Add The Grid provider

[CastilloLuis]

Summary

Adds The Grid as a provider.

The Grid is an OpenAI-compatible inference API whose model ids are market instruments rather than fixed foundation models. A caller selects a quality tier — text-standard, code-prime, agent-max — and The Grid acquires qualifying inference on its market to serve the request. Lab-specific markets are also available (claude-opus-latest, kimi-latest, gpt-sol-latest, and others).

Mirrors the existing Eden AI provider: openai.OpenAI against a base_url, with OpenAICompliantMessageConverter on the response so tool calls and usage come back normalized. Auto-discovered by ProviderFactory via the *_provider.py filename convention, so no registry change is needed.

Files

File
aisuite/providers/thegrid_provider.py	the provider
tests/providers/test_thegrid_provider.py	5 mock-based tests
guides/thegrid.md	setup guide
guides/README.md	link to it
README.md	one word added to the inline provider list

One behaviour worth calling out

Because an instrument pools several backing models, the model field of a response names the model that actually served the request, not the instrument requested. Asking for text-standard can come back as openai/gpt-oss-120b. That is intentional on The Grid's side — it is how they stay transparent about what served you — but it will surprise a caller who assumes the two match, so it is documented in the guide, in the class docstring, and pinned by a test.

Verification

Mock-based tests, per the repo convention:

$ pytest tests/providers/test_thegrid_provider.py -q
.....                                                                    [100%]
5 passed in 1.01s

$ black --check aisuite/providers/thegrid_provider.py tests/providers/test_thegrid_provider.py
All done! ✨ 🍰 ✨
2 files would be left unchanged.

And end to end against the live API, since mocks cannot tell you the base URL or auth is right:

provider discovered: True
  thegrid:text-standard   -> content='OK'
  thegrid:code-prime      -> content='OK'
  tool_calls: get_weather({"city": "San Francisco"})

Tool calling is exercised deliberately — #121 asked for it on a previous provider, so it is here from the start rather than as a follow-up.

Notes

• No new dependency. Reuses the existing openai client, so no pyproject.toml extra is needed (same as deepseek, ollama, lmstudio).
• No registry, enum or __init__.py change — discovery is by filename.