v0.4.18

What's Changed

• doc: add 0.4.17 release blog post by @nicoloboschi in #538
• perf: replace window-function retrieval with UNION ALL + per-bank HNSW indexes by @nicoloboschi in #541
• feat: make recall max query tokens configurable via env var by @nicoloboschi in #544
• feat: add jina-mlx reranker provider for Apple Silicon by @nicoloboschi in #542
• doc: Run Hindsight with Ollama: Local AI Memory, No API Keys Needed by @benfrank241 in #536
• doc: What's New in Hindsight Cloud — Programmatic API Key Management by @benfrank241 in #543
• fix: cancel in-flight async ops when bank is deleted by @nicoloboschi in #545
• fix: register embedded profiles in CLI metadata on daemon start by @nicoloboschi in #546
• fix(openclaw): keep recalled memories out of the visible transcript by @stablegenius49 in #548
• blog: Time-Aware Spreading Activation for Memory Graphs by @benfrank241 in #547
• feat: add MiniMax LLM provider support by @octo-patch in #550
• fix: truncate documents exceeding LiteLLM reranker context limit by @nicoloboschi in #549
• Upgrade GitHub Actions to latest versions by @salmanmkc in #553
• chore: add dependabot config for GitHub Actions updates by @nicoloboschi in #554
• chore(deps): bump actions/cache from 4 to 5 by @dependabot[bot] in #559
• chore(deps): bump actions/setup-go from 5 to 6 by @dependabot[bot] in #558
• chore(deps): bump actions/upload-artifact from 4 to 7 by @dependabot[bot] in #555
• chore(deps): bump actions/checkout from 4 to 6 by @dependabot[bot] in #556
• feat: introduce hindsight-api-slim and hindsight-all-slim packages by @nicoloboschi in #560
• fix: remove broken minimax test and enhance slim smoke tests by @nicoloboschi in #564
• chore(deps): bump actions/setup-node from 4 to 6 by @dependabot[bot] in #557
• docs: revamp sidebar with icon grids and language support by @nicoloboschi in #563
• feat: compound tag filtering via tag_groups by @nicoloboschi in #562

New Contributors

• @stablegenius49 made their first contribution in #548
• @octo-patch made their first contribution in #550
• @salmanmkc made their first contribution in #553
• @dependabot[bot] made their first contribution in #559

Full Changelog: v0.4.17...v0.4.18

v0.4.17

What's Changed

• doc: add 0.4.16 release blog post and changelog by @nicoloboschi in #502
• fix: truncate long bank names in selector dropdown by @dcbouius in #505
• fix: refresh bank list when dropdown is opened by @dcbouius in #504
• fix: openclaw tests + split doc-examples CI per language by @nicoloboschi in #503
• doc: document all missing bank config fields in memory-banks.mdx by @nicoloboschi in #508
• Add on_file_convert_complete extension hook after file-to-markdown conversion by @cdbartholomew in #507
• feat: add source facts token limits to consolidation and recall by @nicoloboschi in #509
• refactor: remove dead code and clarify observations vs mental models by @nicoloboschi in #512
• feat: per-request file parser selection with fallback chains by @nicoloboschi in #514
• feat: observation history tracking and diff UI by @nicoloboschi in #513
• feat: mental model history tracking and UI diff view by @nicoloboschi in #516
• doc: Upgrading OpenClaw's Memory with Hindsight by @benfrank241 in #515
• Fix GCS auth for Workload Identity Federation credentials by @cdbartholomew in #518
• feat: change tags for a document by @nicoloboschi in #517
• fix: serialize Alembic upgrades in-process by @cdbartholomew in #521
• Fix Iris parser httpx read timeout by @cdbartholomew in #524
• fix: migrate mental_models.embedding dimension alongside memory_units by @nicoloboschi in #526
• feat: filter operations by type + fix stale auto-refresh closure (#522) by @nicoloboschi in #527
• fix(consolidation): respect bank mission over ephemeral-state heuristic by @nicoloboschi in #525
• docs: add FAQ entry for conversation retain format by @nicoloboschi in #529
• fix: normalize named tool_choice to avoid HTTP 400 on LM Studio / Ollama by @nicoloboschi in #528
• doc: Your Pydantic AI Agent Forgets You After Every Run. Fix It in 5 Lines. by @benfrank241 in #531
• Fix run-db-migration for all-tenant upgrades by @cdbartholomew in #530
• doc: What's New in Hindsight — Document File Upload by @benfrank241 in #532
• doc: split blog index into Hindsight and Hindsight Cloud sections by @nicoloboschi in #534
• fix: strip null bytes from parsed file content before retain by @nicoloboschi in #535
• fix: resolve remaining webhook schema issues in multi-tenant retain by @Andreymi in #533
• feat: manual retry for failed async operations by @nicoloboschi in #537

