v0.7.2

What's Changed

• docs: add 0.7.1 changelog and release blog post by @nicoloboschi in #1818
• chore: regenerate docs-skill references after v0.7.1 by @Sanderhoff-alt in #1822
• docs(multilingual): add pg_search backend to BM25 selector and comparison table by @r266-tech in #1824
• docs(grok-build): add Grok Build integration page by @benfrank241 in #1793
• feat(gemini-spark): add Hindsight integration for Gemini Spark via MCP by @benfrank241 in #1779
• fix(api): isolate claude-code provider subprocess from user plugins by @nicoloboschi in #1825
• chore: regenerate docs skill changelog index after v0.7.1 by @benfrank241 in #1827
• fix(api): honor explicit daemon host and port by @Sanderhoff-alt in #1821
• fix(api): vchord ANN — use cosine opclass and dispatch tuning GUCs per backend by @isac322 in #1668
• feat(vapi): add Vapi voice AI webhook memory integration by @benfrank241 in #923
• feat(roo-code): add Roo Code integration with MCP + rules by @benfrank241 in #920
• fix(directives): honor tag_groups in list_directives and reflect by @nicoloboschi in #1831
• docs(models): list openrouter, google, and jina-mlx in Cross-Encoder Supported Providers table by @r266-tech in #1832
• docs(mcp): document sync_retain tool and correct tool counts (26/29 -> 27/30) by @r266-tech in #1834
• chore(deps): bump the uv group across 1 directory with 2 updates by @dependabot[bot] in #1836
• fix: preserve raw reranker scores for calibrated [0,1] providers by @nicoloboschi in #1846
• test(reflect): regression test that tag_groups reaches internal recall by @nicoloboschi in #1828
• Bound trigram entity resolution batch size during retain by @voarsh2 in #1841
• fix(control-plane): honor basePath for auth redirects by @xmh1011 in #1845
• fix(control-plane): validate login returnTo to prevent open redirect by @nicoloboschi in #1848
• chore(control-plane): move tests out of src/ into tests/ by @nicoloboschi in #1850
• fix(retain): never silently drop memory on a fact-extraction failure (#1833) by @nicoloboschi in #1852
• fix(retain): apply batching to Oracle entity resolution + guarantee pg_trgm RESET by @nicoloboschi in #1847
• fix(consolidation): propagate round-limit re-queue failure to worker retry by @nicoloboschi in #1857
• fix(retain): preserve full document body when splitter chunks oversized input by @nicoloboschi in #1855
• fix(consolidation): shorten retry backoff base from 60s to 5s by @nicoloboschi in #1854
• ci(test-api): shard pytest 3 ways + cache resolved .venv by @nicoloboschi in #1856
• fix: wait for daemon health before reclaiming occupied port by @s09x in #1858
• docs(faq): explain Hindsight's event-centric graph vs. traditional KGs by @cdbartholomew in #1837
• docs(blog): 15k stars milestone post by @benfrank241 in #1835
• fix(consolidation): scope-locked parallel dispatch (alternative to #1843) by @nicoloboschi in #1853
• docs(configuration): document HINDSIGHT_API_WORKER_TASK_RETRY_BACKOFF_SECONDS in Distributed Workers by @r266-tech in #1861
• docs(configuration): document link_expansion per-entity-limit and timeout knobs by @r266-tech in #1864
• chore(deps): bump uv from 0.11.6 to 0.11.15 in /hindsight-integrations/crewai in the uv group across 1 directory by @dependabot[bot] in #1865
• docs(api): fix broken Worker Configuration anchor in operations page by @r266-tech in #1869
• docs(retrieval): document calibrated [0,1] reranker score passthrough by @r266-tech in #1870
• docs(cli): correct bank set-disposition flags to --skepticism/--literalism/--empathy by @r266-tech in #1871
• docs(configuration): document HINDSIGHT_CP_DATAPLANE_API_KEY for Control Plane by @r266-tech in #1872
• fix(db_utils): make acquire_with_retry yield exactly once by @slayoffer in #1880
• fix(cli): hindsight memory retain --timestamp + correct fact-type values by @slayoffer in #1881
• fix(embedded-db): bump pg0-embedded to 0.14.2 for clean stop/restart by @nicoloboschi in #1892
• fix(embed): launch Windows daemon via pythonw to stop ConPTY terminal tab by @nicoloboschi in #1890
• fix(retain): close to_unit_id deferred-FK race on memory_links inserts (#1882) by @nicoloboschi in #1894
• ci: add daily Windows smoke test (API + Python client integration) by @nicoloboschi in #1895
• feat(api): add Fireworks AI batch inference provider by @Fyko in #1860
• fix(api): robust retain/recall on special-token literals and lone surrogates by @nicoloboschi in #1891
• ci(windows-smoke): pin extras on uv run so deps survive by @nicoloboschi in #1900
• chore(docs): regenerate hindsight-docs skill references by @nicoloboschi in #1899
• fix(retain): offset chunk_index across sub-batches of an oversized document (#1888) by @nicoloboschi in #1896
• docs: drop removed opinion/agent fact_type from MCP/SDK/integration references by @r266-tech in #1893
• fix(consolidation): make per-bank consolidation submit atomic + scope-aware (#1842) by @nicoloboschi in #1898
• fix(backup): include all 7 missing tables in backup/restore by @nicoloboschi in #1903
• fix(db): unblock PostgreSQL upgrade to v0.7.x (sqlalchemy<2.1 + autocommit_block migrations) by @nicoloboschi in #1904
• docs(integrations): drop removed opinion fact_type from recall_types by @r266-tech in #1905
• feat(flowise): add Flowise integration with Hindsight memory tools by @benfrank241 in #1436
• fix(api): bound native ML thread pools to available CPUs by @nicoloboschi in #1901
• chore(dev): one-shot dev setup script + fix control-plane production build by @nicoloboschi in #1910
• feat(google-adk): add Hindsight integration for Google ADK by @benfrank241 in #1862
• blog: Hermes Agent on Windows — Set Up Persistent Memory with Hindsight by @benfrank241 in #1913
• blog: Building a Hermes Coding Assistant on Windows That Remembers Your Codebase by @benfrank241 in #1912
• perf(graph-maintenance): cast ANN seed embeddings once + add perf suite (#1919) by @nicoloboschi in #1924
• refactor(api): move audit-logs endpoint queries into MemoryEngine by @nicoloboschi in #1925
• docs(docker): drop --rm, add --name + restart policy in run examples by @nicoloboschi in #1927
• fix(control-plane): pin next to 16.2.5 to fix standalone i18n redirect loop (#1926) by @nicoloboschi in #1928

