-
Notifications
You must be signed in to change notification settings - Fork 33.6k
Proxy support for LLM API calls #5454
Copy link
Copy link
Open
Labels
P3Low — cosmetic, nice to haveLow — cosmetic, nice to havearea/configConfig system, migrations, profilesConfig system, migrations, profilescomp/agentCore agent loop, run_agent.py, prompt builderCore agent loop, run_agent.py, prompt buildertype/featureNew feature or requestNew feature or request
Description
Metadata
Metadata
Assignees
Labels
P3Low — cosmetic, nice to haveLow — cosmetic, nice to havearea/configConfig system, migrations, profilesConfig system, migrations, profilescomp/agentCore agent loop, run_agent.py, prompt builderCore agent loop, run_agent.py, prompt buildertype/featureNew feature or requestNew feature or request
Type
Fields
Give feedbackNo fields configured for issues without a type.
Problem
Only the Telegram adapter respects HTTP_PROXY / HTTPS_PROXY env vars. The main LLM API client (OpenAI SDK) does not have explicit proxy configuration. Users behind corporate proxies or VPNs cannot route API calls through their proxy.
Proposed Solution
The OpenAI SDK uses httpx under the hood, which respects standard proxy env vars by default. Verify this works end-to-end, and add explicit proxy support via
agent.proxyconfig key for cases where env vars are not sufficient (e.g. different proxy for API vs web tools). Also ensure the Anthropic native client path respects the same proxy config.