Issues

• hermes: cloud mode accepts a URL without a key and reports available, and names no credential when it is unavailable

  integration:hermes
  vectorize-io/hindsight#6438 · Filter by author woshicby opened on Nov 7, 2026

• 1.2.0: recall queries bypass EMBEDDINGS_MAX_INPUT_TOKENS, an oversized query 400s the whole recall (HTTP 500)

  core
  vectorize-io/hindsight#6437 · Filter by author e-wag opened on Nov 7, 2026

• test_78_split_embeddings is flaky: unstubbed consolidate call when the worker runs before the bank is deleted

  core
  vectorize-io/hindsight#6436 · Filter by author nicoloboschi opened on Nov 7, 2026

• Audit PATCH endpoints that replace a whole list instead of updating part of it

  core
  vectorize-io/hindsight#6435 · Filter by author nicoloboschi opened on Nov 7, 2026

• flaky: test-python-client-oracle hits ORA-00060 deadlock in TestDocuments::test_get_document

  core
  vectorize-io/hindsight#6434 · Filter by author nicoloboschi opened on Nov 7, 2026

• test(system): test_24_observation_scopes intermittently finds no observations after settled()

  core
  vectorize-io/hindsight#6432 · Filter by author nicoloboschi opened on Nov 7, 2026

• coding-agents: reflect still hard-capped at 300s under Bun (opencode) — Bun fetch idle timeout bypasses reflectToolTimeoutMs

  integration:coding-agents
  vectorize-io/hindsight#6431 · Filter by author baktiaditya opened on Nov 7, 2026
  2

• stats: total_links under-counts entity links in 1.x (sum is missing the per-unit factor n)

  core
  vectorize-io/hindsight#6430 · Filter by author sprezz1 opened on Nov 7, 2026

• Operations stuck in 'processing' are never reclaimed once the claiming worker is gone (still present in v1.2.0)

  core
  vectorize-io/hindsight#6429 · Filter by author hermes-3640 opened on Nov 7, 2026

• feat(retain): add retain_prompt_metadata_keys to filter which document metadata keys are injected into extraction prompts

  core
  vectorize-io/hindsight#6428 · Filter by author tt2-woo opened on Nov 7, 2026

• anthropic: serialize non-strict prompt-injected schema as compact JSON instead of indent=2

  core
  vectorize-io/hindsight#6427 · Filter by author tt2-woo opened on Nov 7, 2026

• retain: stop replaying exhausted provider request-timeout retries through the worker retry schedule

  core
  vectorize-io/hindsight#6426 · Filter by author tt2-woo opened on Nov 7, 2026

• Recall has no degraded mode: an embeddings outage fails every recall, though the BM25 arm needs no query vector

  core
  vectorize-io/hindsight#6424 · Filter by author Egoushka opened on Nov 6, 2026
  2

• [Bug] openai-compatible: per-turn soft-path schema re-bills reflect's whole conversation, and a mid-prompt timestamp truncates the cache before it

  bugcore
  vectorize-io/hindsight#6422 · Filter by author JasonSwindle opened on Nov 6, 2026
  2

• coding-agents: Claude AskUserQuestion answers are dropped from retained transcripts

  integration:coding-agents
  vectorize-io/hindsight#6421 · Filter by author azohra opened on Nov 6, 2026

• 1.2.0 ollama-cloud regression: schema tool turns and misleading done error on GPT-OSS/DeepSeek

  core
  vectorize-io/hindsight#6420 · Filter by author Korayem opened on Nov 6, 2026

• [hermes integration] opt-in retain_kanban so kanban worker runs can auto-retain

  integration:hermes
  vectorize-io/hindsight#6419 · Filter by author biondogs opened on Nov 6, 2026
  1

• Memory curation (PATCH /memories/{memory_id}) runs synchronously on the API pod with no async or bulk mode; a bulk edit job turns into a fleet-wide incident

  core
  vectorize-io/hindsight#6418 · Filter by author doconnell2020 opened on Nov 6, 2026

• enqueue_graph_maintenance row locks are held for the whole retain/edit transaction; concurrent same-bank writers convoy and exhaust the pool

  core
  vectorize-io/hindsight#6417 · Filter by author doconnell2020 opened on Nov 6, 2026

• add profile mentions in #6406 and #6410

  core
  vectorize-io/hindsight#6416 · Filter by author harshitgupta31415 opened on Nov 6, 2026

• count_unconsolidated() seq-scans memory_units on every consolidation job (inner LIMIT has no ORDER BY)

  core
  vectorize-io/hindsight#6415 · Filter by author Komzpa opened on Nov 6, 2026

• Flaky: test_admin_reembed fails with 'could not open relation with OID' under xdist

  core
  vectorize-io/hindsight#6412 · Filter by author nicoloboschi opened on Nov 6, 2026

• Flaky: upgrade test v0.3.0-HEAD fails when the old version's LLM extraction drops a fact

  core
  vectorize-io/hindsight#6397 · Filter by author nicoloboschi opened on Nov 6, 2026

• Flaky on Windows: test_windows_probes_get_a_deadline_longer_than_powershell_start

  core
  vectorize-io/hindsight#6396 · Filter by author nicoloboschi opened on Nov 6, 2026

• Flaky: test_concurrent_submits_queue_one_refresh queues 2 refreshes

  core
  vectorize-io/hindsight#6395 · Filter by author nicoloboschi opened on Nov 6, 2026