feat: Add ModelsLab model client (autogen-ext)

[adhikjoshi]

feat: Add ModelsLab chat completion client to autogen-ext

Summary

Adds ModelsLabChatCompletionClient to autogen-ext, giving AutoGen agents access to ModelsLab's uncensored Llama 3.1 8B and 70B models with 128K context windows via an OpenAI-compatible API.

What is ModelsLab?

ModelsLab is an AI API platform providing uncensored language models at competitive pricing — ideal for agents performing creative writing, research, and tasks where standard content filters are too limiting.

• Docs: https://docs.modelslab.com/uncensored-chat
• API key: https://modelslab.com/dashboard

Files

New files

File	Description
src/autogen_ext/models/modelslab/__init__.py	Package exports
src/autogen_ext/models/modelslab/_client.py	ModelsLabChatCompletionClient
tests/models/test_modelslab_client.py	15 test cases

Modified files (one-line patch)

src/autogen_ext/models/__init__.py:

+from autogen_ext.models.modelslab import ModelsLabChatCompletionClient

Implementation

Subclasses OpenAIChatCompletionClient with preset base_url, model_info, and MODELSLAB_API_KEY env var support:

class ModelsLabChatCompletionClient(OpenAIChatCompletionClient):
    def __init__(self, model="llama-3.1-8b-uncensored", api_key=None, ...):
        super().__init__(
            model=model,
            api_key=api_key or os.environ["MODELSLAB_API_KEY"],
            base_url="https://modelslab.com/uncensored-chat/v1",
            model_info=_MODELSLAB_MODEL_INFO[model],
            ...
        )

Pre-registered ModelInfo for both models (128K context, no vision, function calling enabled).

Usage

import asyncio
from autogen_agentchat.agents import AssistantAgent
from autogen_agentchat.ui import Console
from autogen_ext.models.modelslab import ModelsLabChatCompletionClient

model_client = ModelsLabChatCompletionClient(model="llama-3.1-8b-uncensored")

async def main():
    agent = AssistantAgent("assistant", model_client=model_client)
    await Console(agent.run_stream(task="Write a Python quicksort implementation."))

asyncio.run(main())

Models

Model	Context	Notes
llama-3.1-8b-uncensored	128K	Default — fast
llama-3.1-70b-uncensored	128K	Higher quality

Checklist

• Subclasses OpenAIChatCompletionClient — same pattern as Azure client
• Pre-registered ModelInfo for both models (context, capabilities)
• MODELSLAB_API_KEY env var support with clear error message
• Explicit api_key overrides env var
• Custom model_info override accepted
• 15 unit tests
• Docstring with usage example, model table, API links
• No new dependencies beyond autogen-ext[openai]

[adhikjoshi]

I agree to the CLA. @microsoft-github-policy-service agree

[adhikjoshi]

@microsoft-github-policy-service agree