AI - CHATBOT - Complete Guide - 1

1. Agent --> Agent will move towards your goal.
2. LLM --> Giving just the answer. Also AWS LLM ( Bedrock Service Eg., DMART own products ) which is provided by AWS which doesn't require any API keys from outside.
3. RAG --> Providing Additional Knowledge to LLM. Eg., Database, PDF , Doc , Text , website , etc
4. LangChain --> Used to connect the components of AI application
5. LangGraph --> It decides which tool needs to be executed based on the situation - Orchestration
6. MCP Server --> Contains Tools ( Postgres / Google Drive / etc )
7. MCP Client --> A plugin which is used to connect the MCP Server.
8. Memory --> To remember the previous information --> LangGraph , in the form of Dictionary.
9. Loop Engineering --> If the output is not correct , to execute the task again we need Loop which will be carried out by LangGraph.
10. FastAPI
11. Security --> Guardrails ( If the prompt contain any threat information or not )
12. Harness Engineering --> To validate or evaluate RAG.
13. React --> Chatbot Application
14. Go-Live or Production

Diagram

• Using FastAPI, Agent is called.
• Fast API is nothing but REST service.
• In this Agent who is the main person , i.e the Orchestration - LangGraph
• LangGraph will decide the Tool.
• So , MCP server will have the tools.
• MCP Server will connect to Postgres.
• Postgres is nothing but the Knowledge.
• Postgres then connects to RAG.
• To generate the answer , RAG calls the LLM.
• LLM will provide the answer to the ChatBot.

RAGAS

• Its an evaluation Framework.

Notes

• AWS server will provide inbuilt LLM , which is called Bedrock service.