New Contributors

• @Fyko made their first contribution in #1860

Full Changelog: v0.7.1...v0.7.2

v0.7.1

What's Changed

• docs(retrieval): add pgroonga to the BM25 backends table by @r266-tech in #1783
• docs: add 0.7.0 changelog and release blog post by @nicoloboschi in #1781
• fix(control-plane): add graph_maintenance to operations type filter by @nicoloboschi in #1785
• fix(control-plane): polish zh translation for naturalness by @XIYBHK in #1791
• docs(blog): OpenClaw agent that remembers your codebase by @benfrank241 in #1768
• fix(openclaw): stop silently skipping dispatch on synthetic-main + static-banking setups by @nicoloboschi in #1802
• fix markdown tables in mental models and cosmetic issues by @chrislatimer in #1800
• docs(retrieval): note pg_search configurable tokenizer in BM25 backends table by @r266-tech in #1790
• fix(openclaw): label "Current time" as UTC in injected memory context by @nicoloboschi in #1804
• fix(api): anchor recall recency to query timestamp by @Sanderhoff-alt in #1788
• fix(openclaw): flush un-retained turns on session_end by @nicoloboschi in #1806
• feat(embeddings): add token refresh to CodexOAuthEmbeddings; extract CodexAuthManager by @DK09876 in #1712
• feat(control-plane): add Chinese locale variants by @MapleEve in #1784
• docs(models): add claude-code Docker recipe with host Max Plan auth by @benfrank241 in #1526
• feat(integrations): default recallTypes to ['observation'] for openclaw + claude-code by @nicoloboschi in #1808
• feat(reranker): per-provider HTTP timeout env vars by @nicoloboschi in #1810
• fix(embeddings): clean up CodexOAuthEmbeddings token-refresh follow-up by @nicoloboschi in #1809
• docs(docker): add local llama.cpp sidecar compose example by @nicoloboschi in #1814
• feat(api): configure LLM reasoning effort via environment by @s9rkn in #1815
• docs(integrations): default recallTypes to ["observation"] for openclaw + claude-code (#1808) by @r266-tech in #1812
• fix(retain): keep oversized items in one async child to stop FK race (#1795) by @nicoloboschi in #1805
• feat(worker): add priority-based consolidation bank scheduling by @nicoloboschi in #1813
• fix(consolidation): indefinite retry with backoff + dedup-by-bank guard by @nicoloboschi in #1811
• docs(models): list openai-codex and openrouter in embeddings Supported Providers table by @r266-tech in #1792

New Contributors

• @XIYBHK made their first contribution in #1791
• @MapleEve made their first contribution in #1784
• @s9rkn made their first contribution in #1815

