[integration] v1.4.5 — bug fixes, dead code removal, lifespan refactor#6
Open
Atrv-Shrn wants to merge 18 commits into
Open
[integration] v1.4.5 — bug fixes, dead code removal, lifespan refactor#6Atrv-Shrn wants to merge 18 commits into
Atrv-Shrn wants to merge 18 commits into
Conversation
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
…o lifespan
Move watcher and consolidator startup/shutdown from deprecated
@app.on_event("startup"/"shutdown") handlers into the existing
_app_lifespan async context manager with try/finally.
Also replace 6 hardcoded version string assertions with __version__
import and rename 2 stale test methods.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
#4) - Change env_file_encoding from "utf-8" to "utf-8-sig" so .env files with a UTF-8 BOM are parsed correctly (vault_path no longer gets BOM prefix) - Remove dead obsidian_mode field from Settings class, .env.example, and all test fixtures (never read by any code path) - Add BOM-stripping and BOM-free regression tests - Update CHANGELOG with v1.4.5 section Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
…#3) 6 hardcoded "1.4.3" assertions replaced with __version__ from memstack. Renamed test_health_version_is_1_4_2 → test_health_version_matches_package and test_mcp_server_version_is_1_4_2 → test_mcp_server_version_matches_package. Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
…efactor PR Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
- Bump version from 1.4.4 to 1.4.5 in __init__.py, pyproject.toml, .env.example - Add v1.4.5 changelog entry to README.md - Update health example version string in README.md - Update STATUS.md and progress.md with integration status - Fix ruff format in index.py, test_async_execution.py Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
- README.md: complete rewrite as self-contained user guide with glossary, all-platform instructions, full API reference, e2e walkthrough, troubleshooting with exact fix commands, known warnings - STATUS.md: update with /finish status and v1.4.5 verification - progress.md: append /start-int and /finish entries - uv.lock: version bump from 1.4.4 to 1.4.5 Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Updated .gitignore to track dev files on feat/integration: - reference/ docs, CLAUDE.md, CONTEXT.md, STATUS.md, progress.md - ARCHITECTURE.md, CONTRIBUTING.md, .claude/ specs and settings - tests/test_consolidation.py, tests/test_synthesis.py Updated branch-plan.md for v1.4.5 reactive summit. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
…ng model change Adds SearchIndex.validate_dimension() that checks if existing LanceDB vector dimensions match the current embedding provider. On mismatch, logs a warning and triggers a full reindex from vault. Skips gracefully when no table exists or no provider is configured. Called at startup in both REST app and MCP server lifespan, before vault scan. 5 tests covering no-table, matching, mismatch, no-provider, and embed-failure cases. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Remove embedding_autofallback config field and silent fallback logic. Default embedding provider changed from ollama to fastembed, default model changed from nomic-embed-text to BAAI/bge-small-en-v1.5. When the configured provider is unavailable, raise RuntimeError immediately instead of silently falling back. Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
Combined continuity docs after merging feat/dimension-validation into feat/integration alongside feat/embedding-config. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
- Fix stale embedding defaults and AUTOFALLBACK references in README.md - Remove autofallback mentions from ARCHITECTURE.md - Add v1.4.5 entries for embedding-config and dimension-validation in CHANGELOG.md - Update STATUS.md and progress.md with merged PR status and verification results Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
…milarity_ignore_enabled LLM responses wrapped in ```json``` code fences caused json.loads() to fail, silently falling back to wrong behavior: duplicates in smart write (add instead of merge), skipped extraction in synthesis, and skipped batches in consolidation. Strip fences before parsing in llm.py, synthesis.py, and consolidation.py. The similarity_ignore_enabled flag bypassed LLM consultation for high-similarity matches, discarding genuinely new information. Removed the flag and the early-return block so the LLM always decides between add/merge/update/ignore. The reference architecture requires LLM consultation for all scores at or above the add threshold. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
