Search results

Hindsight Anthropic LLM provider sends deprecated temperature param, silently breaks reflection on Opus 4.x

bugSomething isn't workingSomething isn't working

Status: Open.

#1749 In vectorize-io/hindsight;

· chrislatimer opened on May 25, 2026

HINDSIGHT_API_EMBEDDINGS_LITELLM_SDK_API_KEY required even for IAM creds use

bugSomething isn't workingSomething isn't working

Status: Open.

#1741 In vectorize-io/hindsight;

· jakub-qg opened on May 25, 2026

Single retain() call produces multiple near-duplicate experience entries with identical timestamps

Status: Open.

#1730 In vectorize-io/hindsight;

· sirotaku opened on May 25, 2026

fix(openclaw): SessionEnd does not flush un-retained turns — short conversations lose all memory

Status: Open.

#1726 In vectorize-io/hindsight;

· eightHundreds opened on May 25, 2026

docs(hermes): missing concurrency guidance for local LLM backends causes silent slot starvation

bugSomething isn't workingSomething isn't working

Status: Open.

#1720 In vectorize-io/hindsight;

· GregChinn opened on May 24, 2026

FlashRank ONNX reranker: glibc allocator retains ~50MB/recall, RSS grows unbounded

Status: Open.

#1717 In vectorize-io/hindsight;

· aliatx2017 opened on May 24, 2026

Add per-bank consolidation slot allocation (HINDSIGHT_API_WORKER_CONSOLIDATION_SLOTS_PER_BANK)

Status: Open.

#1715 In vectorize-io/hindsight;

· hunterleesoik opened on May 23, 2026

VectorChord recall may over-expand semantic/BM25 candidates and degrade answer relevance before reranking

bugSomething isn't workingSomething isn't working

Status: Open.

#1707 In vectorize-io/hindsight;

· Sanderhoff-alt opened on May 22, 2026

Discussion: Narrator line ties first-person attribution to banks.name (often bank_id), causing entity pollution and transcript misclassification

Status: Open.

#1680 In vectorize-io/hindsight;

· eightHundreds opened on May 21, 2026

Async operations need poison quarantine / degraded queue health

Status: Open.

#1671 In vectorize-io/hindsight;

· ai-ag2026 opened on May 20, 2026

batch_retain can pass zero-length embeddings to pgvector (vector(384) vs vector(0))

Status: Open.

#1669 In vectorize-io/hindsight;

· ai-ag2026 opened on May 20, 2026

vchord backend: ANN queries cannot use vchordrq indexes due to opclass/operator mismatch (vector_l2_ops vs <=>)

Status: Open.

#1667 In vectorize-io/hindsight;

· isac322 opened on May 20, 2026