Full Changelog: v0.7.0...v0.7.1

v0.7.0

What's Changed

• docs: add 0.6.2 changelog and release blog post by @benfrank241 in #1633
• docs: attribute 0.6.2 release post to benfrank241 by @benfrank241 in #1635
• blog: What's New in Hindsight Cloud by @benfrank241 in #1636
• feat(typescript-client): replace Promise with concrete generated types by @solidprinciples in #1640
• docs(changelog): correct openai-agents v0.1.1 entry by @benfrank241 in #1639
• docs: add HINDSIGHT_API_WORKER_ID tip to API quickstart by @benfrank241 in #1617
• fix(api): strip Gemma thought tags by @xmh1011 in #1653
• feat: add opencode-go LLM provider by @tevenfeng in #1652
• fix(api): lazy load reflect tiktoken encoding by @xmh1011 in #1654
• fix(api): preserve tag group or triggers by @xmh1011 in #1655
• fix(api): auto-refresh openai-codex OAuth access_token by @benfrank241 in #1661
• fix(typescript-client): update repository URL to correct repo by @solidprinciples in #1657
• Blog: Agent Memory Consolidation framework by @benfrank241 in #1672
• style: apply ruff format to openai_compatible_llm.py by @benfrank241 in #1703
• docs(crewai): prioritize Hindsight Cloud in quickstart by @benfrank241 in #1689
• docs(strands): prioritize Hindsight Cloud in quickstart by @benfrank241 in #1690
• docs(pipecat): prioritize Hindsight Cloud in quickstart by @benfrank241 in #1695
• Regenerate hindsight-docs skill references by @benfrank241 in #1686
• docs(pydantic-ai): prioritize Hindsight Cloud in quickstart by @benfrank241 in #1691
• docs(smolagents): prioritize Hindsight Cloud in quickstart by @benfrank241 in #1692
• docs(codex): add Cloud API URL to quickstart and config default by @benfrank241 in #1693
• docs(agentcore): add Cloud API URL to quickstart and config default by @benfrank241 in #1694
• docs(agno): add Cloud API URL to quickstart and config default by @benfrank241 in #1696
• docs(n8n): add Cloud API URL to quickstart and config default by @benfrank241 in #1697
• docs(dify): add Cloud API URL to quickstart and config default by @benfrank241 in #1698
• docs(paperclip): add Cloud API URL to quickstart and config default by @benfrank241 in #1699
• docs(nemoclaw): add Cloud API URL to quickstart and config default by @benfrank241 in #1700
• docs(chat): add Cloud API URL to quickstart by @benfrank241 in #1701
• fix(api): ignore null bank config overrides by @xmh1011 in #1664
• fix(api): reject blank retain content by @xmh1011 in #1685
• fix: derive HINDSIGHT_API_HEALTH_URL default from HINDSIGHT_API_PORT by @sg-shag in #1709
• fix(mcp): omit reflect provenance by default by @xmh1011 in #1665
• fix: break mutual recursion in batch API fallback for non-batch providers by @mssteuer in #1463
• feat(reranker): support alibaba qwen3-rerank by @quicklyfast in #1501
• feat(extensions): add OperationValidator.precheck pre-body-parse hook by @cdbartholomew in #1548
• fix(claude-code-mcp): make run_mcp.sh bootstrap idempotent on Windows by @ottopichlhoefer in #1565
• fix(integrations-claude-code): label 'Current time' as UTC in recall context by @valda in #1568
• fix(reranker): detect pre-normalized scores and use rank-based normalization by @xuli500177 in #1512
• feat(right-agent): add Right Agent integration by @onsails in #1599
• fix(docker): detect nested pg0 data directories by @xmh1011 in #1650
• chore(deps): bump openssl from 0.10.79 to 0.10.80 in /hindsight-clients/rust in the cargo group across 1 directory by @dependabot[bot] in #1663
• fix(helm): remove stale Chart.lock that pulls in conflicting Bitnami postgresql sub-chart by @jay-mtl in #1632
• fix(mental-models): cap history array length to prevent jsonb overflow by @cdbartholomew in #1593
• fix(control-plane): surface upstream errors via respondWithSdk helper by @offendingcommit in #1678
• fix(embed): resolve npx absolute path on Windows before spawning UI by @tuancookiez-hub in #1682
• fix: escape literal braces in all user-supplied prompt fields by @nicoloboschi in #1728
• fix(mental-models): full refresh pending delta baselines by @xmh1011 in #1684
• fix: avoid retrying invalid embedding dimensions by @ai-ag2026 in #1687
• chore(deps): bump the uv group across 2 directories with 2 updates by @dependabot[bot] in #1705
• fix(openclaw): update Hindsight dependency ranges by @de1tydev in #1716
• fix(worker): handle stale pending schema routines by @xmh1011 in #1666
• fix(api): enable gzip middleware to keep graph payload parseable by @nicoloboschi in #1731
• fix(docs): correct DATABASE_URL env var name and fix invisible code in tip titles by @nicoloboschi in #1733
• fix(control-plane): verify signed session cookie instead of presence by @nicoloboschi in #1739
• fix: skip fuzzy entity resolution for user-defined label entities (#1558) by @nicoloboschi in #1737
• fix(api): wire up per-operation LLM concurrency caps by @nicoloboschi in #1738
• docs(blog): Hermes coding assistant codebase memory by @benfrank241 in #1710
• fix(retain): split oversized single items in batch retain (#1571) by @nicoloboschi in #1736
• feat(api): add clear endpoint for mental model content by @nicoloboschi in #1743
• test: stabilize pre-existing CI flakes by @nicoloboschi in #1742
• fix(api): make litellm-sdk embeddings api_key optional for Bedrock IAM auth by @jakub-qg in #1744
• fix(reflect): hide disabled tools from the agent's system prompt by @nicoloboschi in #1740
• fix(ollama): add ollama-cloud provider and fix native API auth by @nicoloboschi in #1734
• feat(api): add targeted consolidation by observation scopes by @nicoloboschi in #1746
• fix(api): release glibc heap pages after local reranker batches by @nicoloboschi in #1745
• docs(mcp): document clear_mental_model tool by @r266-tech in #1750
• fix(api): ignore inherited v1 base URL for Codex by @de1tydev in #1718
• fix(webhooks): avoid duplicate retain batch deliveries by @xmh1011 in #1683
• fix(api): stop sending temperature param to Anthropic API by @nicoloboschi in #1753
• feat(bm25): configurable native language + opt-in pgroonga backend by @nicoloboschi in #1538
• feat: support Codex OAuth embeddings by @haha0815 in #1704
• fix(ci): align config field count and CLI consolidation call with #1746 by @nicoloboschi in #1757
• chore(db): drop 9 unused/redundant indexes by @nicoloboschi in #1762
• feat(control-plane): surface clear_mental_model in UI by @nicoloboschi in #1764
• feat(paperclip): add per-user memory isolation via bankGranularity by @benfrank241 in #1761
• feat(api): add ParadeDB pg_search as Citus-compatible BM25 backend by @nicoloboschi in https://github.com/vectorize-io/hindsight/pull...

