feat: Add multi-provider LLM support (Fixes #1)

[shanto12]

Overview

This PR implements a provider-agnostic LLM client interface that enables seamless switching between different LLM providers without code changes, addressing Issue #1.

Changes

1. Provider-Agnostic Interface (src/llm/client.py)

• Created LLMClient abstract base class with methods:
  • chat() - for chat completions
  • embed() - for embeddings
  • tts() - for text-to-speech (optional)
• Defined data classes for ChatMessage, ChatResponse, EmbeddingResponse, and TTSResponse

2. Provider Adapters (src/llm/adapters.py)

Implemented concrete adapters for three providers:

• OpenAICompatibleAdapter: Works with OpenAI API and any OpenAI-compatible services (xAI, DeepSeek, etc.)
  • Supports custom base_url configuration
  • Implements chat, embed, and tts methods
• GroqAdapter: Fast inference with Groq API
  • Implements chat method
  • Embedding not supported (raises NotImplementedError)
• OllamaAdapter: Local model inference
  • Implements chat and embed methods
  • Default base URL: http://localhost:11434

3. Factory Pattern (src/llm/factory.py)

• LLMClientFactory class for creating provider-specific clients
• create_client() - Create client with explicit configuration
• create_from_env() - Create client from environment variables
• Environment variable support:
  • LLM_PROVIDER - Provider name (openai, groq, ollama)
  • LLM_API_KEY - API key
  • LLM_BASE_URL - Custom base URL (optional)
  • LLM_MODEL - Default model

4. Tests (tests/test_llm_client.py)

• Factory tests for all three providers
• Chat completion tests with mocking
• Error handling tests
• Provider validation tests

5. Documentation (README.md)

• Added comprehensive multi-provider LLM support section
• Configuration examples for all three providers
• Usage examples showing how to switch providers
• Clear instructions for environment variable setup

How to Switch Providers Without Code Changes

Simply update environment variables:

# Switch to Groq
export LLM_PROVIDER=groq
export LLM_API_KEY=your_groq_key

# Switch to Ollama (local)
export LLM_PROVIDER=ollama
export LLM_BASE_URL=http://localhost:11434  # Optional

# Switch to OpenAI
export LLM_PROVIDER=openai
export LLM_API_KEY=your_openai_key

Then use the factory method:

from src.llm import LLMClientFactory

client = LLMClientFactory.create_from_env()

The client will automatically use the configured provider!

Testing

Run tests with:

pytest tests/test_llm_client.py

Related Issue

Fixes #1 - Multi-provider LLM support implementation

[vercel]

@shanto12 is attempting to deploy a commit to the md777 Team on Vercel.

A member of the Team first needs to authorize it.