Adaptive Chatbot System

This project focuses on building an intelligent chatbot system that can autonomously decide which tools to use based on user queries. The system leverages vectordb to classify intent and dynamically selects the appropriate tools or APIs to provide accurate responses. Key features include: - Intent recognition (Synthetically generated queries for transactional_queries and service_queries) - Trained a model with this hand curated dataset to generate more similar and diverse queries - Embedded the quries in milvusDB - A similarity search is performed to find whether the query is transactional or service based - Using milvus saves costs for an LLM call in prod - Backend and frotend deployed on docker - LLM serving via vLLM on WSL - model: phi3:7b - Whenever user clicks on the chatbot button, a JWT session is created with a 10 minute expiry - This makes the overall pipeline secure