molecule-ai/hermes-agent
38
0
Fork 0
Code Issues 5 Pull Requests 7 Actions Packages Projects Releases Wiki Activity

fix(agent): disable stale stream timeout for local providers (#6368)

Browse Source 
Local inference providers (Ollama, oMLX, llama-cpp) can take 300+ seconds
for prefill on large contexts. The 180s stale stream detector was killing
these connections while the provider was still processing.

Uses the existing is_local_endpoint() (proper URL parsing with RFC-1918,
localhost, WSL detection) instead of ad-hoc substring matching. The stale
timeout is only disabled when the user hasn't explicitly set
HERMES_STREAM_STALE_TIMEOUT — explicit user config is always honored.

Fixes #5889
This commit is contained in:
Teknium 2026-04-08 19:53:39 -07:00 committed by GitHub
parent 6e3f7f3610
commit ae4a884e8d
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 18 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

29
run_agent.py
View File

@ -4728,18 +4728,25 @@ class AIAgent:
self._close_request_openai_client(request_client, reason="stream_request_complete")
_stream_stale_timeout_base = float(os.getenv("HERMES_STREAM_STALE_TIMEOUT", 180.0))
# Scale the stale timeout for large contexts: slow models (like Opus)
# can legitimately think for minutes before producing the first token
# when the context is large. Without this, the stale detector kills
# healthy connections during the model's thinking phase, producing
# spurious RemoteProtocolError ("peer closed connection").
_est_tokens = sum(len(str(v)) for v in api_kwargs.get("messages", [])) // 4
if _est_tokens > 100_000:
_stream_stale_timeout = max(_stream_stale_timeout_base, 300.0)
elif _est_tokens > 50_000:
_stream_stale_timeout = max(_stream_stale_timeout_base, 240.0)
# Local providers (Ollama, oMLX, llama-cpp) can take 300+ seconds
# for prefill on large contexts. Disable the stale detector unless
# the user explicitly set HERMES_STREAM_STALE_TIMEOUT.
if _stream_stale_timeout_base == 180.0 and self.base_url and is_local_endpoint(self.base_url):
_stream_stale_timeout = float("inf")
logger.debug("Local provider detected (%s) — stale stream timeout disabled", self.base_url)
else:
_stream_stale_timeout = _stream_stale_timeout_base
# Scale the stale timeout for large contexts: slow models (like Opus)
# can legitimately think for minutes before producing the first token
# when the context is large. Without this, the stale detector kills
# healthy connections during the model's thinking phase, producing
# spurious RemoteProtocolError ("peer closed connection").
_est_tokens = sum(len(str(v)) for v in api_kwargs.get("messages", [])) // 4
if _est_tokens > 100_000:
_stream_stale_timeout = max(_stream_stale_timeout_base, 300.0)
elif _est_tokens > 50_000:
_stream_stale_timeout = max(_stream_stale_timeout_base, 240.0)
else:
_stream_stale_timeout = _stream_stale_timeout_base
t = threading.Thread(target=_call, daemon=True)
t.start()