Unexpected endpoint or method

[inquam]

Doens't seem to locate models. In the dev-tab of LM Studio I see:

Unexpected endpoint or method. (GET /v1/models). Returning 200 anyway

Using LM Studio 0.4.2

[HarelMil]

+1

I noticed that it happens on my configuration where an Authorization header is required.
In other words, if I enforce Authorization with a Bearer token the plugin is probably not using it (a guess). When I downgraded the security by omitting this requirement it worked.

[agustif]

opencode-lmstudio@1.0.0-rc.1 is now available and should address both behaviors reported here.

The RC performs discovery with GET /api/v0/models, while the provider's inference URL remains OpenAI-compatible under /v1. When an apiKey is configured, discovery sends it as an Authorization: Bearer ... header. Both endpoint construction and authenticated discovery are covered by the release tests.

Could you retest with the exact RC, including the authenticated configuration if applicable?

{
  "$schema": "https://opencode.ai/config.json",
  "plugin": ["opencode-lmstudio@1.0.0-rc.1"]
}

Please confirm whether LM Studio now receives a successful GET /api/v0/models, whether models appear in OpenCode, and your OpenCode/LM Studio versions. Do not post the token itself. If discovery works without the unexpected /v1/models warning or authentication failure, please confirm that this issue can be closed; otherwise, include the sanitized provider configuration and current LM Studio status/error.

See #34 for the full RC configuration and feedback template.

[agustif]

opencode-lmstudio@1.0.0-rc.2 now implements LM Studio 0.4's documented endpoint and authentication boundary.

Discovery uses native GET /api/v1/models; inference remains /v1/chat/completions. A configured provider apiKey is sent as a Bearer token for discovery and inference. Local/private automatic lookup uses LM_API_TOKEN, then LMSTUDIO_API_KEY. Authentication failures and invalid native responses do not silently fall back to another endpoint.

Please pin "plugin": ["opencode-lmstudio@1.0.0-rc.2"], restart OpenCode, and retest with your LM Studio 0.4.x setup, including required authentication if applicable. If models appear without an unexpected /v1/models request or 401/403, please confirm this issue can be closed. Otherwise, include the exact versions, sanitized provider configuration, requested path, and HTTP status—never the token itself. See #34 for the shared template.

[aelfwyne]

Please just add proper support for the request. End users need it to work in a compatible way.

[agustif]

Please just add proper support for the request. End users need it to work in a compatible way.

could you explain exactly what you mean? Does it not work now for you while using the latest rc.2 version?

will make a regular 1.0 release soon, just wanted to daily-drive it myself for a while before pushing it out of beta.

Did you find any issues with it?