fleet-memory/hindsight-api-slim/hindsight_api/engine
Kagura 8e2e2d5bf2
fix(reflect): disable source facts in search_observations to prevent context overflow (#669)
search_observations in the reflect agent hardcoded include_source_facts=True
with max_source_facts_tokens=-1 (unlimited). For banks with many observations
backed by thousands of facts, a single tool call could produce 300K+ tokens,
exceeding the default 100K context budget and causing forced synthesis with
an empty 'Retrieved Data' section.

The reflect agent synthesizes from observations, not raw backing facts.
Disable source facts to keep payloads proportional to observation count
(~6K vs ~310K in the reporter's case).

The consolidation path already has configurable source fact limits (PR #509,
v0.4.17). The reflect path was not updated.

Fixes #668

Co-authored-by: Kagura Chen <daniyuu19@sjtu.edu.cn>
2026-03-24 09:12:54 +01:00
..
consolidation fix: prevent silent memory loss on consolidation LLM failure (#601) 2026-03-17 20:15:33 +01:00
directives feat: introduce hindsight-api-slim and hindsight-all-slim packages (#560) 2026-03-13 13:50:03 +01:00
parsers feat: introduce hindsight-api-slim and hindsight-all-slim packages (#560) 2026-03-13 13:50:03 +01:00
providers fix: strip markdown code fences from all LLM providers, not just local (#646) 2026-03-22 21:29:16 +01:00
reflect fix(reflect): disable source facts in search_observations to prevent context overflow (#669) 2026-03-24 09:12:54 +01:00
retain fix(entity_resolver): prevent _pending_stats/_pending_cooccurrences memory leak (#662) 2026-03-23 16:06:04 +01:00
search fix(recall): reject empty queries with 400 and fix SQL parameter gap (#632) 2026-03-21 20:24:36 +01:00
storage feat: introduce hindsight-api-slim and hindsight-all-slim packages (#560) 2026-03-13 13:50:03 +01:00
__init__.py feat: introduce hindsight-api-slim and hindsight-all-slim packages (#560) 2026-03-13 13:50:03 +01:00
cross_encoder.py docs: add config vars for local reranker FP16 and bucket batching (#589) 2026-03-16 17:35:09 +01:00
db_budget.py feat: introduce hindsight-api-slim and hindsight-all-slim packages (#560) 2026-03-13 13:50:03 +01:00
db_utils.py feat: introduce hindsight-api-slim and hindsight-all-slim packages (#560) 2026-03-13 13:50:03 +01:00
embeddings.py feat: introduce hindsight-api-slim and hindsight-all-slim packages (#560) 2026-03-13 13:50:03 +01:00
entity_resolver.py fix(entity_resolver): prevent _pending_stats/_pending_cooccurrences memory leak (#662) 2026-03-23 16:06:04 +01:00
interface.py feat: introduce hindsight-api-slim and hindsight-all-slim packages (#560) 2026-03-13 13:50:03 +01:00
jina_mlx_reranker.py feat: introduce hindsight-api-slim and hindsight-all-slim packages (#560) 2026-03-13 13:50:03 +01:00
llm_interface.py feat: introduce hindsight-api-slim and hindsight-all-slim packages (#560) 2026-03-13 13:50:03 +01:00
llm_wrapper.py fix: prevent silent memory loss on consolidation LLM failure (#601) 2026-03-17 20:15:33 +01:00
memory_engine.py test: add unit tests for pg_trgm auto-detection and ValidationResult.accept_with() enrichment (#650) 2026-03-23 10:33:09 +01:00
operation_metadata.py feat: introduce hindsight-api-slim and hindsight-all-slim packages (#560) 2026-03-13 13:50:03 +01:00
query_analyzer.py feat: introduce hindsight-api-slim and hindsight-all-slim packages (#560) 2026-03-13 13:50:03 +01:00
response_models.py fix: support gemini-3.1-flash-lite-preview by preserving thought_signature in tool calls (#568) 2026-03-13 16:43:01 +01:00
task_backend.py feat: introduce hindsight-api-slim and hindsight-all-slim packages (#560) 2026-03-13 13:50:03 +01:00
utils.py feat: introduce hindsight-api-slim and hindsight-all-slim packages (#560) 2026-03-13 13:50:03 +01:00