v0.6.2

What's Changed

• blog: How Hindsight Scales by @nicoloboschi in #1539
• fix(claude-code): get_page detail=content + handle tool-result spillover by @cdbartholomew in #1543
• docs: add 0.6.1 changelog and release blog post by @nicoloboschi in #1542
• blog: add cover image for How Hindsight Scales by @benfrank241 in #1545
• fix(claude-code-mcp): rename recall max_results→max_tokens by @offendingcommit in #1544
• Fix strands client lifecycle leak for internally-owned clients by @benfrank241 in #1547
• blog: add category filter to blog landing page by @benfrank241 in #1580
• fix(paperclip): align with Paperclip's actual event payloads by @amirhmoradi in #1560
• blog: the case against external vector DBs for agent memory by @benfrank241 in #1594
• Docs: add Windows + China deployment guidance for embeddings setup by @benfrank241 in #1549
• docs(strands): document FastAPI lifecycle pattern from #1547 by @r266-tech in #1581
• docs(paperclip): align integration guide + README with #1560 lifecycle by @r266-tech in #1596
• fix(agent-sdk): agent_knowledge_recall — rename max_results to max_tokens, default 1024 by @r266-tech in #1552
• fix(ci): paperclip lint formatting + openclaw hook test expectations by @dcbouius in #1601
• security: bump vulnerable dependencies across npm and pip by @dcbouius in #1600
• security: bump urllib3 to 2.7.0 in integration lockfiles by @dcbouius in #1603
• security: bump remaining high/critical deps across all lockfiles by @dcbouius in #1609
• security: bump litellm to >=1.83.14 by @dcbouius in #1610
• feat(claude-code): expose configurable MCP request timeout (#1575) by @rsaulo in #1591
• fix(agent-sdk): agent_knowledge_get_page request detail=content (sister of #1543) by @r266-tech in #1557
• fix: handle transient OID errors in embedding dimension migration by @dcbouius in #1612
• fix(ci): use frozen lockfile in lint.sh during CI by @dcbouius in #1618
• fix(docs): use real GitHub handle for ContextForge integration author by @benfrank241 in #1621
• fix(cli, control-plane): make retain Event Date / timestamp actually reach the API by @benfrank241 in #1622
• docs(claude-code): document requestTimeoutSeconds option from #1591 by @r266-tech in #1626
• docs(cli): document --timestamp flag on memory retain (#1622) by @r266-tech in #1623
• fix(migrations): repair mental_models.subtype at current head (#1553) by @benfrank241 in #1627
• fix(ci): set UV_FROZEN=1 on verify-generated-files job by @dcbouius in #1629
• chore(deps): bump the uv group across 3 directories with 8 updates by @dependabot[bot] in #1630
• blog: onboarding a new engineer onto five months of OpenCode memory by @benfrank241 in #1628

New Contributors

• @offendingcommit made their first contribution in #1544
• @amirhmoradi made their first contribution in #1560
• @rsaulo made their first contribution in #1591

Full Changelog: v0.6.1...v0.6.2

v0.6.1

What's Changed

• docs: add 0.6.0 changelog and release blog post by @nicoloboschi in #1458
• chore: remove self-driving-agents CLI — moved to own repo by @nicoloboschi in #1461
• fix(retain): reduce memory pressure by clearing content references after use by @nicoloboschi in #1455
• fix(openclaw): mission semantics + retainQueue config whitelist by @nicoloboschi in #1473
• feat(claude-code): create-agent skill understands SDA layout by @nicoloboschi in #1475
• chore(deps): bump pg0-embedded to >=0.14.0 by @nicoloboschi in #1476
• feat(openclaw): drop redundant before_agent_start + add debugPerfTiming by @nicoloboschi in #1477
• feat(control-plane): enrich bank dropdown with memory stats by @nicoloboschi in #1479
• feat(engine): optional read-only backend for recall queries by @cdbartholomew in #1460
• fix(claude-code): bootstrap 'mcp' Python package on first run by @nicoloboschi in #1485
• feat(perf): publish perf-test results to external dashboard by @nicoloboschi in #1474
• fix(packaging): hard-pin meta packages to matching hindsight-api-slim by @nicoloboschi in #1486
• docs(claude-code): document knowledge tools and subagent skill by @nicoloboschi in #1487
• ci(release): add --provenance to npm publish (required for n8n Verified) by @benfrank241 in #1491
• chore: stabilize CI — docs-skill pre-commit hook + retain dict-mutation fix by @nicoloboschi in #1490
• fix(docker): chmod 755 /home/hindsight for --user UID:GID support by @nicoloboschi in #1493
• feat(cli): add --strategy flag to memory retain-files by @nicoloboschi in #1494
• blog: Your Claude Code Subagents Don't Share What They Learn by @benfrank241 in #1456
• docs(cli): document --strategy flag for memory retain-files by @r266-tech in #1499
• docs(env): document HINDSIGHT_API_READ_DATABASE_URL in .env.example by @r266-tech in #1496
• ci: cosign-sign published GHCR images (manual workflow) by @nicoloboschi in #1495
• ci: cosign-sign release images + document verification by @nicoloboschi in #1502
• Add AlloyDB ScaNN vector index support by @can1357 in #1459
• fix(worker): probe pg_proc before calling optional schemas_with_pending_work() by @nicoloboschi in #1503
• docs(installation): bake custom models into image instead of PVC by @nicoloboschi in #1504
• follow-up to #1459: AlloyDB ScaNN docs + review-nit cleanups by @nicoloboschi in #1506
• feat: add map-type entity labels for structured entity extraction by @nicoloboschi in #1505
• fix(openclaw): pass enableKnowledgeTools through getPluginConfig by @nicoloboschi in #1507
• docs(entity-labels): document type="map" structured entity groups by @r266-tech in #1508
• perf(worker): scope progress-stats fanout via existing routine by @nicoloboschi in #1509
• fix(n8n): drop hindsight-client runtime dep, inline HTTP calls by @benfrank241 in #1513
• fix(openclaw): write hooks.allowConversationAccess in setup wizard by @nicoloboschi in #1514
• fix(openclaw): reuse existing token + URL when re-running setup wizard by @nicoloboschi in #1518
• feat(claude-code): resolve git worktrees + explicit directory-bank mapping by @nicoloboschi in #1520
• fix(openclaw): backfill plugins.allow with hindsight-openclaw in setup wizard by @nicoloboschi in #1521
• fix(engine): remove multiplicative retry layers in fact extraction by @nicoloboschi in #1516
• docs: surface stable worker_id guidance and zombie-operation recovery by @nicoloboschi in #1522
• fix(daemon): replace os.fork() with subprocess.Popen to fix MPS on macOS by @nicoloboschi in #1519
• blog: n8n Workflows Are Stateless. Hindsight Makes Them Compound. by @benfrank241 in #1511
• fix(reflect): read document metadata from retain params by @nepenth in #1523
• fix(worker): propagate child error_message to failed batch_retain parent by @cdbartholomew in #1527
• fix(claude-code): use detail=metadata for agent_knowledge_list_pages by @cdbartholomew in #1528
• feat(cp): add optional access-key login for Control Plane by @ariel-ai-bot in #1530
• feat: add z.ai as LLM provider by @Burgunthy in #1529
• feat(openclaw): prepend session-context block to retained transcripts by @kryptt in #1439
• fix(cp): improve access-key auth UX and harden middleware by @nicoloboschi in #1533
• docs(zai): document z.ai provider and add default model by @nicoloboschi in #1532
• chore: fix formatting in llm_wrapper.py to pass verify-generated-files by @nicoloboschi in #1534
• perf(locomo): scope CI to a 3-conversation curated subset by @nicoloboschi in #1536
• fix(typescript-client): derive CLIENT_VERSION via tsup define by @nicoloboschi in #1540
• feat(api): add litellmrouter provider for LLM fallback chains by @nicoloboschi in #1537

