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(proxy): classify cancelled relay tasks correctly in termination cause
Adryan Eka Vandra6ae0fcd8b52f
diff --git a/headroom/proxy/handlers/openai.py b/headroom/proxy/handlers/openai.pyindex b34afa977..54aaa95f9 100644--- a/headroom/proxy/handlers/openai.py+++ b/headroom/proxy/handlers/openai.py@@ -1949,8 +1949,19 @@ class OpenAIHandlerMixin:# (e.g. server shutdown).for t in done:exc = None- with contextlib.suppress(Exception):- exc = t.exception()+ # Cancelled tasks raise CancelledError from+ # .exception(); surface it explicitly so the+ # downstream ``isinstance(exc, CancelledError)``+ # branches actually run. For any other+ # unexpected state (``InvalidStateError`` if+ # the task somehow isn't done — shouldn't+ # happen post-gather but defensive), we+ # suppress and leave ``exc`` as ``None``.+ if t.cancelled():+ exc = asyncio.CancelledError()+ else:+ with contextlib.suppress(asyncio.InvalidStateError):+ exc = t.exception()task_name = t.get_name() or ""if t is client_task:if client_relay_error is not None: