Q & A Based Chatbot on Custom Knowledgebase

Key Features

• CSV data handling with Q&A pairs and categories.
• Embedding storage and management using Azure Cognitive Search.
• Query processing with vector embeddings for relevant answer selection.
• Response generation using LLAMA3 70B and GPT-4o models.
• Performance testing framework for various question types (same, similar, hybrid, new).
• Dynamic knowledge base updates with auto-indexing in Azure Cognitive Search.
• User authentication and session management with JWT for secure access.
• Graph-based data representation with Neo4j for knowledge relationships.

Technologies Used

Challenges

Handling various types of questions with diverse queries, ensuring the accuracy of embedding storage, and maintaining efficient real-time knowledge base updates were significant challenges.

Solution

The system integrates embedding generation, Azure Cognitive Search for fast retrieval, and graph technology to visualize and query knowledge relationships, making it a powerful tool for querying and updating a knowledge base.

Results

The chatbot provides a seamless way to interact with and manage a custom knowledge base, improving knowledge retrieval and offering valuable insights from real-time interactions.