…egory classification to pipeline
Pagination:
- Add limit/offset params to VaultStore.list_memories(), REST GET /memories, MCP memory_list
- Add MemoryListResponse model with memories, total, limit, offset fields
- Default limit=50; limit=0 returns all (used by consolidator)
- Update cache key from str to (agent_id, limit, offset) tuple
Category classification:
- New intelligence/classify.py with classify_content() calling LLM for categorisation
- 10 categories: food, fitness, work, entertainment, tools, relationships, health, travel, education, other
- Pipeline auto-tags every memory with cat:{category} via classify_content()
- New SearchIndex.find_by_tag() for tag-based memory lookup
- Same-category memories injected into LLM consultation as merge candidates
- consult_llm() now accepts incoming_category parameter
Bug fixes (from bug report analysis):
- Watcher _handle_add_or_update() now always deletes old index entries before re-indexing
- scan_vault() now cleans up orphaned index entries for deleted vault files
- New SearchIndex.list_memory_ids() for orphan cleanup
- LLM consultation prompt improved for categorical grouping and contradiction handling
- LLM consultation now produces merged_content for coherent merge rewrites
- VaultStore.merge() accepts optional merged_body for LLM-rewritten content
- Synthesis prompt updated to preserve specifics that pin down preferences
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The prompt asked for a bare category name but format="json" forces Ollama to produce JSON. This mismatch caused json.loads() to crash on non-JSON output, silently falling back to "other" for every memory. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
…owing The agent_end handler's empty catch block made connection failures and HTTP errors completely invisible — auto-capture would silently stop working whenever the MemStack server was unreachable. Now logs network errors and non-2xx responses via console.error while remaining fire-and-forget (does not block turn completion). Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
…ection Delete the TypeScript bridge plugin (openclaw-bridge/) entirely. The new model connects MemStack to OpenClaw as a native MCP server — the agent calls memory tools explicitly as tool calls, with no hooks, no auto-capture, and no auto-recall. Rewrote docs/connect-openclaw.md with explicit memory_write + tags:["auto-capture"] workflow. Updated README, CONTRIBUTING, CHANGELOG, .gitignore, and spec files. Fixed pre-existing unused import in memories.py. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
…P config The disable-native-memory step was marked optional and placed after the MCP server config step. Without disabling OpenClaw's native memory first, the framework-level pipeline intercepts all memory operations and MemStack tools are never called. Moved to Step 3 (before MCP config) and removed the optional label. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
env_file_encodingfrom"utf-8"to"utf-8-sig"—.envfiles saved with a UTF-8 BOM no longer causeMEMSTACK_VAULT_PATHvalidation errorsobsidian_modeconfig field fromSettings,.env.example, and all test fixtures — the field was declared but never read by any code path__version__import — tests no longer break on version bumps@app.on_eventto_app_lifespanasync context managerMerged PRs
Changes
src/memstack/core/config.py—utf-8-sigencoding, removedobsidian_modesrc/memstack/interfaces/rest/app.py— lifespan refactor, removed 4@app.on_eventhandlerstests/test_config.py— 2 new BOM tests, removedobsidian_modeassertionstests/test_routes_health.py—__version__assertions, renamed test methodtests/test_mcp_server.py—__version__assertions, renamed test methodtests/test_cli.py—__version__assertiontests/test_app.py— lifespan test updatetests/test_vault.py,tests/test_watcher.py— removedobsidian_modefrom fixtures.env.example— removedMEMSTACK_OBSIDIAN_MODE, version header → 1.4.5__init__.py,pyproject.toml— version → 1.4.5README.md— full rewrite as self-contained user guideCHANGELOG.md— v1.4.5 sectionARCHITECTURE.md,CONTRIBUTING.md— updated (gitignored, on disk)STATUS.md,progress.md— updated with integration statusAcceptance criteria verified
ruff check src/ tests/clean,ruff format --check src/ tests/cleangit grep on_event -- src/→ zerogit grep obsidian_mode -- src/ tests/→ zerogit grep '"1\.4\.[0-3]"' -- tests/→ zeroKnown limitations
hostparameter after first creation🤖 Generated with Claude Code