🤖 LLM Provider Module Documentation

Location: app/core/providers.py This module provides a standardized and extensible interface for interacting with multiple Large Language Model (LLM) APIs. It enables pluggable support for models like DeepSeek and Gemini using a unified abstraction layer.

1. ⚙️ Configuration

The module behavior is controlled by environment variables, primarily for authentication and model selection.

📌 Environment Variables

💡 Tip: Set these variables in your .env file or environment before running the application.

2. 🧱 Core Components

The system uses a provider interface pattern to support multiple LLMs with a consistent API.

🧩 LLMProvider (Abstract Base Class)

• Base class for all LLM providers.

• Defines the required method:

  def generate_response(self, prompt: str) -> str

• All concrete providers must implement this method.

🧠 DeepSeekProvider

• Implements the LLMProvider interface.
• Uses:
  • DEEPSEEK_API_KEY
  • DEEPSEEK_MODEL_NAME
• Handles API calls to DeepSeek's LLM service.

🌟 GeminiProvider

• Implements the LLMProvider interface.
• Uses:
  • GEMINI_API_KEY
  • GEMINI_MODEL_NAME
• Handles API calls to Google Gemini's LLM service.

🏭 get_llm_provider(model_name: str) -> LLMProvider

• Factory function for obtaining a provider instance.

Usage:

provider = get_llm_provider("deepseek")
response = provider.generate_response("Tell me a joke.")

Supported values for model_name:

• "deepseek"
• "gemini"

Raises:

• ValueError if an unsupported model_name is passed.

3. 🚀 Usage Example

Use the get_llm_provider() factory function to get a provider instance. Then call generate_response() on it:

from app.core.providers import get_llm_provider

# Get provider
llm = get_llm_provider("gemini")

# Generate response
response = llm.generate_response("What is the capital of France?")
print(response)

This design allows you to easily swap or extend LLMs (e.g., add OpenAI, Anthropic) by simply implementing a new provider class.