Full Changelog: v0.4.16...v0.4.17

v0.4.16

What's Changed

• docs: 0.4.15 release blog post and changelog by @nicoloboschi in #477
• doc: update cookbook by @nicoloboschi in #479
• blog: add LiteLLM persistent memory post by @benfrank241 in #481
• doc: fix LiteLLM blog — OPINION → OBSERVATION, update title by @benfrank241 in #482
• cli: add linux-arm64 binary to release and CI by @nicoloboschi in #484
• fix: resolve all Dependabot security vulnerabilities by @nicoloboschi in #486
• doc: add ZeroEntropy reranker to models.md by @nicoloboschi in #489
• feat: webhook system with retain.completed event, UI, and docs by @nicoloboschi in #487
• fix: resolve TypeError when LLM returns invalid JSON across all retries by @nicoloboschi in #490
• Fix bank-level MCP tool filtering for FastMCP 3.x by @cdbartholomew in #491
• doc: add MCP blog post by @benfrank241 in #492
• perf: add GIN index on source_memory_ids for observation lookup by @nicoloboschi in #485
• fix: replace additive combined scoring with multiplicative CE boosts by @nicoloboschi in #494
• fix: cap uvicorn graceful shutdown at 5s and enable force-kill on double Ctrl+C by @nicoloboschi in #495
• fix: resolve chunks for observation results via source_memory_ids by @nicoloboschi in #496
• feat(openclaw): v2 recall/retention controls, scalability fixes, and Gemini safety settings by @mysteriousHerb in #480
• fix: use correct schema name in webhook outbox callback by @cdbartholomew in #499
• fix: run schema migrations in thread to prevent event loop deadlock by @nicoloboschi in #500
• doc: Give Your OpenAI App a Memory in 5 Minutes by @benfrank241 in #498
• fix: preserve None temporal fields for observations without source dates by @nicoloboschi in #501

New Contributors

• @mysteriousHerb made their first contribution in #480

Full Changelog: v0.4.15...v0.4.16

v0.4.15

What's Changed

• doc: changelog and blog post by @nicoloboschi in #445
• doc: improvements by @nicoloboschi in #448
• doc: 0.4.14 by @nicoloboschi in #449
• feat: observation_scopes field to drive observations granularity by @nicoloboschi in #447
• feat(openclaw): retain last n+2 turns every n turns (default n=10) by @fabioscarsi in #452
• Add bank-scoped validation to engine and HTTP handlers by @cdbartholomew in #454
• fix: JSON serialization and logging exception propagation in claude_code_llm by @nicoloboschi in #461
• fix: zeroentropy rerank URL missing /v1 prefix and MCP retain async_processing param by @nicoloboschi in #460
• fix(control-plane): observations count always showing 0 by @nicoloboschi in #464
• fix: resolve consolidation deadlock caused by zombie processing tasks on retry by @nicoloboschi in #463
• fix(reflect): prevent context_length_exceeded on large memory banks by @nicoloboschi in #462
• feat: support timestamp="unset" to retain content without a date by @nicoloboschi in #465
• feat: entity labels — optional, free_values, multi_value, UI polish by @nicoloboschi in #450
• docs: entities vs tags vs metadata by @nicoloboschi in #466
• feat: add Pydantic AI integration for persistent agent memory by @benfrank241 in #441
• feat: add Pydantic AI integration to CI, release pipeline, and docs by @nicoloboschi in #467
• feat: add tags filtering and q description fix for list documents API by @nicoloboschi in #468
• blog: add CrewAI persistent memory post by @benfrank241 in #471
• feat: add extension hooks for root routing and error headers by @DK09876 in #470
• refactor(openclaw): replace console.log with debug() helper by @slayoffer in #456
• fix(performance): improve recall and retain performance on large banks by @nicoloboschi in #469
• feat: configurable Gemini/Vertex AI safety settings by @nicoloboschi in #473
• perf(recall): improve single-query chunk fetch by @nicoloboschi in #475
• fix(ts-sdk): send null instead of undefined when includeEntities is false by @nicoloboschi in #476
• refactor: replace set_gemini_safety_settings() with LLMProvider.with_config() by @nicoloboschi in #474

