Model discovery via /model

[OriLevy]

I noticed the v0.1.8 release stated that /model should now support openAI compatible model discovery. Does this require any configuration? Not working by default with openrouter as the provider for me on the new release

[OriLevy]

Thought the issue might be that model discovery doesn't respect .openclaude-profile.json so I tested with local environment variables and still no luck

[Vasanthdev2004]

The v0.1.8 note means /model can do discovery for some OpenAI-compatible backends, but it is not guaranteed to work identically across every provider that speaks the OpenAI-style API.

OpenRouter is the main case where this can be confusing:

• it is supported through the OpenAI-compatible path
• but model discovery may still depend on the exact endpoint behavior and what the current implementation is querying
• saved profile vs env var differences can also make it look like discovery is failing for one path but not another

Since you already tested both profile and direct env vars, this sounds less like a config typo and more like either:

• an implementation gap in OpenRouter model discovery, or
• a provider-specific response shape/expectation mismatch

If you can, please paste:

• the exact provider/model config you used (with secrets redacted)
• whether normal prompting through OpenRouter works even though /model discovery does not
• latest release vs latest main

If prompting works but /model discovery still does not, we should probably track that as a concrete provider-discovery bug rather than setup confusion.

[Vasanthdev2004]

Following up on the earlier reply here — OpenRouter model discovery via /model is a known gap right now. The v0.1.8 release added OpenAI-compatible discovery, but OpenRouter's model listing endpoint returns a different response shape than vanilla OpenAI, so discovery can silently fail even though regular prompting works fine.

Current workaround:

• Set your OpenRouter model explicitly via env vars or /provider rather than relying on /model discovery
• Normal prompting through OpenRouter works; it is only the /model listing that is broken

This is on the radar as a provider-discovery bug. If you want to track progress, the issue is effectively the same as what was described in the earlier reply — /model discovery for OpenRouter specifically.