desktop: model picker omits custom openai-compatible provider models; search finds nothing

Summary

Models declared under custom providers using package: "@opencode/ai/providers/openai-compatible" do not appear in the desktop model picker, and picker search returns nothing for them. Only the currently active session's model is shown. The CLI lists all declared models correctly.

Environment

opencode version: 2.2.0
OS: macOS 26.x (Darwin 25.6.0, arm64)
Terminal: OpenCode Desktop (Electron)
Shell: /bin/bash
Install/channel: latest
Active plugins: none

Reproduction

Declare a custom OpenAI-compatible provider in ~/.config/opencode/opencode.json:

{
  "providers": {
    "bigbot": {
      "package": "@opencode/ai/providers/openai-compatible",
      "settings": { "baseURL": "http://<lan-host>:8080/v1" },
      "models": {
        "deepseek-v4.1-flash": {},
        "glm-5.3-flash": {}
      }
    }
  }
}

Restart the desktop app cold (also tried: config reload, killing the opencode server process).
Open the model picker and search for the declared models.

Expected Behavior

All declared custom provider models appear in the picker and are selectable/searchable, matching CLI behavior.

Actual Behavior

Only the active session's model (deepseek-v4.1-flash) appears in the picker; glm-5.3-flash and a second custom provider's models never appear, and picker search finds nothing for them. Persists across reload, app restart, and server restart.

Server log shows:

SessionRunnerModel.ModelInitializationError: Cannot initialize bigbot/deepseek-v4.1-flash:
No versions available for openai-compatible

The CLI (opencode TUI) lists all three declared models correctly, and the running service's config API also returns all of them, so the service sees the models — the desktop picker's resolution path appears to drop custom openai-compatible providers.

Related: the desktop also cannot add local providers through its provider-adding UI (known issue); the config had to be hand-edited.

Additional Context

Endpoints verified working: /v1/models returns the loaded models on both endpoints.
Requesting a declared-but-unloaded model ID succeeds (server serves whatever is loaded), so this is purely a picker/resolver-side issue.
Possibly relevant: ModelResolver.loadCatalogPackage treating openai-compatible as a catalog package with "no versions".

Thanks for the detailed report! I think this is a duplicate of #50726.

Possible duplicates and related issues, ordered from most to least likely:

desktop: custom provider models missing from /models dialog #50726: the same problem. Models from a config-declared @opencode/ai/providers/openai-compatible provider are missing from the desktop model picker and search, while the server returns them and inference works. A workaround was reported there: in the providers UI, turn the Personal / OpenCode group off and back on. After that, the custom models showed up in the picker. Could you try that and see if it helps?
opencode2: user-defined local provider models never enter the model catalog #42856: config-declared custom providers never make it into the model catalog. Related, but in your case the server does list the models.
v2: config-declared custom-provider models silently become 0 in catalog when models.opencode.ai fetch fails at boot (provider survives, resolution-by-ID still works) #41360: config-declared custom models get dropped from the catalog when the models.opencode.ai fetch fails at startup.

The No versions available for openai-compatible log line might be a separate problem. If it still shows up once the picker lists your models, or if models from one provider work and the other's don't, please say so here.

The bot will close this issue as a duplicate in 1 day unless you reply explaining how your problem is different.

Actually it was simpler than that. They were turned off in the model section of the settings menu. They were added by hand to opencode.json. Having them come in switched off by default is an odd choice. The No versions available happened when I tried to add the provider/model in the UI, and that's why I needed to add them manually to opencode.json. I believe that's a known issue, and if that's the case, you can close mine entirely.

…

On Wed, Oct 28, 2026 at 12:09 AM opencodejr[bot] ***@***.***> wrote: *opencodejr[bot]* left a comment (anomalyco/opencode#59840) <#59840 (comment)> Thanks for the detailed report! I think this is a duplicate of #50726 <#50726>. Possible duplicates and related issues, ordered from most to least likely: - #50726 <#50726>: the same problem. Models from a config-declared @opencode/ai/providers/openai-compatible provider are missing from the desktop model picker and search, while the server returns them and inference works. A workaround was reported there: in the providers UI, turn the Personal / OpenCode group off and back on. After that, the custom models showed up in the picker. Could you try that and see if it helps? - #42856 <#42856>: config-declared custom providers never make it into the model catalog. Related, but in your case the server does list the models. - #41360 <#41360>: config-declared custom models get dropped from the catalog when the models.opencode.ai fetch fails at startup. The No versions available for openai-compatible log line might be a separate problem. If it still shows up once the picker lists your models, or if models from one provider work and the other's don't, please say so here. The bot will close this issue as a duplicate in 1 day unless you reply explaining how your problem is different. — Reply to this email directly, view it on GitHub <#59840?email_source=notifications&email_token=BRUOWJ54MWX4DG27QKHDAZ35WGESJA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTMNBWGIYDKOBVGI2KM4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2KYZTPN52GK4S7MNWGSY3L#issuecomment-6462058524>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/BRUOWJZMSP7LI3CRILKYYDT5WGESJAVCNFSNUABFKJSXA33TNF2G64TZHM4TONJXGM2DGMJZHNEXG43VMU5TMMBXGM3DCMRQHE32C5QC> . You are receiving this because you authored the thread.Message ID: <anomalyco/opencode/issues/59840/6462058524 ***@***.***>

We haven't heard back, so I'm closing this as a duplicate of #50726. That issue covers the same problem: models from config-declared openai-compatible providers are missing from the desktop model picker and search, even though the server lists them. Please follow #50726 for updates. The workaround reported there may help in the meantime: in the providers UI, turn the Personal / OpenCode group off and back on.

If your problem turns out to be different, please reply here and we can reopen this. That includes the case where the No versions available for openai-compatible error remains after your models show up in the picker. Thanks again for the detailed report!

Navigation Menu

Repository navigation

desktop: model picker omits custom openai-compatible provider models; search finds nothing #59840

Summary

Environment

Reproduction

Expected Behavior

Actual Behavior

Additional Context

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

desktop: model picker omits custom openai-compatible provider models; search finds nothing #59840

Description

Summary

Environment

Reproduction

Expected Behavior

Actual Behavior

Additional Context

Activity

opencodejr commented on Oct 28, 2026

nakedhavoc commented on Oct 29, 2026

opencodejr commented on Oct 29, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions