Repository navigation
Releases
Releases · aiconnai/engram
v0.24.0
Engram v0.24.0
Release candidate prepared for the v0.24.0 GitHub Release (v0.21.2 remains the latest published release until the signed v0.24.0 tag is pushed; registry publication to crates.io, PyPI, and npm is a separate decision).
Security
palace_visualizesaving is a filesystem write — a call withoutput_pathnow requiresscoped_writewithinENGRAM_EXPORT_BASE_DIR, otherwisemaintenance, likememory_export_markdown, andpermission_mode_statusexposesoutput_pathinwrite_flag_params. Paths outside the export base are refused before any directory is created,..components are rejected unconditionally, active SQLite artifacts are refused both before directory creation and after parent canonicalization across CLI and MCP callers, symlinks and multi-link files (nlink > 1) are rejected, and outputs are published via an owner-only (0600) sibling temp file and atomic rename.GET /v1/eventsis workspace-scoped for restricted principals — a keyless (anonymous loopback) stream opened withoutworkspacenow receivesdefaultevents only instead of every workspace's events and content previews.- Context assembly and graph expansion stay in scope —
memory_build_context,memory_get_injection_prompt, andmemory_prepare_contextnormalize and authorizeworkspace,workspaces, andfilters.workspaceconsistently with central MCP permission checks, keep restricted principals within their allowed workspace, advancedepth2–3 graph frontiers only through authorized in-scope memories, and filterinclude_graphcrossrefsso foreign-workspace edges are never exposed. - Keyless HTTP and CORS align on
HostandOrigin— without an API key,/mcp,/v1/mcpand/v1/eventsanswer403to requests whoseHostorOrigindoes not name this machine (DNS rebinding, cross-site requests).ENGRAM_HTTP_ALLOWED_HOSTSextends theHostallowlist; blankENGRAM_CORS_ORIGINSnormalizes to unset, empty comma tokens and emptyOriginheaders are rejected, and bothCorsLayerpreflight (OPTIONS) and keyless checks accept case-insensitive loopbackOriginURIs (localhost,127.0.0.1,[::1], any port) by default; see MCP_AUTH. - Multimodal path confinement and handler scope guards (
A-5,A-6,A-7,Q-A) —memory_search_by_imagerequiresENGRAM_MEDIA_BASE_DIRinread_only/scoped_writemodes (escalating tomaintenancewhen unset or blank), all media path validators refuse active SQLite artifacts before canonicalization,memory_digestauthorizes innercontext_build_bundlecalls,harness_statusrestrictsinclude_gitto unrestricted writers, andmemory_agent_statusscopestotal_memoriesto the target workspace. - Bundled SQLite 3.54.1 (RUSTSEC-2026-0459, CVE-2026-119375) —
rusqlite0.31 → 0.41.1 andlibsqlite3-sys0.28.0 → 0.39.1 in the root andfuzz/lockfiles. The previous bundled SQLite 3.45.0 could return uninitialized page-cache memory for a database opened with a-walof a different page size. No schema, MCP or SDK change;rusqlitekeeps the same feature surface (cacheandfallible_uintare now explicit opt-ins). - Markdown filesystem import (RISK-0011) — scoped writers require an administrator-configured root; unconfined reads require maintenance even during review. Bounded directory-relative reads reject links, special files and SQLite artifacts before opening regular files. Existing IDs are authorized by their stored workspace, cross-workspace ID overwrites against an explicit target
workspaceare rejected, and confirmed batches authorize/classify/apply within one transaction. Pre-commit failures roll back; completion uncertainty is explicit. See compatibility notes. - Markdown filesystem export —
memory_export_markdownis a write operation: scoped writers need an administrator-configuredENGRAM_EXPORT_BASE_DIR; unconfined destinations require maintenance privileges. Linux/macOS directory-relative writes reject unsafe components, symlinks, hard links, active SQLite artifacts and directory ACLs, and publish private temporary files by atomic rename. Other platforms fail closed. See compatibility notes. - Batch tools enforce per-item workspaces —
memory_create_batchandmemory_ingest_fact_batchauthorize every item'sworkspace(central check, gRPC pre-check and handlers), closing a cross-workspace write introduced by the search/auth hardening commite04d3307. Batch item arrays are derived from the catalog schema and fail closed on unreadable nestedworkspaceclaims. - Resource subscriptions are principal-aware —
