feat: add LiteLLM as AI gateway provider

[RheagalFire]

Summary

Adds LiteLLM as a new LLM provider, giving MIRIX users access to 100+ LLM providers (OpenAI, Anthropic, Azure, Bedrock, Vertex AI, Groq, Ollama, etc.) through a single unified interface via the LiteLLM Python SDK.

Changes

• mirix/llm_api/litellm_client.py -- new LiteLLMClient extending LLMClientBase, uses litellm.acompletion() with drop_params=True
• mirix/llm_api/llm_client.py -- added "litellm" case to factory
• mirix/schemas/llm_config.py -- added "litellm" to model_endpoint_type Literal
• pyproject.toml -- added litellm>=1.80.0,<1.87.0 as optional dependency

Tests

Live E2E against Azure Foundry (Anthropic Claude Sonnet 4.6):

config = LLMConfig(
    model='anthropic/claude-sonnet-4-6',
    model_endpoint_type='litellm',
    model_endpoint='https://azure-endpoint.services.ai.azure.com/anthropic',
    context_window=200000,
    api_key='...',
)
client = LLMClient.create(config)
msgs = [Message(role='user', content=[TextContent(type='text', text='What is 2+2? Reply with just the number.')])]
resp = await client.send_llm_request(msgs)

Client type: LiteLLMClient
Response: 4
Model: claude-sonnet-4-6
Tokens: 20 in, 5 out

Lint: make format passes clean.

Example usage

from mirix.schemas.llm_config import LLMConfig

# Use any provider supported by LiteLLM (100+)
# Model format: provider/model-name
config = LLMConfig(
    model="anthropic/claude-haiku-4-5",
    model_endpoint_type="litellm",
    context_window=200000,
)

# Provider API keys are read from environment variables
# e.g. ANTHROPIC_API_KEY, OPENAI_API_KEY, etc.
# Or pass api_key directly in the config

Risk / Compatibility

• Additive only -- no existing providers modified
• litellm is an optional dependency (pip install mirix[litellm])
• Lazy-imported inside request() so base install is unaffected

[CLAassistant]

All committers have signed the CLA.

[RheagalFire]

cc @LiaoJianhe