Feat/add litellm provider

[RheagalFire]

Summary

Adds LiteLLM as an alternative LLM provider for MemoryOS, enabling access to 100+ LLM providers (OpenAI, Anthropic, Google Gemini, AWS Bedrock, Azure OpenAI, Mistral, Cohere, and more) through a unified interface.

Changes

• memoryos-pypi/utils.py - New LiteLLMClient class with chat_completion() method, specific exception handling for 7 error types (AuthenticationError, RateLimitError, APIConnectionError, Timeout, NotFoundError, BadRequestError, ContextWindowExceededError), drop_params=True for cross-provider compatibility
• memoryos-pypi/memoryos.py - Provider dispatch: provider="litellm" routes to LiteLLMClient instead of OpenAIClient
• memoryos-pypi/__init__.py - Export LiteLLMClient
• memoryos-pypi/requirements.txt - Added litellm>=1.80.0
• memoryos-pypi/tests/test_litellm.py - 12 comprehensive unit tests
• pytest.ini - Test configuration

Tests

Unit tests (12/12 passing):

test_returns_content PASSED
test_authentication_error PASSED
test_not_found_error PASSED
test_rate_limit_error PASSED
test_context_window_exceeded PASSED
test_none_content_returns_error PASSED
test_timeout_error PASSED
test_connection_error PASSED
test_bad_request_error PASSED
test_drop_params_true PASSED
test_api_key_forwarded PASSED
test_api_key_not_in_kwargs_when_none PASSED
12 passed

Risk / Compatibility

• Additive only. Existing OpenAIClient path completely untouched.
• litellm is an optional dependency - existing installs unaffected.
• Provider selection is explicit via provider="litellm" parameter.

Example usage

from memoryos import Memoryos

# Use LiteLLM with Anthropic Claude
memory = Memoryos(
    user_id="user_123",
    assistant_id="assistant_456",
    llm_model="anthropic/claude-sonnet-4-20250514",
    provider="litellm",
    api_key="sk-your-anthropic-key",
    storage_path="./memory_data"
)

# Use LiteLLM with AWS Bedrock
memory = Memoryos(
    user_id="user_123",
    assistant_id="assistant_456",
    llm_model="bedrock/anthropic.claude-3-haiku-20240307-v1:0",
    provider="litellm",
    storage_path="./memory_data"
    # Uses AWS credentials from environment (AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY)
)

# Use LiteLLM with Google Gemini
memory = Memoryos(
    user_id="user_123",
    assistant_id="assistant_456",
    llm_model="gemini/gemini-2.5-flash",
    provider="litellm",
    api_key="your-google-api-key",
    storage_path="./memory_data"
)

# Then use normally - all memory operations work the same regardless of provider
memory.add_interaction(
    user_input="I prefer dark mode and use vim",
    assistant_response="Got it! I'll remember your preferences."
)