LLM Agent Routing System

This project is an intelligent routing system for user queries, using the LangChain framework and Groq's Gemma2-9b-It model. It allows queries to be directed to the most relevant data source either arxiv_search (for AI research papers), Wikipedia (for human-related information), or an LLM (for general queries).

Features

• Context-Aware Routing: Routes user queries based on the context to either arxiv, Wikipedia, or an LLM.
• Flexible Query Handling: Supports different types of user queries related to AI research, human information, or general knowledge.
• Powered by LLM: Utilizes Groq's Gemma2-9b-It model for intelligent query understanding and routing.

Project Components

1. LangChain: Used for building the agent that handles the query routing process.
2. Groq's Gemma2-9b-It Model: The large language model used to interpret user queries and generate structured output.
3. arxiv_search: Searches for AI-related research papers.
4. Wikipedia Search: Handles queries related to human knowledge.
5. LLM Search: A fallback option for general queries that dont fit into the other categories.

Installation

1. Clone the repository:

   git clone https://github.com/MadhanMohanReddy2301/SmartChainAgents.git
   

2. Navigate to the project directory:

   cd SmartChainAgents
   

3. Install the required dependencies:

   pip install -r requirements.txt
   

4. Set up your environment variables by adding your Groq API key:

   export GROQ_API_KEY=your_groq_api_key
   

Usage

1. Initialize the LLM router system by running the script:

   python graph.py
   

2. Test the routing functionality with some example queries:

   question_router.invoke({"question": "Who is Shahrukh Khan?"})
   question_router.invoke({"question": "What are the types of agent memory?"})
   

3. The system will route the query to the appropriate source and return the result accordingly.

Example

# Example query for Wikipedia search
{
  "datasource": "wiki_search"
}

# Example query for arxiv search
{
  "datasource": "arxiv_search"
}