New Contributors

• @fabioscarsi made their first contribution in #452

Full Changelog: v0.4.14...v0.4.15

v0.4.14

What's Changed

• doc: 0.4.13 changelog by @nicoloboschi in #411
• fix(chore): include hindsight-crewai in release workflow by @nicoloboschi in #412
• feat(openclaw): add autoRecall toggle and excludeProviders schema by @slayoffer in #413
• doc: improve api explanation by @nicoloboschi in #415
• misc: fix vertex/gemini errors and use it for ci tests by @nicoloboschi in #414
• Fix bank config API for multi-tenant schema isolation by @cdbartholomew in #417
• feat: add ZeroEntropy reranker provider support by @franchb in #420
• Fix reflect based_on population and enforce full hierarchical retrieval by @cdbartholomew in #421
• fix: improve memory footprint of recall by @nicoloboschi in #423
• feat: include doc metadata in fact extraction by @nicoloboschi in #424
• feat: increase customization for reflect, retain and consolidation by @nicoloboschi in #419
• feat: enable bank config API by default by @nicoloboschi in #426
• feat: add reflect mode to LoComo benchmark and improve reflect agent by @nicoloboschi in #428
• fix(openclaw): pass auth token to health check endpoint by @slayoffer in #427
• fix: handle observations regeneration when memories get deleted by @nicoloboschi in #429
• feat: batch observations consolidation by @nicoloboschi in #430
• feat: expand MCP tool surface area by @DK09876 in #435
• feat: filter graph memories with tags by @nicoloboschi in #431
• fix: pass encoding_format="float" in LiteLLM SDK embedding calls by @franchb in #434
• fix: catch ValueError instead of bare except in date parsing by @haosenwang1018 in #438
• feat: configure exposed mcp tools per bank by @nicoloboschi in #439
• fix(storage): use dynamic schema_getter in PostgreSQLFileStorage for multi-tenant by @Andreymi in #440
• fix: fail fast HNSW index when embedding dimension exceeds 2000 by @slayoffer in #361
• feat: add Chat SDK integration for persistent chat bot memory by @benfrank241 in #442
• chore: integrate chat with release by @nicoloboschi in #443
• fix: doc build and add chat doc by @nicoloboschi in #444

New Contributors

• @haosenwang1018 made their first contribution in #438
• @Andreymi made their first contribution in #440

Full Changelog: v0.4.13...v0.4.14

v0.4.13

What's Changed

• doc: changelog for 0.4.12 by @nicoloboschi in #397
• fix: document not tracked if has 0 extracted facts by @nicoloboschi in #399
• feat: add CrewAI integration for persistent crew memory by @benfrank241 in #319
• fix: clients don't respect timeout setting by @nicoloboschi in #400
• fix: reduce temporal ordering offset from 10s to 10ms by @dcbouius in #402
• fix: reranker crashes on provider error by @nicoloboschi in #403
• fix(mcp): stateless param not supported anymore by @nicoloboschi in #406
• feat: include source facts in observation recall by @nicoloboschi in #404
• fix: docker startup fails with named docker volumes by @nicoloboschi in #405
• fix(mcp): unify hindsight-mcp-local and server mcp by @nicoloboschi in #407
• fix: npx hindsight-control-plane fails by @nicoloboschi in #408
• feat: switch default model to gpt-4o-mini by @nicoloboschi in #410

New Contributors

• @benfrank241 made their first contribution in #319

Full Changelog: v0.4.12...v0.4.13

v0.4.12

What's Changed

• doc: changelog and blog post for 0.4.11 by @nicoloboschi in #363
• feat: allow chunks-only in recall (max_tokens=0) by @nicoloboschi in #364
• fix(openclaw): remove unused imports, retry health check, suppress unhandled rejection by @slayoffer in #373
• fix: propagate document_tags in async retain path by @abix5 in #374
• feat: add Go client SDK with ogen code generation by @franchb in #375
• fix: improve async batch retain with large payloads by @nicoloboschi in #366
• feat: support Batch API for retain (openai/groq) by @nicoloboschi in #365
• feat: use official go generator for Go client by @nicoloboschi in #377
• feat: support for pgvectorscale (DiskANN) by @nicoloboschi in #378
• doc: add go client examples by @nicoloboschi in #380
• ci: ensure docs get created with no extracted facts by @nicoloboschi in #379
• feat: support azure pg_diskann by @nicoloboschi in #381
• doc: add faq page by @nicoloboschi in #383
• fix(openclaw): error E2BIG on large content ingested by @nicoloboschi in #389
• fix(python-client): async method parity and server keepalive timeout by @cdbartholomew in #387
• feat: accept pdf, images and office files by @nicoloboschi in #390
• fix: restore entity retrieval in recall by @dcbouius in #391
• fix(go-client): use monorepo-compatible module path by @franchb in #392
• feat: improve ai sdk tools by @nicoloboschi in #394
• fix(go-client): add go build to CI by @nicoloboschi in #393
• fix(openclaw): shell safety, HTTP dual-mode, lazy reinit, per-user banks by @slayoffer in #388
• feat: add iris as file parser by @nicoloboschi in #395
• fix: improve openclaw test coverage by @nicoloboschi in #396

New Contributors

• @abix5 made their first contribution in #374
• @franchb made their first contribution in #375

