fleet-memory/hindsight-api/hindsight_api/engine
Nicolò Boschi 5d05962db0
feat: add source facts token limits to consolidation and recall (#509)
* feat: add source facts token limits to consolidation and recall

- Add two new configurable (per-bank) parameters:
  - consolidation_source_facts_max_tokens: total token budget for source
    facts across all observations in the consolidation prompt (-1 = unlimited)
  - consolidation_source_facts_max_tokens_per_observation: per-observation
    cap so each observation gets a fair share of source facts (-1 = unlimited,
    default 256)
- Both are also exposed as recall API parameters via SourceFactsIncludeOptions
  (max_tokens and max_tokens_per_observation)
- Consolidation now uses resolve_full_config to respect bank-level overrides
- Improve consolidation prompt: temporal metadata (occurred_start=, | Involving:)
  is now clearly separated from observation text, with a concrete example showing
  the expected synthesis style and explicit rules not to copy raw fact lines
- Add tests for recall source facts capping and consolidation config forwarding
- Expose all three new fields in the control plane bank config UI
- Document new env vars in configuration.md
- Regenerate OpenAPI spec and all SDK clients

* fix: reorder observations UI fields and rename Label Groups to Entity Labels

* fix: revert Entities section title (only rename inner label)

* doc: add consolidation source facts and batch size fields to memory-banks docs
2026-03-06 13:01:33 +01:00
..
consolidation feat: add source facts token limits to consolidation and recall (#509) 2026-03-06 13:01:33 +01:00
directives feat: revisit mental models, directives and reflections (#179) 2026-01-22 17:13:16 +01:00
mental_models feat: revisit mental models, directives and reflections (#179) 2026-01-22 17:13:16 +01:00
parsers Add on_file_convert_complete extension hook after file-to-markdown conversion (#507) 2026-03-06 09:56:53 +01:00
providers refactor: replace set_gemini_safety_settings() with LLMProvider.with_config() (#474) 2026-03-03 15:00:32 +01:00
reflect feat: entity labels — optional, free_values, multi_value, UI polish (#450) 2026-03-02 13:05:25 +01:00
retain fix: resolve TypeError when LLM returns invalid JSON across all retries (#488) (#490) 2026-03-04 14:17:55 +01:00
search fix: replace additive combined scoring with multiplicative CE boosts (#494) 2026-03-05 11:12:28 +01:00
storage fix(storage): use dynamic schema_getter in PostgreSQLFileStorage for multi-tenant (#440) 2026-02-25 15:19:08 +01:00
__init__.py feat: extensions (#54) 2025-12-22 11:05:23 +01:00
cross_encoder.py fix: zeroentropy rerank URL missing /v1 prefix and MCP retain async_processing param (#460) 2026-03-02 10:32:01 +01:00
db_budget.py fix: batch queries on recall (#149) 2026-01-13 13:20:22 +01:00
db_utils.py fix(performance): improve recall and retain performance on large banks (#469) 2026-03-03 13:35:22 +01:00
embeddings.py fix: pass encoding_format="float" in LiteLLM embedding calls (#434) 2026-02-25 10:32:05 +01:00
entity_resolver.py fix(performance): improve recall and retain performance on large banks (#469) 2026-03-03 13:35:22 +01:00
interface.py feat: add tags filtering and q description fix for list documents API (#468) 2026-03-02 17:03:16 +01:00
llm_interface.py feat: support Batch API for retain (openai/groq) (#365) 2026-02-16 13:31:50 +01:00
llm_wrapper.py refactor: replace set_gemini_safety_settings() with LLMProvider.with_config() (#474) 2026-03-03 15:00:32 +01:00
memory_engine.py feat: add source facts token limits to consolidation and recall (#509) 2026-03-06 13:01:33 +01:00
operation_metadata.py fix: improve async batch retain with large payloads (#366) 2026-02-16 12:51:42 +01:00
query_analyzer.py fix(performance): improve recall and retain performance on large banks (#469) 2026-03-03 13:35:22 +01:00
response_models.py feat: include source facts in observation recall (#404) 2026-02-19 14:54:19 +01:00
task_backend.py fix: retain async fails if timestamp is set (#251) 2026-01-30 13:04:33 +01:00
utils.py feat: implement hierarchical configuration (system, tenant, bank) (#329) 2026-02-12 13:14:57 +01:00