Compress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.
Python agentaianthropicclaude-codecompressioncontext-engineeringcontext-windowcursorfastapilangchainllmmcpopenaiprompt-engineeringproxypythonragtoken-optimizationtokenstypescript
fix(memory): filter inactive graph-expanded results (#2210)
Chesteraa4515cf7aaf
diff --git a/headroom/memory/backends/local.py b/headroom/memory/backends/local.pyindex 1dca91b8e..6fafeb91b 100644--- a/headroom/memory/backends/local.py+++ b/headroom/memory/backends/local.py@@ -488,20 +488,26 @@ class LocalBackend:new_memory_ids = related_memory_ids - seen_memory_idsfor mem_id in new_memory_ids:memory = await self._hierarchical_memory.get(mem_id)- if memory and memory.user_id == user_id:- # Filter by session_id if specified (security: prevent session leakage)- if session_id is not None and memory.session_id != session_id:- continue- # Add with lower score since it's from graph expansion- results.append(- MemorySearchResult(- memory=memory,- score=0.5, # Default score for graph-expanded results- related_entities=list(memory.entity_refs),- related_memories=[],- )+ if (+ memory is None+ or memory.user_id != user_id+ or memory.valid_until is not None+ or memory.superseded_by is not None+ ):+ continue+ # Filter by session_id if specified (security: prevent session leakage)+ if session_id is not None and memory.session_id != session_id:+ continue+ # Add with lower score since it's from graph expansion+ results.append(+ MemorySearchResult(+ memory=memory,+ score=0.5, # Default score for graph-expanded results+ related_entities=list(memory.entity_refs),+ related_memories=[],)- seen_memory_ids.add(mem_id)+ )+ seen_memory_ids.add(mem_id)# Filter by specified entities if providedif entities:diff --git a/tests/test_memory/test_local_backend_search.py b/tests/test_memory/test_local_backend_search.pynew file mode 100644index 000000000..5783fa04d--- /dev/null+++ b/tests/test_memory/test_local_backend_search.py@@ -0,0 +1,76 @@+"""Focused tests for LocalBackend search result filtering."""++from __future__ import annotations++from datetime import datetime, timezone+from types import SimpleNamespace+from unittest.mock import AsyncMock++import pytest++from headroom.memory.backends.local import LocalBackend+from headroom.memory.models import Memory+++def _backend_with_related_memory(related_memory: Memory) -> LocalBackend:+ seed = Memory(+ id="seed-memory",+ content="Alice manages Project X",+ user_id="alice",+ entity_refs=["Project X"],+ )+ vector_result = SimpleNamespace(memory=seed, similarity=0.9)++ backend = LocalBackend()+ backend._initialized = True+ backend._hierarchical_memory = SimpleNamespace(+ search=AsyncMock(return_value=[vector_result]),+ get=AsyncMock(return_value=related_memory),+ )+ backend._graph = SimpleNamespace(+ get_entity_by_name=AsyncMock(return_value=SimpleNamespace(id="project-x")),+ query_subgraph=AsyncMock(+ return_value=SimpleNamespace(+ entities=[SimpleNamespace(metadata={"source_memory_id": related_memory.id})],+ relationships=[],+ )+ ),+ )+ return backend+++async def test_graph_expansion_includes_active_related_memory() -> None:+ related = Memory(+ id="related-memory",+ content="Project X uses Python",+ user_id="alice",+ entity_refs=["Project X", "Python"],+ )+ backend = _backend_with_related_memory(related)++ results = await backend.search_memories("Alice's work", "alice", include_related=True)++ assert [result.memory.id for result in results] == ["seed-memory", "related-memory"]++[email protected]("inactive_field", ["valid_until", "superseded_by"])+async def test_graph_expansion_excludes_inactive_related_memory(+ inactive_field: str,+) -> None:+ related = Memory(+ id="related-memory",+ content="Outdated Project X detail",+ user_id="alice",+ entity_refs=["Project X"],+ )+ if inactive_field == "valid_until":+ related.valid_until = datetime.now(timezone.utc).replace(tzinfo=None)+ else:+ related.superseded_by = "replacement-memory"+ backend = _backend_with_related_memory(related)++ results = await backend.search_memories("Alice's work", "alice", include_related=True)++ assert [result.memory.id for result in results] == ["seed-memory"]