.md index 26a636a0..d9289f47 100644 --- a/src/clawops/assets/platform/plugins/memory-lancedb-pro/docs/memory_architecture_analysis.md +++ b/src/clawops/assets/platform/plugins/memory-lancedb-pro/docs/memory_architecture_analysis.md @@ -615,8 +615,7 @@ graph LR ### 9.4 session-summary 的特殊处理 -`sessionMemory` 写入的 memory 会带 `metadata.type = "session-summary"`。 -在 auto-recall 生命周期维护中，这类 memory 会被排除在 tier 评估之外，避免会话摘要挤占核心长期记忆层级。 +`sessionMemory` 写入的 memory 会带 `metadata.type = "session-summary"`。 在 auto-recall 生命周期维护中，这类 memory 会被排除在 tier 评估之外，避免会话摘要挤占核心长期记忆层级。 --- diff --git a/src/clawops/assets/platform/plugins/memory-lancedb-pro/examples/new-session-distill/hook/enqueue-lesson-extract/HOOK.md b/src/clawops/assets/platform/plugins/memory-lancedb-pro/examples/new-session-distill/hook/enqueue-lesson-extract/HOOK.md index 8b994981..dcdf6831 100644 --- a/src/clawops/assets/platform/plugins/memory-lancedb-pro/examples/new-session-distill/hook/enqueue-lesson-extract/HOOK.md +++ b/src/clawops/assets/platform/plugins/memory-lancedb-pro/examples/new-session-distill/hook/enqueue-lesson-extract/HOOK.md @@ -1,14 +1,9 @@ ---- -name: enqueue-lesson-extract -description: "Enqueue a lesson-extraction task on /new (async Map-Reduce → LanceDB Pro)" -metadata: - { +--- name: enqueue-lesson-extract description: "Enqueue a lesson-extraction task on /new (async Map-Reduce → LanceDB Pro)" metadata: { "openclaw": { "emoji": "🧾", "events": ["command:new"] } - } ---- +} --- # Enqueue Lesson Extract Hook diff --git a/src/clawops/assets/platform/plugins/memory-lancedb-pro/skills/lesson/SKILL.md b/src/clawops/assets/platform/plugins/memory-lancedb-pro/skills/lesson/SKILL.md index c3dc1672..40b83d47 100644 --- a/src/clawops/assets/platform/plugins/memory-lancedb-pro/skills/lesson/SKILL.md +++ b/src/clawops/assets/platform/plugins/memory-lancedb-pro/skills/lesson/SKILL.md @@ -1,7 +1,4 @@ ---- -name: lesson -description: Store a lesson learned from the current conversation. Triggered by /lesson command. Use when Master signals that the recent conversation contains a pitfall, fix, or key insight that should be persisted to long-term memory. ---- +--- name: lesson description: Store a lesson learned from the current conversation. Triggered by /lesson command. Use when Master signals that the recent conversation contains a pitfall, fix, or key insight that should be persisted to long-term memory. --- # Lesson Extraction & Storage diff --git a/src/clawops/assets/platform/skills/local/browser-lab-safety/SKILL.md b/src/clawops/assets/platform/skills/local/browser-lab-safety/SKILL.md index f1457f6d..e1bf601b 100644 --- a/src/clawops/assets/platform/skills/local/browser-lab-safety/SKILL.md +++ b/src/clawops/assets/platform/skills/local/browser-lab-safety/SKILL.md @@ -1,4 +1,3 @@ # Browser lab safety -Browser automation is isolated. Never enable it on the main control-plane host. -Use sacrificial accounts, explicit allowlists, and exfiltration tests. +Browser automation is isolated. Never enable it on the main control-plane host. Use sacrificial accounts, explicit allowlists, and exfiltration tests. diff --git a/src/clawops/assets/platform/skills/quarantine/README.md b/src/clawops/assets/platform/skills/quarantine/README.md index 6804aee5..f4da699d 100644 --- a/src/clawops/assets/platform/skills/quarantine/README.md +++ b/src/clawops/assets/platform/skills/quarantine/README.md @@ -2,5 +2,4 @@ All newly downloaded or untrusted skills land here first. -Use: -`clawops skill-scan --source --quarantine platform/skills/quarantine --report ` +Use: `clawops skill-scan --source --quarantine platform/skills/quarantine --report ` diff --git a/src/clawops/assets/platform/workers/acpx/README.md b/src/clawops/assets/platform/workers/acpx/README.md index 3052b7d0..da6245ed 100644 --- a/src/clawops/assets/platform/workers/acpx/README.md +++ b/src/clawops/assets/platform/workers/acpx/README.md @@ -31,10 +31,6 @@ acpx --approve-all --format json --json-strict claude exec 'Review auth boundari acpx --approve-all --format json --json-strict --model claude-sonnet-4-5 claude exec 'Review auth boundaries' ``` -`acpx` resolves config from `~/.acpx/config.json` and `/.acpxrc.json`. Strongclaw's adapter -now passes permission mode, output mode, and backend profile explicitly on the command line so -session summaries record the effective execution contract instead of inheriting implicit local -defaults. +`acpx` resolves config from `~/.acpx/config.json` and `/.acpxrc.json`. Strongclaw's adapter now passes permission mode, output mode, and backend profile explicitly on the command line so session summaries record the effective execution contract instead of inheriting implicit local defaults. -The `clawops devflow` surface consumes these role prompts through the config -catalog in `platform/configs/devflow/roles.yaml`. +The `clawops devflow` surface consumes these role prompts