fix(honcho): extend local client timeout - #8527
Conversation
|
I am still learning, but I noticed that some checks are failing. Maybe we should fix them before merging. |
These failures come from other upstream changes, not related to the changes introduced in this PR. |
|
Thanks for the focused local-Honcho fix. Current main still gives an unconfigured loopback client the generic 30-second default at Problems
Suggested changes
Automated hermes-sweeper review. |
|
Superseded by configuration: after #66052 (commit b170f52) merged #57437 by @liuhao1024 and #61675 by @Elektrofussel, self-hosted users can set a long timeout in config.yaml (honcho.timeout) and it now takes effect live in long-lived processes — and local-host detection covers LAN/VPN addresses, not just localhost. That covers the slow-local-model case this PR targeted without a hardcoded blanket 300s override. Your diagnosis that local dialectic calls can exceed 60s was right and informed the direction — thanks @DomGrieco! |
Summary
Why
Local/self-hosted Honcho deployments can take longer than the SDK's default 60 second HTTP timeout for dialectic-style calls. When Hermes is pointed at a localhost/127.0.0.1 Honcho instance, this can cause Hermes to give up before Honcho responds even though the local server is still working.
This change keeps the scope narrow:
timeout=parameter rather than introducing Hermes-specific retry logicTest Plan
/Users/dgrieco/.hermes/hermes-agent/venv/bin/python -m py_compile plugins/memory/honcho/client.py tests/honcho_plugin/test_client.py/Users/dgrieco/.hermes/hermes-agent/venv/bin/python -m pytest tests/honcho_plugin/test_client.py -q