vectorize-io/hindsight v1.1.0 on GitHub

What's Changed

docs: v1.0.0 changelog, Memlake launch post, and version-1.0 docs snapshot by @nicoloboschi in #5550
fix(client): stop the export poll loop from overrunning its timeout on Windows by @nicoloboschi in #5568
feat/webhook event delivery ids by @nicoloboschi in #5576
feat(control-plane): identify dataplane requests with its own User-Agent by @nicoloboschi in #5536
feat(extensions): LLM request extension — choose the LLM and credentials of each call server-side by @nicoloboschi in #5578
fix(retain): only rebuild observations when a tag change moves their scopes by @nicoloboschi in #5596
ci(triage-labels): stop labelling server issues as coding-agents by @nicoloboschi in #5605
fix(llm): put the mental-model refresh's reflect calls under its own concurrency cap by @nicoloboschi in #5597
fix(litellm): retry on the HTTP status LiteLLM maps, not message keywords by @nicoloboschi in #5604
feat(config): one decision-model setup with per-operation overrides by @nicoloboschi in #5607
fix(worker): drop the overstated claimable count from [PENDING_BREAKDOWN] by @nicoloboschi in #5608
fix(api): stop an abandoned reflect mid-call when its client disconnects by @nicoloboschi in #5610
feat(config): Cloudflare Clef and OpenAI Decisions API as decision-model providers by @nicoloboschi in #5613
fix(retain): fail a chunk whose reply is not JSON, even 'no facts' prose (#5565) by @nicoloboschi in #5615
fix(anthropic): fail fast on refusal stop_reason instead of retrying as invalid JSON by @nicoloboschi in #5616
fix(api): re-check mental-model edits under the row lock so the vector matches the stored text by @nicoloboschi in #5619
fix(retain): write a held batch when the producer is blocked on the memory budget by @nicoloboschi in #5620
feat/recall prefer observations default by @nicoloboschi in #5617
fix(llm): skip json_object response_format for z.ai by @nicoloboschi in #5618
feat(evals): consolidation eval suite by @nicoloboschi in #5626
feat(recall): rank with the decision model when one is set up by @nicoloboschi in #5284
fix(consolidation): keep an UPDATE when the same response deletes that observation (#2550) by @nicoloboschi in #5577
blog: 50,000 GitHub Stars by @benfrank241 in #5625
fix(hermes): default observation_scopes to shared so observations merge across sessions by @nicoloboschi in #5629
fix(local-models): build local torch models one at a time, download in parallel (#4420) by @nicoloboschi in #5609
fix(retain): don't mark facts edited when an unchanged document is re-sent by @nicoloboschi in #5630
fix(reflect): re-ask a done document of the wrong shape instead of failing the refresh (#5202) by @nicoloboschi in #5631
refactor(recall): rank with one decision-model pass, tell it the question date by @nicoloboschi in #5632
fix(consolidation): prompt rules and safety nets found by the consolidation evals by @nicoloboschi in #5627
perf(graph): build the stale-cooccurrence live set from candidate pairs only by @nicoloboschi in #5640
fix(embed): strip dotenv quotes when reading profile .env files by @nicoloboschi in #5642
fix(reflect): a "nothing found" answer needs recall first (#4567) by @nicoloboschi in #5641
fix(control-plane): keep uploads visible until the retain they queue finishes (#5310) by @nicoloboschi in #5643
feat(extensions): let metering extensions see and gate transfer-archive imports by @cdbartholomew in #5579
feat(documents): search document text and show which fields matched by @nicoloboschi in #5633
fix(retain): reject async append without document_id at submit time by @nicoloboschi in #5651
fix/openai service tier chat completions by @nicoloboschi in #5652
fix(retain): reclaim ingress attachments when the write-scope check refuses a retain by @nicoloboschi in #5655
ci(release): stop trying to publish the hermes plugin to PyPI by @nicoloboschi in #5654
fix(transfer): remap mental-model evidence on documents import by @nicoloboschi in #5657
fix(tags): reject an unknown tags_match instead of widening it to "any" (#5371) by @nicoloboschi in #5653
fix(tags): one tag-filter rule across knowledge base, entities and control plane (#5417) by @nicoloboschi in #5656
