docs(openui-chat): document Pokee AI as an OpenAI-compatible provider

[devin-ai-integration]

What

examples/openui-chat already routes through any OpenAI-compatible endpoint via OPENAI_BASE_URL, so running it on Pokee-Isaac needs no code change — only config. This documents that config and adds the missing env.example (other examples like hands-on-table-chat already ship one).

Pokee's /v1/chat/completions is OpenAI-shaped and the only model id is pokee-isaac (verified against api.pokee.ai/v1/models and a live completion), so runTools streaming in src/app/api/chat/route.ts works unchanged.

Changes

• examples/openui-chat/README.md: Pokee example alongside the existing OpenRouter one (OPENAI_BASE_URL=https://api.pokee.ai/v1, OPENAI_MODEL=pokee-isaac)
• examples/openui-chat/env.example: new, with the OpenAI defaults and the Pokee override commented out

Test Plan

• Verified locally

Live pokee-isaac completion through api.pokee.ai/v1 succeeded; the example app was then run against Pokee end to end (recording shared with the requester).

Checklist

• I linked a related issue, if applicable
• I updated docs/README when needed
• I considered backwards compatibility

Link to Devin session: https://app.devin.ai/sessions/0226b24eba644deaa4af169313f98838
Requested by: @zahlekhan

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment, CI, and merge conflict monitoring

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
openui-docs	Ready	Preview	Aug 6, 2026 3:03pm

[devin-ai-integration]

Runtime testing: OpenUI chat example against Pokee AI (pokee-isaac)

Ran examples/openui-chat locally with the README's Pokee config (OPENAI_BASE_URL=https://api.pokee.ai/v1, OPENAI_MODEL=pokee-isaac) and exercised three prompts end-to-end in the browser. The documented config works — pokee-isaac streams OpenUI Lang and the components render, and tool-calling via the OpenAI SDK's runTools is compatible with Pokee.

Prompt	Result
Stock market analysis	✅ full dashboard
Plan a trip to Japan	✅ full itinerary
Latest sports news	⚠️ intermittent empty response

✅ "Stock market analysis" — parallel tool calls → table + chart

8 × get_stock_price (one per ticker), values matching the mocked bases in route.ts (NVDA $878.83, AAPL $188.15):

⚠️ Pre-existing issue this surfaced (not caused by this PR)

"Latest sports news" intermittently (~50%) returns a completely empty assistant bubble with no error. pokee-isaac keeps re-calling search_web until runTools hits DEFAULT_MAX_CHAT_COMPLETIONS = 10; the route's end handler then emits a bare [DONE] having forwarded zero content chunks.

Reproduced at the API level: attempt 1 → 3 tool calls / 288 content chunks (fine); attempt 2 → 16 tool calls / 0 content chunks (empty).

Two independent hardening options for route.ts, both out of scope for this docs change:

1. In the end handler, if zero content chunks were forwarded, emit an error/fallback chunk instead of a bare [DONE].
2. Pass an explicit maxChatCompletions and/or nudge the system prompt to answer after tool results.

✅ "Plan a trip to Japan" — no-loop path

One search_web call, six accordions, numbered day-by-day timeline:

Caveats

• The example's tools are mocked, so content is fabricated by design; assertions target structure and mocked value ranges.
• Dev server logs were not captured, so "no server errors" is unverified.