Full Changelog: v0.4.11...v0.4.12

v0.4.11

What's Changed

• fix(helm): gke overriding HINDSIGHT_API_PORT by @nicoloboschi in #328
• feat(helm): add PDB and per-component affinity support by @nuclon in #327
• feat(openclaw): add excludeProviders config by @GodsBoy in #332
• Fix MCP usage metering: propagate tenant_id through context vars by @DK09876 in #334
• feat: add otel traceability by @nicoloboschi in #330
• feat: add docs skill by @nicoloboschi in #335
• fix: include tiktoken in slim image by @nicoloboschi in #336
• Fix async batch retain incorrectly marked as internal by @cdbartholomew in #338
• feat: add mental model CRUD tools to MCP server by @DK09876 in #337
• feat(helm): add TEI reranker sidecar support by @slayoffer in #333
• Harden MCP server: fix routing, validation, and usage metering by @DK09876 in #341
• Add actual LLM token usage fields to RetainResult by @cdbartholomew in #342
• fix: improve model configuration for litellm gateway by @nicoloboschi in #345
• fix: add trust_code env config by @nicoloboschi in #347
• Replace waitlist links with direct signup URL by @cdbartholomew in #349
• Fix MCP extra args rejection and bank ID resolution priority by @DK09876 in #351
• feat: add reverse proxy support by @nicoloboschi in #346
• Implement the vchord / pgvector support by @qdrddr in #350
• fix: resolve based_on schema/serialization issues in reflect API by @nicoloboschi in #348
• feat: implement hierarchical configuration (system, tenant, bank) by @nicoloboschi in #329
• feat: support for other text and vector search pg extensions by @nicoloboschi in #355
• chore: remove dead code by @nicoloboschi in #356
• Fix incorrect MCP tool parameters in docs by @DK09876 in #358
• feat: support timescale pg_textsearch as text search extension by @nicoloboschi in #359
• feat: support litellm-sdk as reranker and embeddings by @nicoloboschi in #357
• fix(openclaw): avoid memory retain recursion by @nicoloboschi in #362

New Contributors

• @qdrddr made their first contribution in #350

Full Changelog: v0.4.10...v0.4.11

v0.4.10

What's Changed

• fixed cast error by @haydenrear in #300
• fix: tagged directives should be applied to tagged mental models by @nicoloboschi in #303
• docs: add AI SDK integration documentation by @nicoloboschi in #304
• ci: ensure backwards/forward compatibility of the API by @nicoloboschi in #306
• fix(openclaw): remove format:uri to fix ajv warning by @GodsBoy in #309
• feat: support markdown in reflect and mental models by @nicoloboschi in #307
• ci: ensure python 3.14 compatibility by @nicoloboschi in #310
• fix(ci): resolve flaky test failures in api tests by @nicoloboschi in #311
• feat: slim docker distro by @nicoloboschi in #314
• doc: update claude-code usage terms by @nicoloboschi in #315
• fix: hindsight-embed profiles are not loaded correctly by @nicoloboschi in #316
• feat: add TenantExtension auth to MCP endpoint by @DK09876 in #286
• doc: improve Node.js client example by @vanvuongngo in #320
• feat: improve mcp tools based on endpoint by @nicoloboschi in #318
• fix(openclaw): prevent memory wipe on every session by @slayoffer in #323
• feat: add docker-compose example by @vanvuongngo in #313
• fix: do not log db user/password by @vanvuongngo in #312
• Add Supabase tenant extension by @jerryhenley in #267
• fix(helm): improve appVersion usage by @nicoloboschi in #326
• doc: prepare doc for 0.4.10 by @nicoloboschi in #325

New Contributors

• @haydenrear made their first contribution in #300
• @vanvuongngo made their first contribution in #320
• @jerryhenley made their first contribution in #267

Full Changelog: v0.4.9...v0.4.10

v0.4.9

What's Changed

• doc: changelog for 0.4.8 by @nicoloboschi in #283
• doc: update cookbook by @nicoloboschi in #284
• fix(openclaw): improve shell argument escaping by @slayoffer in #288
• feat(openclaw): add external Hindsight API support by @slayoffer in #289
• docs: expand external API configuration for OpenClaw by @slayoffer in #294
• feat(openclaw): add dynamic per-channel memory banks by @slayoffer in #290
• fix: hide hf logging by @nicoloboschi in #295
• fix: improve claude code and codex for /reflect by @nicoloboschi in #285
• feat(hindsight-litellm): support streaming on wrappers by @nicoloboschi in #296
• feat: HindsightEmbedded python SDK by @nicoloboschi in #293
• feat: improve mental models ux on control plane by @nicoloboschi in #297
• Fix recall endpoint timeout handling and add query length validation by @cdbartholomew in #298
• feat: ai sdk integration by @nicoloboschi in #299

Full Changelog: v0.4.8...v0.4.9