resources/subscribeandresources/unsubscribeapply the same workspace policy asresources/read. memory_search_suggestis workspace-scoped — restricted principals get spelling corrections built only from their own workspace vocabulary.permission_denial_for_modefails closed for unknown tools.- One
tools/callunwrap point — HTTP and gRPC unwrap the envelope once and authorize the same bare arguments the dispatcher sees;name/argumentskeys inside real tool arguments are no longer treated as an envelope. A caller's per-call_permission_modeself-restriction is still applied by the dispatcher, so its denial stays in the tool result (HTTP 200). - Log redaction (RISK-0002) — sync/cloud, hooks, multimodal,
engram-serverHTTP/WebSocket transport startup errors, and the watcher (including its daemon) no longer log bucket/object keys, workspace and tool names, filesystem paths,engram_url, dry-run memory content, subprocess stderr or raw error text; errors are logged by class (ENGRAM_LOG_ERROR_DETAIL=1restores detail for debugging). - Cryptographic release-tag verification (
R3) — Addedscripts/verify-release-tag.shanddocs/releases/allowed-signers, wired into.github/workflows/sdk-release.ymlacrosspreflight,publish-python, andpublish-npmwith anti-self-authorization and anti-ref-shadowing (materializing both verifier script and trust store from canonicalrefs/remotes/origin/mainvia--end-of-options), and hardenedscripts/sign-candidate-manifest.shoption/PEM/leading-dash guards. - Lifecycle, policy, compression, and server hook authorization (
Unit 1,Unit 3) — Added 12 lifecycle/policy/compression by-ID tools (memory_set_lifecycle,memory_lifecycle_update,memory_score,memory_promote,memory_explain,memory_reconcile_conflict,memory_promote_to_permanent,memory_boost,memory_set_expiration,memory_compress,memory_decompress,memory_compress_for_context) toID_SCOPED_TOOLSwith in-transactionensure_memory_accesschecks and single-IDdecayscoping;engram-serverlifecycle hooks now enforce_permission_mode/ENGRAM_PERMISSION_MODE, requirePermission::WriteonResourceType::Memory, reject unauthorized fallback to"default", and skipPostToolUsereinforcement on error responses. - Dependency advisories — zip 9.0.5 and duckdb 1.10506.1 (RUSTSEC-2026-0367/0368/0369), tokio 1.53.3 (RUSTSEC-2026-0386), rustls-webpki 0.103.16, webpki-roots 1.0.11 and ctutils 0.4.6 (RUSTSEC-2026-0382/0381/0385), rustix 1.1.6 (RUSTSEC-2026-0486), quinn-proto 0.11.19 (RUSTSEC-2026-0523/0524), and quinn-udp 0.5.16 (RUSTSEC-2026-0538); crypto-bigint 0.5 removed from the graph (RUSTSEC-2026-0387) by not enabling
aws-sdk-s3/sigv4a.
Added
- RFC 0009, 0010, and 0011 Implemented — Promoted and wired MCP Contract Validation (
docs/rfcs/0009-mcp-contract-validation.md), Permission Modes (docs/rfcs/0010-permission-modes.md), and Deterministic Model Routing (docs/rfcs/0011-model-routing.md,governance/models.toml,scripts/select-model.py). - MCP Error Normalizer (
G-6,G-7,G-10) — Addedsrc/mcp/error_normalizer.rsand upgradedcontext/blocks.rs,compression.rs,dream.rs,memory_crud/read_update_delete.rs, andsearch.rsso wrong-typedid/limit/offsetparameters and domain errors emit RFC 0006 structuredToolErrorpayloads (invalid_params,memory_not_found,validation_error). - Zero-Drift Python & TypeScript SDKs (
v0.7.0,G-2,G-4,G-5) — Aligned all Python and TypeScript SDK resource methods with the MCP server tool registry, routedresources/*calls over JSON-RPC (_rpc_call/rpcCall), ratcheteddocs/quality/sdk-contract-drift-baseline.jsonfrom 161 entries down to0(known_drift: []), and added TypeScriptisError: truestructuredEngramError(message, code, details)plustimeout/connection_errorparity. - Server Lifecycle Hooks & Cloud Sync Worker (
#237,C6) — WiredSessionStartHandler,PostToolUseHandler, andSyncWorker::start_backgroundintoengram-serverstartup and shutdown lifecycle.
Changed
- Search (from
e04d3307): catalog defaultlimitis 20,offsetandinclude_archivedare accepted,SearchOptions.offsetis public, andmemory_search_by_identity/memory_session_searchare read-only. Python and TypeScript SDKs defaultsearchto limit 20 and acceptoffset(Python alsoinclude_archived). - HNSW is opt-in —
hnsw_enableddefaults tofalseuntil the benchmark required by RFC 0003 exists; enable it withENGRAM_HNSW_ENABLED/--hnsw-enabled. When off, no index is built at startup or maintained on writes. HNSW widening stops after two empty hydration rounds. - Snapshot loader reads archive entry names from the central directory and rejects undecodable names.
