Knowledge-Retrieval-Chatbot

The knowledge Base Chatbot is a system that will allow the user to query and communicate with a database of uploaded documents. The chatbots are equipped with a repository of structured information, facts, and responses, allowing them to understand and respond to user queries effectively. The knowledge base serves as the bot's memory, enabling it to access and retrieve relevant information swiftly. These chatbots are particularly valuable in customer support, troubleshooting, and information retrieval scenarios, as they can deliver accurate and consistent responses. By continuously learning and updating their knowledge base, these chatbots improve their capabilities over time, making them a versatile tool for enhancing user experience and streamlining interactions in various domains.

Architecture

Features

• Upload documents that will get processed into a representation that the LLM powered chatbot can understand
• Interactive Chatbot interface that allow the user to communicate with their documents easily

Hardware Requirements:

• GPU -> 1-2 A100 with 80 GB VRAM
• CPU -> 128 GB RAM

Software Requirements:

• CUDA installed
• Docker Engine with docker compose for container management

Deployment

server
docker build . -t ghcr.io/rihal-om/knowledge-retrieval-chatbot/server:main
docker push ghcr.io/rihal-om/knowledge-retrieval-chatbot/server:main

web
docker build . -t ghcr.io/rihal-om/knowledge-retrieval-chatbot/web:main
docker push ghcr.io/rihal-om/knowledge-retrieval-chatbot/web:main

airflow
docker build . -t ghcr.io/rihal-om/knowledge-retrieval-chatbot/airflow:main
docker push ghcr.io/rihal-om/knowledge-retrieval-chatbot/airflow:main
docker save  ghcr.io/rihal-om/knowledge-retrieval-chatbot/airflow:main | gzip > krc-airflow.tar.gz


OCR
docker build . -t ghcr.io/rihal-om/knowledge-retrieval-chatbot/ocr:main
docker push ghcr.io/rihal-om/knowledge-retrieval-chatbot/ocr:main