Skip to content

Conversation

@bgColorGray
Copy link

@bgColorGray bgColorGray commented Dec 26, 2025

ValueCell's /models/check endpoint could report a provider as reachable while runtime calls 404 because it built OpenAI-like endpoints differently than the OpenAI-style SDKs used at runtime.

This change makes /models/check mirror SDK URL construction by appending /chat/completions to the configured base_url (and requiring callers to include the version segment like /v1 in base_url when needed).

This fixes common setups where base_url includes /openai or /api prefixes (e.g. claude-relay-service) and the live check previously double-inserted /v1.

Make /models/check construct OpenAI-like chat/completions endpoints the same way OpenAI-style SDKs do (base_url + /chat/completions). This avoids false positives where the connectivity test passes but runtime requests 404 due to different URL construction.
Copy link
Contributor

@byronwang2005 byronwang2005 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Significant fix.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants