[knowhycodata]

Closes #92

What this changes

Desktop setup lists LLMTR beside Google Gemini and xAI. The row uses the same endpoint credential route as those presets, fills in https://llmtr.com/v1, starts the model as qwen/qwen3.8-27b-free, and links to LLMTR's API keys page. Disclosure: I maintain LLMTR.

For hand-configured deployments, the README, docs/configuration.md and .env.example show OPENAI_BASE_URL=https://llmtr.com/v1, OPENAI_API_KEY=llmtr-... and BOT_MODEL=<catalogue id>, with BOT_PROVIDER left as openai and BOT_REASONING_EFFORT unset (LLMTR takes reasoning only as a boolean).

Touched: desktop/src-tauri/src/provider.rs (catalogue row + tests), desktop/src/ProviderPicker.tsx (preset) and its test, README.md, docs/configuration.md, .env.example, CHANGELOG.md.

If you'd rather ship a different default than the free qwen/qwen3.8-27b-free, I'm happy to change it.

Where it runs

• New state that outlives a request? None. A static catalogue row and a frontend preset.
• What happens on the second replica? Nothing different; it produces the same env values any OpenAI-compatible endpoint already does.
• Anything serialised? No.
• Anything fanned out to a browser? No.
• New listener, port, or schedule? No.

Boundary and audit

• Every acting call still goes through the gateway: resolve, decide, audit, then act. (Unchanged; this only picks the model endpoint.)
• New refusals and new failures each write a row. (No new refusals or failures.)
• Nothing new is trusted from the client that the server can resolve itself.

Changelog

• "Desktop setup offers LLMTR" under Unreleased in openbot/CHANGELOG.md.

Proof

• In openbot/: bun run format:check, bun run lint, bun run typecheck pass. In openbot/desktop/: bun run typecheck and bun run build pass.
• bun test desktop/src shared/model-providers.test.ts: 183 pass, 0 fail.
• cargo test --lib provider:: (Tauri crate): 5 pass (on a machine with Rust 1.87+).
• Live check from a fresh clone. I didn't run the full Docker stack; I ran the built-in Bot service directly (cd openbot/agent-bot && bun src/index.ts) with OPENAI_BASE_URL=https://llmtr.com/v1, an LLMTR key in OPENAI_API_KEY and BOT_MODEL=llmtr/qwen3-8-27b, then posted AG-UI runs to /ag-ui:
  • With an add_tasks tool offered, the run streamed TOOL_CALL_START / TOOL_CALL_ARGS ({"tasks":[{"title":"Call the dentist"}]}) / TOOL_CALL_END and RUN_FINISHED in about 2.6 s.
  • With the tool result appended, the next run streamed "Added the task "Call the dentist" to your list." and RUN_FINISHED in about 2 s.
  • The same run with qwen/qwen3.8-27b-free hit a temporary upstream error on the free model at that moment; the Bot surfaced it cleanly as RUN_ERROR.
• No screenshot of the setup screen; the row is covered by the ProviderPicker and provider.rs tests above.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	⚠️ Failed 2026-10-29T08:04:18.611264Z We couldn’t complete this review. Comment @codex review to try again.	0d79e2f	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.