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): don't crash inline memory extraction on a non-object <memory> block (#2470)
Abhay Singhe00c6ff81ce2
diff --git a/headroom/memory/inline_extractor.py b/headroom/memory/inline_extractor.pyindex bcbf6aabf..f560ae3ed 100644--- a/headroom/memory/inline_extractor.py+++ b/headroom/memory/inline_extractor.py@@ -126,9 +126,21 @@ def parse_response_with_memory(response_text: str) -> ParsedResponse:# Parse the JSONtry:data = json.loads(memory_json)- memories = data.get("memories", [])except json.JSONDecodeError as e:logger.warning(f"Failed to parse memory JSON: {e}")+ else:+ # The model controls the block content. A valid-JSON non-object+ # (e.g. a bare array) would make `.get` raise AttributeError, which+ # the JSONDecodeError guard above does not catch; likewise a+ # non-list `memories` would break downstream iteration.+ if isinstance(data, dict):+ extracted = data.get("memories", [])+ if isinstance(extracted, list):+ memories = extracted+ else:+ logger.warning("Memory JSON 'memories' field is not a list; ignoring")+ else:+ logger.warning("Memory JSON is not an object; ignoring")return ParsedResponse(content=content,diff --git a/tests/test_memory_wrapper.py b/tests/test_memory_wrapper.pyindex c09a4c5e3..0bd009713 100644--- a/tests/test_memory_wrapper.py+++ b/tests/test_memory_wrapper.py@@ -205,3 +205,22 @@ def test_memory_api_methods_delegate_to_underlying_memory() -> None:assert api.clear() == 2assert api.stats() == {"total": 2}assert fake_memory.last_clear == {"user_id": "alice"}+++def test_parse_response_with_memory_tolerates_non_object_memory_block() -> None:+ from headroom.memory.inline_extractor import parse_response_with_memory++ # The model controls the <memory> block. A valid-JSON non-object (a bare+ # array) or a non-list `memories` field must not crash the parse: the+ # `json.loads` succeeds, so the JSONDecodeError guard does not apply, and+ # `.get`/iteration on the wrong type would otherwise raise.+ assert parse_response_with_memory('hi <memory>["x"]</memory> bye').memories == []+ assert parse_response_with_memory('<memory>{"memories": "nope"}</memory>').memories == []++ # Malformed JSON is still handled, and a well-formed block still parses.+ assert parse_response_with_memory("<memory>{not json</memory>").memories == []+ parsed = parse_response_with_memory(+ 'text <memory>{"memories": [{"content": "User likes Python"}]}</memory>'+ )+ assert parsed.memories == [{"content": "User likes Python"}]+ assert parsed.content == "text"