Summary

Adds ModelsLab as an LLM provider for browser-use, following the OpenRouter provider pattern.

What this PR adds

• ChatModelsLab class implementing the BaseChatModel protocol
• Uses ModelsLab's OpenAI-compatible chat API: https://modelslab.com/api/v6/llm
• Supports all browser-use features: streaming, structured output, tool calling
• Models available: llama-3-8b, llama-3-70b, mistral-7b, mixtral-8x7b, and more
• 5 unit tests (all passing)

Why ModelsLab

ModelsLab provides open-source LLMs (Llama, Mistral, Mixtral, etc.) at competitive pricing, including uncensored models for use cases requiring less restrictive content policies.

API docs

https://docs.modelslab.com

Usage

from browser_use.llm.modelslab.chat import ChatModelsLab
import os

llm = ChatModelsLab(
    model='llama-3-70b-chat',
    api_key=os.getenv('MODELSLAB_API_KEY'),
)

Testing

python -m pytest browser_use/llm/tests/test_modelslab_chat.py -v

Checklist

• Follows existing provider pattern (OpenRouter)
• Tests included (5 passing)
• No new dependencies
• API key via env var (MODELSLAB_API_KEY)
• Matches PRs merged previously (Mistral Add Mistral chat provider with schema sanitization and presets #3679, Vercel Add provider options to ChatVercel for enhanced model routing #3685)

Summary by cubic

Adds ModelsLab as an OpenAI-compatible LLM provider for browser-use to expand model options and support structured output, tools, and streaming. Follows the existing provider pattern with tests and an example.

• New Features

  • New ChatModelsLab provider using https://modelslab.com/api/v6/llm
  • Supports streaming, tool calling, and JSON-schema structured output
  • Follows OpenRouter provider pattern; no new dependencies
  • Models include llama-3-8b/70b, mistral-7b, mixtral-8x7b, and more
  • Added mocked tests and an example script

• Migration

  • Set MODELSLAB_API_KEY in the environment (optional: override base_url)

^{Written for commit 493f02d. Summary will update on new commits.}