AMITYBOT V1 – CRM-Grounded RAG Assistant

Overview

AMITYBOT V1 is a proof-of-concept AI assistant built around a CRM-grounded Retrieval-Augmented Generation (RAG) pipeline with a minimal yet functional Admin UI. The system enables non-technical users to import leads, manage knowledge base articles, trigger re-indexing, and query CRM-backed data through a conversational interface.

The project is designed for stakeholder demos, internal feedback, and iteration, with a clear separation between backend RAG logic and frontend admin tooling.

---

System Capabilities

Core RAG Backend (PoC)

• FastAPI-based backend with streaming chat responses (SSE)
• CRM-grounded answers using mock lead data
• Chroma vector database for knowledge storage
• Conversational RAG chain with tool calling
• Source-grounded responses with cited excerpts
• CLI demo client for quick testing

Admin UI (MVP)

• Next.js 14 admin dashboard
• Role-protected access (Admin / Editor)
• Global navigation sidebar:
  • Leads
  • Knowledge Base
  • Sync Logs
• Read-only leads table with server-side pagination
• Knowledge Base editor with Markdown support
• CSV-based bulk lead import
• Manual re-index trigger with toast feedback
• Sync logs visibility for re-embedding jobs

Governance & Polish

• Inline lead editing via drawer
• RBAC using NextAuth role claims
• Zod-based form validation
• Dark mode support
• Skeleton loaders and empty states
• Audit history per lead (mock timeline)
• End-to-end smoke tests and load testing
• Demo recording and setup documentation

---

Tech Stack

Backend

• FastAPI
• Python
• ChromaDB (Vector Store)
• RAG pipeline (Conversational Retrieval + Tools)
• Server-Sent Events (Streaming responses)

Frontend

• Next.js 14
• TypeScript
• Tailwind CSS
• shadcn/ui
• React Server Components
• DataGrid for tabular views

Tooling

• Git mono-repo
• CLI demo client
• Zod for validation
• NextAuth for RBAC
• Load testing (parallel query simulation)

---

Key API Endpoints

Chat & RAG

• POST /chat – CRM-grounded conversational endpoint (streaming)

Leads

• GET /crm/leads
• POST /crm/leads/bulk
• PATCH /crm/leads/:id

Knowledge Base

• POST /kb/upload
• POST /kb/reindex

---

Running the Project

Backend

cd backend
pip install -r requirements.txt
uvicorn app.main:app --reload

###Admin UI

cd admin-ui
npm install
npm run dev

###Admin UI runs on: http://localhost:3000