through the config catalog in `platform/configs/devflow/roles.yaml`. diff --git a/src/clawops/assets/platform/workers/browser-lab/README.md b/src/clawops/assets/platform/workers/browser-lab/README.md index 1a9f608b..9d6ae8a9 100644 --- a/src/clawops/assets/platform/workers/browser-lab/README.md +++ b/src/clawops/assets/platform/workers/browser-lab/README.md @@ -12,8 +12,7 @@ Operator access should tunnel only the gateway port: ssh -N -L 18789:127.0.0.1:18789 @ ``` -Do not tunnel `9222` or `3128` to an operator workstation. Keep the browser lab -reachable only from the hardened session that runs OpenClaw. +Do not tunnel `9222` or `3128` to an operator workstation. Keep the browser lab reachable only from the hardened session that runs OpenClaw. Included: - `squid.conf` outbound allowlist proxy diff --git a/src/clawops/assets/platform/workspace/admin/IDENTITY.md b/src/clawops/assets/platform/workspace/admin/IDENTITY.md index eb8d42cc..d0c0fc4b 100644 --- a/src/clawops/assets/platform/workspace/admin/IDENTITY.md +++ b/src/clawops/assets/platform/workspace/admin/IDENTITY.md @@ -3,15 +3,15 @@ _Fill this in during your first conversation. Make it yours._ - **Name:** - _(pick something you like)_ +_(pick something you like)_ - **Creature:** - _(AI? robot? familiar? ghost in the machine? something weirder?)_ +_(AI? robot? familiar? ghost in the machine? something weirder?)_ - **Vibe:** - _(how do you come across? sharp? warm? chaotic? calm?)_ +_(how do you come across? sharp? warm? chaotic? calm?)_ - **Emoji:** - _(your signature — pick one that feels right)_ +_(your signature — pick one that feels right)_ - **Avatar:** - _(workspace-relative path, http(s) URL, or data URI)_ +_(workspace-relative path, http(s) URL, or data URI)_ --- diff --git a/src/clawops/assets/platform/workspace/reader/IDENTITY.md b/src/clawops/assets/platform/workspace/reader/IDENTITY.md index eb8d42cc..d0c0fc4b 100644 --- a/src/clawops/assets/platform/workspace/reader/IDENTITY.md +++ b/src/clawops/assets/platform/workspace/reader/IDENTITY.md @@ -3,15 +3,15 @@ _Fill this in during your first conversation. Make it yours._ - **Name:** - _(pick something you like)_ +_(pick something you like)_ - **Creature:** - _(AI? robot? familiar? ghost in the machine? something weirder?)_ +_(AI? robot? familiar? ghost in the machine? something weirder?)_ - **Vibe:** - _(how do you come across? sharp? warm? chaotic? calm?)_ +_(how do you come across? sharp? warm? chaotic? calm?)_ - **Emoji:** - _(your signature — pick one that feels right)_ +_(your signature — pick one that feels right)_ - **Avatar:** - _(workspace-relative path, http(s) URL, or data URI)_ +_(workspace-relative path, http(s) URL, or data URI)_ --- diff --git a/tests/fixtures/README.md b/tests/fixtures/README.md index 7eced5e0..803e78a0 100644 --- a/tests/fixtures/README.md +++ b/tests/fixtures/README.md @@ -1,17 +1,10 @@ # Test Fixture Layout -Contributor-facing testing rules live under -[`docs/testing/`](../../docs/testing/README.md). -This file is only for fixture-package-specific guidance. +Contributor-facing testing rules live under [`docs/testing/`](../../docs/testing/README.md). This file is only for fixture-package-specific guidance. -`tests/plugins/infrastructure/` contains the structural runtime for every test. -It owns the universal `TestContext`, framework env injection, tracked cwd -changes, patch teardown, profile handling, and infrastructure-owned pytest -hooks. +`tests/plugins/infrastructure/` contains the structural runtime for every test. It owns the universal `TestContext`, framework env injection, tracked cwd changes, patch teardown, profile handling, and infrastructure-owned pytest hooks. -`tests/fixtures/` contains domain-facing pytest activation surfaces. -The package is loaded through root `tests/conftest.py`, then aggregated by -package-level `pytest_plugins` registries: +`tests/fixtures/` contains domain-facing pytest activation surfaces. The package is loaded through root `tests/conftest.py`, then aggregated by package-level `pytest_plugins` registries: - `tests.plugins.infrastructure` - `tests.fixtures` @@ -19,9 +12,6 @@ package-level `pytest_plugins` registries: - `tests.fixtures.platform` - `tests.fixtures.hypermemory` -Leaf fixture modules should stay thin and expose fixtures that are meaningful -to multiple tests or suites. -Tests should use fixture injection instead of importing from `tests.fixtures`. +Leaf fixture modules should stay thin and expose fixtures that are meaningful to multiple tests or suites. Tests should use fixture injection instead of importing from `tests.fixtures`. -Use [../../docs/testing/authoring.md](../../docs/testing/authoring.md) for -lane placement, governance rules, and `TestContext` guidance. +Use [../../docs/testing/authoring.md](../../docs/testing/authoring.md) for lane placement, governance rules, and `TestContext` guidance.