Skip to content

docs: add LLMTR provider - #42867

Open
knowhycodata wants to merge 1 commit into
anomalyco:devfrom
knowhycodata:llmtr-docs
Open

docs: add LLMTR provider#42867
knowhycodata wants to merge 1 commit into
anomalyco:devfrom
knowhycodata:llmtr-docs

Conversation

@knowhycodata

Copy link
Copy Markdown

Issue for this PR

Closes #42866

Type of change

  • Bug fix
  • New feature
  • Refactor / code improvement
  • Documentation

What does this PR do?

Adds the missing LLMTR section to the provider directory in providers.mdx. One file, docs only.

LLMTR already works in opencode — it is in models.dev and live in the catalog mirror (llmtr, npm @ai-sdk/openai-compatible, env LLMTR_API_KEY), so it resolves through the generic openai-compatible path with no runtime code. The only thing missing was the docs entry, so this is purely that.

I put it after LLM Gateway and before Poolside, and followed the same four-step connect/models shape the neighbouring entries use. The config snippet uses the existing "add a model to an existing provider" pattern from the OpenRouter and LLM Gateway sections rather than any new syntax, because the models.dev snapshot only carries the Turkey-hosted llmtr/* ids — anything else from the gateway has to be named in config.

How did you verify your code works?

Docs build, on this branch:

  • bun run --cwd packages/web build[build] Complete!, exit 0. The section renders at /docs/providers with anchor id="llmtr".
  • prettier --check packages/web/src/content/docs/providers.mdxAll matched files use Prettier code style!
  • bun run lint (oxlint) reports one error and 4855 warnings, but they are all pre-existing on dev and in .ts/.tsx files; oxlint does not process .mdx.

Provider behaviour, on the released opencode-ai@1.18.18 with LLMTR_API_KEY set:

  • opencode models lists the llmtr/* catalog models.
  • opencode run --model llmtr/gemma-4 "Reply with exactly: OK"OK.
  • With the exact config snippet from this PR in opencode.json, opencode models gains llmtr/anthropic/claude-opus-4.8 and llmtr/openai/gpt-5.4, and opencode run --model llmtr/anthropic/claude-opus-4.8 returns a real response. Both ids were also checked directly against POST /v1/chat/completions (HTTP 200).

I deliberately did not name llmtr/sincap anywhere: it is retired and returns HTTP 410. It is still in the models.dev snapshot; a separate PR fixes that catalog drift.

Screenshots / recordings

Not a UI change — content added to an existing docs page.

Checklist

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

Context: this was previously #31315 / #31316. The PR was closed by the automated PR-cleanup bot and the issue by the 60-day stale bot, not on technical merit, and neither can be reopened, so this is a refile against current dev.

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.

[FEATURE]: Document LLMTR provider in providers.mdx

1 participant