New Contributors

• @can1357 made their first contribution in #1459
• @nepenth made their first contribution in #1523
• @ariel-ai-bot made their first contribution in #1530
• @Burgunthy made their first contribution in #1529
• @kryptt made their first contribution in #1439

Full Changelog: v0.6.0...v0.6.1

v0.6.0

What's Changed

• Add Pipecat voice AI persistent memory blog post by @benfrank241 in #1300
• docs(guides): add Hindsight update guides by @benfrank241 in #1301
• Fix broken link in CLI ARM64 guide by @benfrank241 in #1305
• Oracle 23ai database backend by @DK09876 in #1307
• feat(agentcore): add hindsight-agentcore integration for Bedrock AgentCore Runtime by @benfrank241 in #822
• fix(agentcore): async-native client, task tracking, drop per-package CHANGELOGs by @nicoloboschi in #1313
• fix(oracle): restore PG query semantics and clean up migration chain by @nicoloboschi in #1312
• refactor(release): single integration table in generate_changelog by @nicoloboschi in #1314
• docs: add scoring & ranking deep dive to recall docs by @nicoloboschi in #1317
• feat: add SmolAgents integration with Hindsight memory tools by @benfrank241 in #658
• fix(oracle): restore exact v0.5.6 PG query shapes by @nicoloboschi in #1321
• fix(test): remove stale profile auto-create assertion from bank stats test by @nicoloboschi in #1323
• fix(embeddings): add allowed_openai_params for OpenAI-compatible embedding dimensions by @zwcf5200 in #1320
• fix: redact database URLs in config logs by @xmh1011 in #1316
• docs(mental-models): document tags?source=mental_models per #1296 by @r266-tech in #1311
• feat(python-client): expose retain_async in retain() and aretain() by @harryplusplus in #1306
• fix(codex): fix encoding with PowerShell by @jerviscui in #1185
• chore: remove stray files accidentally landed on main by @benfrank241 in #1326
• fix(deps): cap cryptography <47 to fix SIGILL on ARM64 Apple Silicon by @nicoloboschi in #1324
• feat: self-driving agents (part1) by @nicoloboschi in #1302
• chore(docs): sync version-0.5 docs from next by @nicoloboschi in #1328
• perf(db): eliminate ResultRow wrapping overhead for PostgreSQL by @nicoloboschi in #1325
• perf: add recall-with-observations & consolidation suites, split CI steps, fix locomo by @nicoloboschi in #1333
• docs(models): document litellm-sdk embeddings provider by @r266-tech in #1336
• feat(perf): add HTTP mode to recall benchmark by @nicoloboschi in #1315
• fix(consolidator): dedupe + ON CONFLICT for observation_sources INSERT by @youchi1 in #1340
• fix(embed): include 'all' extras when spawning hindsight-api from sibling source by @youchi1 in #1341
• fix(async-ops): atomically commit batch_retain parent and child rows by @cdbartholomew in #1343
• docs(python-sdk): document retain_async kwarg per #1306 by @r266-tech in #1347
• docs: add Oracle Database as supported enterprise storage by @nicoloboschi in #1356
• fix(llm): guard against null content from OpenAI-compatible providers by @nicoloboschi in #1355
• feat(oracle): unify migrations under Alembic with dialect dispatcher by @nicoloboschi in #1330
• fix(config): default openai-codex model to gpt-5.4 by @nicoloboschi in #1357
• fix(codex): filter synthetic AGENTS startup messages by @voarsh2 in #1346
• feat(self-driving-agents): add nemoclaw harness support by @nicoloboschi in #1335
• docs: add Pydantic Logfire guide for Hindsight tracing by @benfrank241 in #1339
• Blog: Agent Memory in SmolAgents with Hindsight Tools by @benfrank241 in #1329
• Add AWS AgentCore integration blog post by @benfrank241 in #1377
• fix(embed): drop gpt-4o-mini fallback when hindsight-api import fails by @nicoloboschi in #1363
• feat(typescript-client): add AbortSignal support to all HindsightClient methods by @harryplusplus in #1198
• feat(stats): add time_field toggle to memories-timeseries chart by @aliu-ronin in #1246
• feat(opencode): share memory bank across git worktrees of the same repo by @isac322 in #1352
• docs(self-driving-agents): document nemoclaw harness + --sandbox flag by @r266-tech in #1367
• Harden OpenAI-compatible JSON responses and ensure json_object prompt hint by @voarsh2 in #1368
• fix(hindsight-api-slim): bump opentelemetry-{api,sdk,instrumentation,exporter} floors so PrometheusMetricReader 0.62b1 doesn't crash startup by @nikolay-bratanov in #1373
• fix(oracle): update CHECK constraints in baseline to match PG schema by @DK09876 in #1379
• fix(api): include observations in per-document graph and counts by @youchi1 in #1374
• fix(webhooks): route webhook endpoints through tenant-aware engine methods by @cdbartholomew in #1388
• fix: redirect daemon subprocess stdout/stderr on POSIX to prevent TUI corruption by @laoli-no1 in #1380
• fix(retain): retain all documents in async multi-document batches by @voarsh2 in #1382
• chore(embed): tidy detach-popen helper and close log fds in parent by @nicoloboschi in #1418
• codex: add configurable recall timeout for UserPromptSubmit hooks by @voarsh2 in #1399
• fix(openclaw): add WeakSet registration guard keyed by API instance by @vernmic in #1409
• chore(clients): regenerate clients for time_field timeseries param by @nicoloboschi in #1420
• fix(typescript-client): expose missing recall/reflect params by @nicoloboschi in #1362
• fix: clean up async batch retain test fixtures and comments by @nicoloboschi in #1419
• fix(api): recognize Pydantic aliases in unknown param middleware by @Tosko4 in #1417
• fix(retain): defer memory_links → memory_units FKs to break cascade deadlock by @cdbartholomew in #1398
• fix(daemon): honor --host and HINDSIGHT_API_HOST in daemon mode by @nicoloboschi in #1422
• fix(recall): preserve original exception in recall_async error path by @nicoloboschi in #1421
• fix(worker): handle NotImplementedError from add_signal_handler on Windows by @nicoloboschi in #1423
• fix(self-driving-agents): fail fast when nemoclaw sandbox is missing by @nicoloboschi in #1365
• fix(mcp): expose tag_groups parameter on recall tool by @nicoloboschi in #1424
• chore(generated): regenerate openapi spec + clients post #1246 by @aliu-ronin in #1426
• fix(hindsight-embed): use sysconfig to find scripts dir in daemon start by @nicoloboschi in #1425
• feat(anthropic): env-driven max_retries + default_headers knobs by @TuftyBruno in #1389
• feat(self-driving-agents): add Claude Chat/Cowork harness by @nicoloboschi in #1427
• feat(self-driving-agents): add Hermes Agent harness support by @nicoloboschi in #1431
• fix(entity-resolver): stamp cooccurrences with event_date, not now() by @aliu-ronin in #1247
• blog: Your Agent Harness Has Tools. It Still Needs Memory. by @benfrank241 in #1429
• docs(guides): add framework memory integration guides by @benfrank241 in #1432
• feat(n8n): add n8n community-node package for Hindsight memory by @benfrank241 in #1364
• fix(n8n): add missing entry point, fix auth header, add execution tests by @nicoloboschi in #1444
• ci: stop re-running test wo...

v0.5.6

What's Changed

• docs(release): changelog + blog for v0.5.5 by @nicoloboschi in #1297
• ci: include linux arm64 CLI in release assets by @xmh1011 in #1298

New Contributors

• @xmh1011 made their first contribution in #1298

Full Changelog: v0.5.5...v0.5.6

v0.5.5

What's Changed

• release: 0.5.4 changelog by @nicoloboschi in #1195
• fix: auto-confirm control plane install on first UI launch by @bwjoke in #1197
• feat(worker): per-operation slot reservations by @nicoloboschi in #1199
• feat(perf): add system performance test runner by @nicoloboschi in #1201
• fix(perf): fix perf-test CI install and WorkerPoller kwarg by @nicoloboschi in #1206
• feat(db): configurable Postgres statement_timeout on pool connections by @nicoloboschi in #1200
• fix(worker): prevent child tasks from blocking parent execution by @nicoloboschi in #1210
• chore(perf): default CI perf-test scale to large by @nicoloboschi in #1211
• Blog: Hindsight 10,000 Stars Celebration by @benfrank241 in #1208
• Fix: Update 10k stars blog post date to April 22 by @benfrank241 in #1213
• feat(api): expose retry_count and next_retry_at on operation responses by @cdbartholomew in #1188
• feat(retain): expose processed_content_tokens on RetainResult by @cdbartholomew in #1217
• fix(claude-code): retain on SessionEnd even when retainEveryNTurns > 1 by @starbit-biostar in #1152
• feat(claude-code): recall from additional banks alongside primary by @starbit-biostar in #1153
• feat(perf): add LoComo benchmark as parallel CI job by @nicoloboschi in #1223
• fix(claude-code): prevent compaction from overwriting retained memories by @nicoloboschi in #1222
• fix(integrations): preserve raw UTF-8 in dynamically-derived bank_id by @Desko77 in #1141
• docs(workers): document per-operation slot reservations (#1199) by @r266-tech in #1207
• fix(perf): locomo defaults — all conversations, wait-consolidation, pro model by @nicoloboschi in #1224
• fix(mcp): remove delete_memory tool to close authorization-bypass gap by @nicoloboschi in #1228
• fix(ops): expose processing/cancelled statuses through API and UI by @nicoloboschi in #1231
• feat(api): add exclude_parents filter to list operations by @nicoloboschi in #1230
• feat(api,ui): document chunks API, reprocess, and enhanced document detail by @nicoloboschi in #1236
• fix: correct docker command in 10k stars blog quickstart by @benfrank241 in #1237
• blog: Your Agent Is Not Forgetful. It Was Never Given a Memory. by @benfrank241 in #1235
• guides: add 20 general agent memory guides by @benfrank241 in #1233
• fix(litellm): handle streaming responses in _store_conversation by @DK09876 in #1239
• fix(claude-code): handle list-content tool_results in transcript parsing by @nicoloboschi in #1226
• fix(stats): timeseries buckets should return tz-aware ISO by @aliu-ronin in #1245
• fix(db): accept asyncpg-style URLs for external PostgreSQL by @nicoloboschi in #1225
• feat(embeddings): allow Cohere output dimensions via env var by @nicoloboschi in #1249
• fix(docs): escape curly braces in generated changelog entries by @nicoloboschi in #1248
• fix(tests): repair 9 regressions surfaced on main by @nicoloboschi in #1251
• Update integration author for ContextForge by @ooa-andera in #1254
• feat(embed): full Windows support + prefer sibling hindsight-api over uvx by @nicoloboschi in #1250
• fix(llm): handle DeepSeek tool-call compatibility by @M1p0 in #1253
• docs: add nodes_by_fact_type field to Document Response Format example by @r266-tech in #1243
• feat(embeddings): add HINDSIGHT_API_EMBEDDINGS_GEMINI_FORCE_IPV4 opt-in by @connorblack in #1241
• docs(ops): document processing + cancelled statuses from #1231 by @r266-tech in #1238
• docs(blog): align retainEveryNTurns default with #1186 (10 → 3) by @r266-tech in #1268
• docs(sidecar): mirror HINDSIGHT_API_EMBEDDINGS_GEMINI_FORCE_IPV4 from #1241 by @r266-tech in #1265
• fix(embed): satisfy ty on Windows-only subprocess attrs by @grimmjoww in #1263
• docs(configuration): mirror COHERE_OUTPUT_DIMENSIONS env var into sidecar reference by @r266-tech in #1252
• chore(deps): bump actions/github-script from 8 to 9 by @dependabot[bot] in #1025
• fix(openclaw): stop silently skipping retention on default agent:main:main sessions by @nicoloboschi in #1276
• fix(consolidation): reduce memory fan-out during consolidation recall by @nicoloboschi in #1277
• Add blog post: Connect ChatGPT and Perplexity to Hindsight for Long-Term Memory by @benfrank241 in #1255
• chore(lint): cover all npm packages with prettier in lint.sh by @nicoloboschi in #1279
• fix(mcp): report Hindsight's version in serverInfo, not FastMCP's by @cdbartholomew in #1281
• docs(integrations): add ChatGPT and Perplexity integration guides by @benfrank241 in #1280
• docs(installation): document memory footprint and hardware requirements by @nicoloboschi in #1282
• feat(pipecat): add Pipecat voice AI pipeline memory integration by @benfrank241 in #921
• fix(llm): simplify JSON schemas for better Ollama and LLM compliance by @nicoloboschi in #1292
• fix(parsers): LlamaParse follow-up — reuse client, fix error mapping, add tests by @nicoloboschi in #1293
• fix(api): GET /banks/{bank_id}/profile no longer auto-creates the bank by @cdbartholomew in #1287
• fix(retain): drop strength from causal relations (fixes #1289) by @nicoloboschi in #1295
• fix(llm): omit tool_choice="auto" and add deepseek as first-class provider by @nicoloboschi in #1294
• chore(deps): bump actions/upload-artifact from 4 to 7 by @dependabot[bot] in #1278
• test(integration): add Hermes Agent embedded-mode smoke test by @nicoloboschi in #1283
• feat: mental-models List view + /tags?source=mental_models by @nicoloboschi in #1296

New Contributors

• @bwjoke made their first contribution in #1197
• @starbit-biostar made their first contribution in #1152
• @Desko77 made their first contribution in #1141
• @aliu-ronin made their first contribution in #1245
• @M1p0 made their first contribution in #1253
• @connorblack made their first contribution in #1241

Full Changelog: v0.5.4...v0.5.5

v0.5.4

What's Changed

• release: 0.5.3 notes and blog post by @nicoloboschi in #1126
• blog: OpenAI Agents persistent memory with Hindsight by @benfrank241 in #1129
• fix(openai-agents): review followup — docs, tests, polish by @DK09876 in #1134
• docs: fix HINDSIGHT_API_LLM_MAX_RETRIES default (10 → 3) by @r266-tech in #1137
• docs(sdk): add document CRUD methods to TypeScript client reference by @r266-tech in #1132
• fix(ai-sdk): align ReflectBasedOn types with OpenAPI spec by @nicoloboschi in #1163
• fix(api): populate items_count in list_operations by @nicoloboschi in #1164
• feat(admin): add decommission-workers and worker-status CLI commands by @nicoloboschi in #1165
• feat(claude-code): add {user_id} retainTags template variable by @soichisumi in #1161
• fix(startup): downgrade LLM verify_connection failure to warning by @nicoloboschi in #1166
• feat: disable daemon idle timeout by default by @nicoloboschi in #1162
• blog: OpenCode persistent memory with Hindsight by @benfrank241 in #1172
• guide batch, OpenClaw and Hermes memory by @benfrank241 in #1177
• fix(worker): pass DeferOperation through MemoryEngine.execute_task by @cdbartholomew in #1135
• fix(engine): use ensure_ascii=False in json.dumps for LLM prompts by @harryplusplus in #1169
• docs: add cover images for Apr 20 OpenClaw and Hermes guide batch by @benfrank241 in #1181
• fix(worker): scan for active schemas before claiming by @cdbartholomew in #1109
• fix(mcp): route update_bank through config resolver, expose all fields by @nicoloboschi in #1168
• docs(admin-cli): document decommission-workers and worker-status by @r266-tech in #1180
• fix(reflect): honor reflect_mission identity framing in prompt builder by @nicoloboschi in #1167
• fix(retain): prevent duplicate memory units from delta upsert chunk issues by @nicoloboschi in #1178
• fix(alembic): merge divergent heads for v0.5.3 by @grimmjoww in #1149
• docs(mcp): document update_bank config_updates and configurable fields by @r266-tech in #1183
• blog: add five agent memory articles by @benfrank241 in #1184
• fix(opencode): lower retainEveryNTurns default from 10 to 3 by @DK09876 in #1186
• fix(reflect): prevent directive leakage on empty banks by @nicoloboschi in #1190
• fix: allow reflect-specific LLM config when default is disabled by @zwcf5200 in #1189
• chore(deps): bump actions/upload-pages-artifact from 4 to 5 by @dependabot[bot] in #1170
• feat(embeddings): make OpenAI-compatible batch size configurable (#1142) by @r266-tech in #1143
• fix(embedded): recover from daemon crashes transparently by @nicoloboschi in #1193
• fix(retain): preserve document created_at across upsert + UI edit flow by @nicoloboschi in #1194
• fix(reflect): scope delta mental model recall to new memories only by @nicoloboschi in #1192

New Contributors

• @soichisumi made their first contribution in #1161
• @harryplusplus made their first contribution in #1169
• @zwcf5200 made their first contribution in #1189

Full Changelog: v0.5.3...v0.5.4

v0.5.3

What's Changed

• feat(openclaw): session-scoped document_id and structured per-message timestamp by @nicoloboschi in #1075
• release: 0.5.2 notes and blog post by @nicoloboschi in #1074
• docs: add screenshots to 0.5.2 release post by @nicoloboschi in #1078
• fix(control-plane): encode bank ids in URLs end-to-end by @nicoloboschi in #1079
• fix(reflect): forward mental model max_tokens to refresh by @nicoloboschi in #1076
• feat(cli): add named connection profiles (-p/--profile) by @nicoloboschi in #1080
• feat(api): add tenant field and configurable allowlist to JSON logs by @nicoloboschi in #1085
• feat(paperclip): replace library with Paperclip plugin (v0.2.0) by @benfrank241 in #934
• docs: add /guides section with Hermes how-to guides by @benfrank241 in #1062
• fix(worker): per-tenant fair rotation in claim_batch by @cdbartholomew in #1088
• feat(mental-models): staleness signal + history reflect snapshot + UI revamp by @nicoloboschi in #1089
• fix: handle 'timestamp' field for file retain API by @christerence in #1092
• fix: files/retain upload problems and orphaned retains by @christerence in #1091
• fix(worker): make submit_task idempotent when payload already set by @nicoloboschi in #1097
• test(file-retain): regression for timestamp -> event_date mapping by @nicoloboschi in #1096
• fix(openclaw): resolve full symlink chain in isDirectExecution() by @D2758695161 in #1093
• feat(control-plane): surface failed-consolidation count and drilldown by @nicoloboschi in #1100
• fix(openclaw): per-session retain stops overwriting prior turns by @nicoloboschi in #1102
• fix(ollama): add think=false to _call_ollama_native payload by @karl-8888 in #1099
• fix(consolidation): prevent orphan observations when source deleted mid-consolidation by @nicoloboschi in #1090
• feat(recall): make budget mapping configurable per bank by @nicoloboschi in #1106
• feat(worker): DeferOperation exception for extension-driven requeue by @nicoloboschi in #1105
• feat(mental-models): structured-ops delta refresh + observation cleanup on upsert by @nicoloboschi in #1101
• fix(control-plane): clearer constellation recency legend & node tooltip by @nicoloboschi in #1108
• blog: Constellation View and Entity Co-occurrence Graph by @benfrank241 in #1103
• docs: add /guides section with how-to guides and comparisons by @benfrank241 in #1110
• feat: add OpenAI Agents SDK integration by @DK09876 in #842
• fix(integrations): add encoding="utf-8" to transcript file reads by @tordf in #1119
• fix(ai-sdk): align HindsightClient interface with hindsight-client@0.5.2 by @nicoloboschi in #1118
• fix(migrations): restore broken chain for v0.4.22 to v0.5.x upgrades by @nicoloboschi in #1117
• fix(jina-mlx): serialize Metal GPU ops to prevent SIGSEGV on concurrent reranker calls by @lkttle in #1113
• fix: improve reranker error messages and add configurable TEI timeout by @octo-patch in #1115
• fix(config): reduce default LLM max retries from 10 to 3 by @nicoloboschi in #1121
• feat: add consolidation_max_memories_per_round config by @nicoloboschi in #1123

New Contributors

• @christerence made their first contribution in #1092
• @D2758695161 made their first contribution in #1093
• @karl-8888 made their first contribution in #1099
• @tordf made their first contribution in #1119
• @lkttle made their first contribution in #1113

Full Changelog: v0.5.2...v0.5.3