fix(embeddings): avoid the sentence-transformers 5.4+ dimension FutureWarning by @nicoloboschi in #5660
fix(api): port the LlamaParse connector to the v2 parse API by @nicoloboschi in #5659
fix/import bank bounded batches by @nicoloboschi in #5661
fix(codex): default to a model ChatGPT accounts can use, and stop wasting quota per call by @nicoloboschi in #5658
fix(api): measure durations with a monotonic clock, ban time.time() via ruff by @nicoloboschi in #5662
fix(api): close bank-validator gaps from #5448 by @nicoloboschi in #5663
fix(control-plane): a dataplane 401 no longer sends a valid session to login (#5452) by @nicoloboschi in #5667
fix(mental-models): accept a delta array the provider already parsed by @nicoloboschi in #5669
fix(memories): tell a store which reranker recall uses; export a store-owned fact's id by @nicoloboschi in #5671
feat(documents): change observation_scopes on existing documents with PATCH (#5309) by @nicoloboschi in #5668
fix(retain): keep an unset event_date across reprocess and edit by @nicoloboschi in #5670
feat(control-plane): download a knowledge page or mental model as PDF or Markdown by @nicoloboschi in #5679
fix(reflect): keep facts read through expand in citations and based_on by @nicoloboschi in #5677
fix(api): probe banks for consolidation work without seq-scanning memory_units by @nicoloboschi in #5672
feat(control-plane): newest/oldest toggle on the timeline view by @nicoloboschi in #5681
fix(mcp): let the recall tool return source chunks, entities and source facts by @nicoloboschi in #5678
fix(retain): resume a partially written oversized document slice by slice by @nicoloboschi in #5676
fix(mental-models): delta refresh fixes a changed value everywhere on the page (#5228) by @nicoloboschi in #5680
fix(usage): keep cached and thinking tokens in reflect and retain totals by @nicoloboschi in #5675
feat(reflect): decision-orchestrated reflect when a decision model is set up by @nicoloboschi in #5382
feat(mental-models): separate temperature for refresh delta/retraction calls (#5500) by @nicoloboschi in #5683
feat(helm): add extraObjects for deploying arbitrary extra manifests by @nicoloboschi in #5687
fix(api): validate bank ids on import and reserve the _ prefix by @nicoloboschi in #5673
feat(control-plane): retry failed uploads and retry all failed operations by @nicoloboschi in #5689
perf(api): stop loading unused heavy modules at startup (#5479) by @nicoloboschi in #5686
fix(control-plane): make the Memories tabs switch instantly on big banks by @nicoloboschi in #5692
feat(documents): cap the documents search and say when the total is a lower bound by @nicoloboschi in #5682
fix(claude-code): keep host hooks out of LLM subprocesses when the daemon runs from $HOME by @nicoloboschi in #5691
feat(anthropic): allow Workload Identity Federation instead of a static API key by @nicoloboschi in #5690
feat(mental-models): scheduled full refresh for delta models by @nicoloboschi in #5674
feat(helm): configurable revisionHistoryLimit and Gateway API HTTPRoute by @nicoloboschi in #5695
fix(embed): apply --profile when the CLI runs via python -m by @nicoloboschi in #5694
fix(llm): report real token usage for Codex, Claude Code and Anthropic by @nicoloboschi in #5693
feat(consolidation): decision-model triage settles easy facts before the LLM call by @nicoloboschi in #5664
feat(api): NVIDIA NIM reranker and embeddings providers by @nicoloboschi in #5697
fix(transfer): import an archive without holding it in memory (#5524) by @nicoloboschi in #5696
feat(api): chatgpt LLM provider via Sign in with ChatGPT by @nicoloboschi in #5688
blog: four things to build with Vercel Eve and Hindsight by @benfrank241 in #5699
fix(retain): bound multi-document fan-out on the SQL path by @nicoloboschi in #5740
fix(control-plane): sidebar and banks list reachable by keyboard and screen readers by @nicoloboschi in #5741
chore(skills): ship-it cleanup + issue filing, new /triage skill by @nicoloboschi in #5739
docs(config): OpenRouter + models that reject temperature by @nicoloboschi in #5743
fix(api): a refresh cut off during a retry backoff no longer pauses automatic refresh (#5561) by @nicoloboschi in #5745
perf(consolidation): embed a response's actions in one embedder call by @nicoloboschi in #5744
fix(worker): one SIGTERM runs the graceful shutdown instead of force-exiting by @nicoloboschi in #5747
fix(llm): cool down a failed chain member instead of retrying it on every request by @nicoloboschi in #5742
feat(reranker): onnx provider for any cross-encoder exported to ONNX by @nicoloboschi in #5748
fix(llm): retry a 200 that ends with finish_reason=error instead of reading it as no tool call (#5612) by @nicoloboschi in #5751
fix(retain): a failed chunk no longer replaces the document (#5589) by @nicoloboschi in #5753
fix(retain): fail once on a rejected LLM key instead of replaying it (#5644) by @nicoloboschi in #5749
fix(worker): wake the idle poll loop when a task finishes by @nicoloboschi in #5755
fix(consolidation): ease DB pool pressure and honour auto-consolidation off on re-queue by @nicoloboschi in #5756
feat(curation): let a memory edit change its mentioned_at by @nicoloboschi in #5750
fix(retain): a reply without a top-level facts list is an error, not 0 facts by @nicoloboschi in #5754
fix(codex): replay tool-call history with call ids instead of flattening it to text by @nicoloboschi in #5761
feat(mental-models): decision-model gate in front of a delta page refresh by @nicoloboschi in #5698
fix(mental-models): tell knowledge pages apart from standalone mental models (#5665) by @nicoloboschi in #5759
fix(metrics): set the request schema when the backlog gauges ask the memories store by @nicoloboschi in #5757
docs(api): say the memory list/graph q filter is an unindexed scan, for browsing only by @nicoloboschi in #5758
fix(migrations): no global memory_units vector index on per-bank backends after an embedding resize (#5707) by @nicoloboschi in #5760
test: mark pgvector-index tests store-incompatible; isolate extension alembic tests by @nicoloboschi in #5762
perf(consolidation): embed a batch's recall queries in one embedder call by @nicoloboschi in #5769
feat(operations): retry runs a deferred operation now by @nicoloboschi in #5772
fix(minimax): send thinking=adaptive to MiniMax-M3.1, which rejects disabled by @nicoloboschi in #5774
fix(api): return consolidation state on GET /memories/{id} by @nicoloboschi in #5767
fix(reflect): accept done after a search that found nothing (#5702) by @nicoloboschi in #5777
fix(api): honor a systemd CPUQuota and bound ONNX Runtime threads (#5725) by @nicoloboschi in #5779
fix(banks): create a bank's storage before its row, not after by @nicoloboschi in #5763
fix(api): stop a big graph read from freezing the event loop (#5703) by @nicoloboschi in #5782
fix(llm): mental-model refresh inherits the global failover chain by @nicoloboschi in #5780
fix(consolidation): lock in the writers' order so it cannot deadlock against retain or delete (#5433) by @nicoloboschi in #5768
fix(control-plane): make the control plane usable on phones by @nicoloboschi in #5776
feat(documents): per-document change history rebuilt from operations (#5638) by @nicoloboschi in #5783
docs(eve): document the eve add memory/hindsight install by @benfrank241 in #5789
fix(coding-agents): append only new commits to the git-log document by @cdbartholomew in #5711
fix(transfer): export and import a bank's growing tables a page at a time (#5737) by @nicoloboschi in #5792
fix(typesafe): shrink and retry a request Jev finds over its window, stop dropping decision answers by @nicoloboschi in #5794
fix(observations): explicit clear resets the memory even with no observations to delete by @nicoloboschi in #5775
fix(deps): slim server installs from uv.lock on Python 3.14 (#5709) by @nicoloboschi in #5770
fix(reranker): stop a cancelled in-process rerank from starting more batches by @nicoloboschi in #5795
test(consolidation): give the deadlock tests' fake engine an embedder (#5797) by @nicoloboschi in #5798
fix(migrations): build a new schema's indexes plainly so provisioning never waits on other transactions by @cdbartholomew in #5700
fix(embeddings): defer operations when OpenAI embeddings run out of credits by @nicoloboschi in #5746