- DuckDB bundled engine 1.4.4 → 1.5.6;
sqlite-vecis a dev-dependency (benchmarks only). - CI pins cargo-audit 0.22.0 and cargo-outdated 0.19.0 (
--locked); the TypeScript live lane uses Node 24; the docs job runs the public-claims, advertised-contracts and release-version checks. bump-version.pyverifies lockfiles and__version__and adds ...
v0.21.2
Engram v0.21.2
Release date: 2026-06-16
Changed
- Release notes source — GitHub Release bodies now come from the reviewed
CHANGELOG.mdsection for the requested tag instead of GitHub-generated
notes, preventing reruns from expanding old PR lists. - Release automation hygiene — Release preflight now blocks existing tags
unlessoverwrite_assets=trueis explicitly requested, and Homebrew formula
updates skip no-op commits while checking for the tap token up front. - Public claims hygiene — README, roadmap, and release docs now use more
conservative wording and document which public-use claims still need
publication-time fact-checking.
Full changelog: v0.21.1...v0.21.2
v0.21.1
Engram 0.21.1
This patch release focuses on release readiness, public documentation hygiene, and the first extracted product strategy notes from the old competitive-research branch.
Highlights
- Adds conservative public strategy documentation under
docs/strategy/and links it from the docs index and roadmap. - Refreshes README, roadmap, AI guide, and changelog wording to avoid stale transport, version, performance, and competitor-positioning claims.
- Keeps required GitHub PR checks reportable for documentation-focused changes.
- Updates release automation so Homebrew tap updates run when
HOMEBREW_TAP_TOKENis configured and skip cleanly when it is not. - Updates release workflow actions for the GitHub Actions Node 20 to Node 24 migration.
Verification
git diff --checkbash docs/harness/bin/doctor.shcargo publish -p engram-core --dry-run --allow-dirtymake ci- GitHub required checks passed on PRs #81, #82, #83, and #84.
- Manual Release workflow with
HOMEBREW_TAP_TOKENpassed and pushed the Homebrew formula.
Distribution
- Crate:
engram-core = 0.21.1 - GitHub release assets: macOS and Linux tarballs for x86_64 and aarch64
- Homebrew tap:
aiconnai/homebrew-engramupdated to the published asset checksums
Public marketing, competitive, performance, and benchmark claims should still be fact-checked before reuse outside the repository docs.
Full changelog: v0.21.0...v0.21.1
v0.21.0
v0.20.0
v0.15.0
v0.11.0 — Cognitive Memory Engine
Engram v0.11.0 — Cognitive Memory Engine
All 19 phases complete. 672+ tests, schema v30, 207+ MCP tools.
New Phases (E–K)
- Phase E: Memory Compression & Consolidation — 30x semantic compression, online synthesis, sleep-time consolidation
- Phase F: Agentic Memory Evolution — Historical memory update, retrieval utility scoring (MemRL Q-values), emotional & reflective memory
- Phase G: Advanced Graph Intelligence — Conflict detection/resolution, Hebbian coactivation, semantic triplet matching (SPARQL-like)
- Phase H: Autonomous Memory Agent — Proactive acquisition, 4-pass gardening pipeline, observe→decide→act agent loop,
engram-agentbinary - Phase I: Retrieval Excellence — Multi-provider embeddings (Ollama, Cohere, Voyage, ONNX), MMR diversity, semantic cache, neural reranking, explainability, feedback loop
- Phase J: Context Engineering — Fact extraction, prompt construction (3 strategies), self-editing memory blocks (Letta-inspired)
- Phase K: Temporal Graph & Platform Maturity — Temporal knowledge graph with validity periods, 5-level hierarchical scoping, LOCOMO/LongMemEval/MemBench benchmark suite
Highlights
- 207+ MCP tools (46 new across 7 feature flags)
- Schema v30 (15 new tables, 11 new columns)
- 672+ tests passing
- New binary:
engram-agentfor autonomous memory management - New binary:
engram-benchfor standardized benchmarks - SDKs updated to v0.2.0 (Python + TypeScript) with all new methods
Feature Flags
| Flag | Phase | Description |
|---|---|---|
compression |
E | Semantic compression & consolidation |
agentic-evolution |
F | Utility scoring, sentiment, reflection |
advanced-graph |
G | Conflicts, coactivation, triplets |
autonomous-agent |
H | Agent loop, gardening, proactive scan |
retrieval-excellence |
I | Multi-provider embeddings, MMR, neural rerank |
context-engineering |
J | Facts, prompt builder, memory blocks |
temporal-graph |
K | Temporal edges, scoping, benchmarks |
Install
cargo install engram-coreOr with all features:
cargo install engram-core --features fullLinks
Full Changelog: v0.6.0...v0.11.0