feat(opencode): add dynamic model fetching for OpenAI-compatible provider by shivamashtikar · Pull Request #14277 · anomalyco/opencode

shivamashtikar · 2026-02-19T13:43:47Z

Issue for this PR

Type of change

Bug fix
New feature
Refactor / code improvement
Documentation

What does this PR do?

Adds a shouldFetchModels config option that dynamically fetches available models from the provider's OpenAI-compatible API at startup. This eliminates the need to manually list every model in opencode.json for custom providers.

Comparison with PR #13896:
PR #13896 implements LiteLLM-specific model loading. This PR provides a generic implementation supporting any OpenAI-compatible provider (LiteLLM, Ollama, vLLM, custom proxies, etc.). Both try /model/info first, but this PR adds proper fallback to the standard /models endpoint.

How it works:

When a custom provider has baseURL configured, models are fetched in the background (unless shouldFetchModels: false)
First tries /model/info for rich metadata (limits, costs, capabilities)
Falls back to /models endpoint if /model/info unavailable
Manual config models always override fetched ones
Non-blocking - startup proceeds immediately, models merged when response arrives

Example:

{
  "provider": {
    "my-proxy": {
      "npm": "@ai-sdk/openai-compatible",
      "options": {
        "baseURL": "https://my-proxy.com/v1",
        "apiKey": "{env:API_KEY}"
      }
    }
  }
}

To disable:

{
  "provider": {
    "my-proxy": {
      "shouldFetchModels": false,
      "options": { "baseURL": "https://my-proxy.com/v1" }
    }
  }
}

How did you verify your code works?

bun test test/provider/fetch-models.test.ts - 4 tests pass
bun test test/session/llm.test.ts - 10 tests pass (no regressions)
Built standalone executable with --single flag and verified it works

Screenshots / recordings

N/A - no UI changes

Checklist

I have tested my changes locally
I have not included unrelated changes in this PR

github-actions · 2026-02-19T13:44:24Z

The following comment was made by an LLM, it may be inaccurate:

Found a potential duplicate:

PR #13896 - feat(opencode): add auto loading models for litellm providers

feat(opencode): add auto loading models for litellm providers #13896

Why it's related: This PR appears to implement very similar functionality - automatic model loading for LiteLLM providers. Given that PR #14277 also adds dynamic model fetching for OpenAI-compatible providers (including LiteLLM), these PRs may be addressing the same or overlapping feature requests. You should verify whether PR #13896 was already merged or closed and check if #14277 extends/improves upon it.

shivamashtikar · 2026-02-19T13:49:12Z

this PR provides compatibility with any open-ai compatible endpoints along with litellm as opposed to #13896 which only handles for litellm

PratikNarola1 · 2026-02-19T13:54:18Z

LGTM

8x22b · 2026-02-19T19:41:17Z

PLEASE MERGE IT

espetro · 2026-02-20T12:30:05Z

up! we're looking forward to it!

shivamashtikar · 2026-02-25T18:59:36Z

Hi @alexyaroshuk @adamdotdevin could you please check this PR and let me know if it can be merged or if you need any more changes in it

alexyaroshuk · 2026-02-26T17:14:43Z

Hi @alexyaroshuk @adamdotdevin could you please check this PR and let me know if it can be merged or if you need any more changes in it

unit tests failing:

4 tests failed:
(fail) session.llm.stream > sends temperature, tokens, and reasoning options for openai-compatible models [31.00ms]
(fail) session.llm.stream > sends responses API payload for OpenAI models [63.00ms]
(fail) session.llm.stream > sends messages API payload for Anthropic models [94.00ms]
(fail) session.llm.stream > sends Google API payload for Gemini models [109.00ms]

I checked with upstream and confirmed that upstream does not have these failing. Your PR has these failing. You can verify by running bun test from packages/opencode

…iders Add fetchModels config option that fetches available models from a provider's API at startup. Tries LiteLLM /model/info first for rich metadata (limits, costs, capabilities), falls back to standard /models endpoint. Manually configured models override fetched ones.

…startup Move dynamic model fetching to a fire-and-forget background task within state(). Models are merged into the providers object asynchronously after state initialization completes, so startup is not delayed by network requests.

…to true Rename config field to shouldFetchModels for clarity. Default to true so custom providers with a baseURL automatically discover models without explicit opt-in. Update PR docs accordingly.