fix(mental-models): sibling pages no longer rewrite a document whose memory scope is empty (#5728) by @nicoloboschi in #5802
feat(api): tag globs in observation_scopes + per-bank default_observation_scopes (#5666) by @nicoloboschi in #5800
docs(api): runbook for splitting a memory bank in place by @benfrank241 in #5803
fix(mental-models): widen the columns that reference mental_models.id (#5788) by @nicoloboschi in #5793
fix: flaky tests #5752, #5764 (deadlock retry on re-retain, delta prompt) by @nicoloboschi in #5784
test: seed store-neutral fixtures through the bank's memories store by @nicoloboschi in #5799
fix(worker): a terminal write to a dropped schema drops the task instead of the worker by @cdbartholomew in #5813
fix: store-owned banks match Postgres on re-retain labels, failed re-retain, temporal recall and mental model clear by @nicoloboschi in #5796
fix(gemini): fail fast on content blocks instead of retrying them by @nicoloboschi in #5824
feat(worker): expose the operation id to validators via RequestContext by @cdbartholomew in #5807
fix(control-plane): neutral description for Default observation scopes by @nicoloboschi in #5827
chore(deps): bump claude-agent-sdk to >=0.2.121 (CVE-2026-96620) by @cdbartholomew in #5810
fix/audit log list preview by @nicoloboschi in #5826
fix(recall): a year-1/9999 date no longer breaks temporal recall and consolidation (#5820) by @nicoloboschi in #5828
feat(reranker): cap in-flight requests for the cohere reranker by @nicoloboschi in #5831
blog: Thirty Agents, One Memory by @benfrank241 in #5808
fix/audit action filter all by @nicoloboschi in #5834
feat(llm-trace): record embedding, rerank and decision-model calls in the request history by @nicoloboschi in #5835
fix(reflect): keep memory ids out of document-mode text (#5833) by @nicoloboschi in #5837
fix: flaky tests #5804, #5806, #5830; name the cause of a failed bank delete (#5805) by @nicoloboschi in #5838
chore(deps): bump hono in coding-agents lockfile by @cdbartholomew in #5844
chore(deps): bump aiohttp in agent-framework lockfile by @cdbartholomew in #5847
chore(deps): bump aiohttp in agentcore lockfile by @cdbartholomew in #5848
chore(deps): bump aiohttp in aider lockfile by @cdbartholomew in #5850
chore(deps): bump aiohttp in autogen lockfile by @cdbartholomew in #5851
chore(deps): bump aiohttp, mcp in claude-agent-sdk lockfile by @cdbartholomew in #5852
chore(deps): bump aiohttp in composio lockfile by @cdbartholomew in #5853
chore(deps): bump aiohttp in continue lockfile by @cdbartholomew in #5854
chore(deps): bump aiohttp, mcp in crewai lockfile by @cdbartholomew in #5855
chore(deps): bump aiohttp in google-adk lockfile by @cdbartholomew in #5858
chore(deps): bump aiohttp in haystack lockfile by @cdbartholomew in #5859
chore(deps): bump aiohttp in langgraph lockfile by @cdbartholomew in #5860
chore(deps): bump aiohttp in litellm lockfile by @cdbartholomew in #5861
chore(deps): bump aiohttp, tinytag in llamaindex lockfile by @cdbartholomew in #5862
chore(deps): bump aiohttp in meta-muse lockfile by @cdbartholomew in #5863
chore(deps): bump aiohttp, mcp in openai-agents lockfile by @cdbartholomew in #5864
chore(deps): bump aiohttp in pipecat lockfile by @cdbartholomew in #5865
chore(deps): bump aiohttp in pydantic-ai lockfile by @cdbartholomew in #5866
chore(deps): bump aiohttp in smolagents lockfile by @cdbartholomew in #5867
chore(deps): bump aiohttp in superagent lockfile by @cdbartholomew in #5869
chore(deps): bump aiohttp in vapi lockfile by @cdbartholomew in #5870
chore(deps): bump aiohttp in agno lockfile by @cdbartholomew in #5849
chore(deps): bump aiohttp, mcp in strands lockfile by @cdbartholomew in #5868
chore(deps): bump aiohttp in system-evals lockfile by @cdbartholomew in #5871
chore(deps): bump aiohttp in cursor lockfile by @cdbartholomew in #5856
chore(deps): bump aiohttp in dify lockfile by @cdbartholomew in #5857
chore(deps): bump aiohttp, mcp in root lockfile by @cdbartholomew in #5872
chore(deps): bump aiohttp in system-tests lockfile by @cdbartholomew in #5873
chore(deps): bump mcp in integration-tests lockfile by @cdbartholomew in #5874
chore(deps): bump aiohttp in hermes lockfile by @cdbartholomew in #5875
chore(deps): bump vite in paperclip lockfile by @cdbartholomew in #5876
chore(deps): bump vite in openclaw lockfile by @cdbartholomew in #5877
chore(deps): bump aiohttp in ag2 lockfile by @cdbartholomew in #5846
chore(deps): bump certifi in ag2 lockfile by @cdbartholomew in #5879
chore(deps): bump certifi in claude-agent-sdk lockfile by @cdbartholomew in #5881
chore(deps): bump certifi in composio lockfile by @cdbartholomew in #5882
chore(deps): bump certifi, uv in crewai lockfile by @cdbartholomew in #5883
chore(deps): bump certifi in google-adk lockfile by @cdbartholomew in #5885
chore(deps): bump certifi in haystack lockfile by @cdbartholomew in #5886
chore(deps): bump certifi in langgraph lockfile by @cdbartholomew in #5887
chore(deps): bump certifi in litellm lockfile by @cdbartholomew in #5888
chore(deps): bump certifi in llamaindex lockfile by @cdbartholomew in #5889
chore(deps): bump certifi in openai-agents lockfile by @cdbartholomew in #5890
chore(deps): bump certifi in pipecat lockfile by @cdbartholomew in #5891
chore(deps): bump certifi in pydantic-ai lockfile by @cdbartholomew in #5892
chore(deps): bump certifi in smolagents lockfile by @cdbartholomew in #5893
chore(deps): bump certifi in superagent lockfile by @cdbartholomew in #5895
chore(deps): bump certifi in system-tests lockfile by @cdbartholomew in #5897
chore(deps): bump certifi in agno lockfile by @cdbartholomew in #5880
chore(deps): bump certifi in strands lockfile by @cdbartholomew in #5894
chore(deps): bump certifi in system-evals lockfile by @cdbartholomew in #5896
chore(deps): bump certifi in dify lockfile by @cdbartholomew in #5884
chore(deps): bump certifi in integration-tests lockfile by @cdbartholomew in #5878
chore(deps): bump vite in ai-sdk lockfile by @cdbartholomew in #5899
chore(deps): bump vite in chat lockfile by @cdbartholomew in #5900
chore(deps): bump vite in cloudflare-oauth-proxy lockfile by @cdbartholomew in #5901
chore(deps): bump vite in coding-agents lockfile by @cdbartholomew in #5902
chore(deps): bump vite in eliza lockfile by @cdbartholomew in #5903
chore(deps): bump vite in eve lockfile by @cdbartholomew in #5904
chore(deps): bump vite in flowise lockfile by @cdbartholomew in #5905
chore(deps): bump vite in n8n lockfile by @cdbartholomew in #5906
chore(deps): bump vite in obsidian lockfile by @cdbartholomew in #5908
chore(deps): bump vite in opencode lockfile by @cdbartholomew in #5909
chore(deps): bump vite in embed-ui lockfile by @cdbartholomew in #5910
chore(deps): bump vite, launch-editor, follow-redirects in root lockfile by @cdbartholomew in #5911
chore(deps): bump vite in nemoclaw lockfile by @cdbartholomew in #5907
fix(reflect): work on models that reject a forced tool_choice; no 500 on an empty observation by @nicoloboschi in #5917
fix(reflect): answer once — free turns signal answer_ready instead of writing a discarded answer (#5832) by @nicoloboschi in #5916
chore(deps): bump certifi, pillow in root lockfile by @cdbartholomew in #5898
chore(deps): bump toktok-rs to 0.3.0 by @nicoloboschi in #5921
fix/store owned delta chunk ids by @nicoloboschi in #5920
fix(memories): page whole-set reads to the end; subset export reads only its documents by @nicoloboschi in #5918
blog: We Got Early Access to NVIDIA Nemotron 3.5 Embed by @Neeeser in #5841
fix(migrations): clear placeholder dates without scanning memory_units; consolidation reads them as unknown by @nicoloboschi in #5922
chore(deps): bump pillow to 13.0.0 in crewai integration by @cdbartholomew in #5930
chore(deps): bump pillow to 13.0.0 and nltk to 3.10.4 in llamaindex integration by @cdbartholomew in #5931
chore(deps): bump pillow to 13.0.0 in smolagents integration by @cdbartholomew in #5932
chore(deps): bump pillow to 13.0.0 in autogen integration by @cdbartholomew in #5929
chore(deps): bump pillow to 13.0.0 in pipecat integration by @cdbartholomew in #5933
feat(llm): let live requests jump ahead of background jobs for LLM slots by @nicoloboschi in #5928
fix(memories): make the memory list's default order total so offset pages don't overlap by @cdbartholomew in #5934

New Contributors

@Neeeser made their first contribution in #5841

Full Changelog: v1.0.0...v1.1.0