…iders - Add shouldFetchModels option to provider config (defaults to true) - Update tests to disable model fetching in test fixtures

- Test shouldFetchModels: false prevents API calls - Test manual models work when fetching is disabled - Test blacklist/whitelist filtering with manual models

shivamashtikar · 2026-03-02T23:23:34Z

Hi @alexyaroshuk , apologies i missed those. I've fixed the test cases and and also added new test cases related to my change, can you please review and merge if everything seems fine

github-actions · 2026-03-02T23:37:18Z

Thanks for updating your PR! It now meets our contributing guidelines. 👍

alexyaroshuk · 2026-03-03T00:33:18Z

Hi @alexyaroshuk , apologies i missed those. I've fixed the test cases and and also added new test cases related to my change, can you please review and merge if everything seems fine

i cannot merge, i simply try to help with review. Looks like the test issue is fixed, but merging is up to Adam

shivamashtikar · 2026-03-03T10:07:37Z

Sure @alexyaroshuk please help with the review

Mushoz mentioned this pull request Feb 19, 2026

feat(opencode): add auto loading models for litellm providers #13896

Open

2 tasks

balcsida mentioned this pull request Feb 20, 2026

feat(opencode): add LiteLLM provider with auto model discovery #14468

Open

6 tasks

shivamashtikar force-pushed the feat/dynamic-model-fetch branch 2 times, most recently from 6210e12 to 91ef011 Compare February 24, 2026 09:47

shivamashtikar added 3 commits March 3, 2026 04:13

refactor(opencode): rename fetchModels to shouldFetchModels, default …

765aa51

…to true Rename config field to shouldFetchModels for clarity. Default to true so custom providers with a baseURL automatically discover models without explicit opt-in. Update PR docs accordingly.

shivamashtikar force-pushed the feat/dynamic-model-fetch branch from 584e91e to 66329b2 Compare March 2, 2026 22:50

feat(opencode): add dynamic model fetching for OpenAI-compatible prov…

b1b97d6

…iders - Add shouldFetchModels option to provider config (defaults to true) - Update tests to disable model fetching in test fixtures

shivamashtikar force-pushed the feat/dynamic-model-fetch branch from 66329b2 to b1b97d6 Compare March 2, 2026 22:54

test(opencode): add tests for shouldFetchModels feature

e3d8e0a

- Test shouldFetchModels: false prevents API calls - Test manual models work when fetching is disabled - Test blacklist/whitelist filtering with manual models

github-actions bot added needs:compliance This means the issue will auto-close after 2 hours. and removed needs:compliance This means the issue will auto-close after 2 hours. labels Mar 2, 2026

github-actions bot mentioned this pull request Mar 3, 2026

📊 AI CLI 工具社区动态日报 2026-03-03 duanyytop/agents-radar#45

Closed

github-actions bot mentioned this pull request Mar 3, 2026

docs(providers): document Responses API support for custom providers #15847

Open

6 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(opencode): add dynamic model fetching for OpenAI-compatible provider#14277

feat(opencode): add dynamic model fetching for OpenAI-compatible provider#14277
shivamashtikar wants to merge 5 commits intoanomalyco:devfrom
shivamashtikar:feat/dynamic-model-fetch

shivamashtikar commented Feb 19, 2026 •

edited

Loading

Uh oh!

github-actions bot commented Feb 19, 2026

Uh oh!

shivamashtikar commented Feb 19, 2026

Uh oh!

PratikNarola1 commented Feb 19, 2026

Uh oh!

8x22b commented Feb 19, 2026

Uh oh!

espetro commented Feb 20, 2026

Uh oh!

shivamashtikar commented Feb 25, 2026 •

edited

Loading

Uh oh!

alexyaroshuk commented Feb 26, 2026

Uh oh!

shivamashtikar commented Mar 2, 2026 •

edited

Loading

Uh oh!

github-actions bot commented Mar 2, 2026

Uh oh!

alexyaroshuk commented Mar 3, 2026

Uh oh!

shivamashtikar commented Mar 3, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Conversation

shivamashtikar commented Feb 19, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Issue for this PR

Type of change

What does this PR do?

How did you verify your code works?

Screenshots / recordings

Checklist

Uh oh!

github-actions bot commented Feb 19, 2026

Uh oh!

shivamashtikar commented Feb 19, 2026

Uh oh!

PratikNarola1 commented Feb 19, 2026

Uh oh!

8x22b commented Feb 19, 2026

Uh oh!

espetro commented Feb 20, 2026

Uh oh!

shivamashtikar commented Feb 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

alexyaroshuk commented Feb 26, 2026

Uh oh!

shivamashtikar commented Mar 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions bot commented Mar 2, 2026

Uh oh!

alexyaroshuk commented Mar 3, 2026

Uh oh!

shivamashtikar commented Mar 3, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

shivamashtikar commented Feb 19, 2026 •

edited

Loading

shivamashtikar commented Feb 25, 2026 •

edited

Loading

shivamashtikar commented Mar 2, 2026 •

edited

Loading