🧠 AI Knowledge Base

A full-stack RAG (Retrieval-Augmented Generation) document assistant powered by FastAPI, React/Vite, Qdrant, and Google Gemini.

Upload PDF documents and ask natural-language questions — the AI retrieves relevant context from your documents and generates grounded answers.

---

🏗️ Architecture

React (Vite) ←→ FastAPI ←→ Qdrant (vector DB)
                   ↕
             Google Gemini API
             (embeddings + chat)

Pipeline:

1. PDF → text extraction (PyMuPDF)
2. Text → overlapping chunks
3. Chunks → Gemini embeddings → Qdrant
4. Query → embed → Qdrant semantic search → top-k chunks
5. Chunks + query → Gemini LLM → grounded answer

---

📁 Project Structure

project/
├── backend/
│   ├── main.py              # FastAPI app entry point
│   ├── config.py            # Settings (pydantic-settings)
│   ├── requirements.txt
│   ├── Dockerfile
│   ├── .env.example
│   ├── routers/
│   │   ├── documents.py     # /upload-pdf, /documents, /document/{id}
│   │   └── chat.py          # /ask
│   └── services/
│       ├── pdf_service.py   # Extraction + chunking
│       ├── embedding_service.py  # Gemini embeddings
│       ├── vector_service.py     # Qdrant CRUD
│       └── rag_service.py        # RAG pipeline
├── frontend/
│   ├── src/
│   │   ├── App.jsx
│   │   ├── index.css
│   │   ├── components/
│   │   │   ├── ChatPanel.jsx
│   │   │   ├── UploadPanel.jsx
│   │   │   ├── DocumentList.jsx
│   │   │   └── Toast.jsx
│   │   └── services/api.js
│   ├── package.json
│   └── vite.config.js
└── docker-compose.yml

---

⚙️ Prerequisites

• Python 3.11+
• Node.js 18+
• Docker Desktop (for Qdrant)
• Google Gemini API key → Get one here

---

🚀 Setup & Running

Option A — Docker Compose (Recommended)

1. Clone / open the project folder.

2. Create backend .env:

   cp backend/.env.example backend/.env
   # Edit backend/.env and set GEMINI_API_KEY=your_key_here

3. Start backend + Qdrant:

   docker-compose up --build

   • API: http://localhost:8000
   • API docs: http://localhost:8000/docs
   • Qdrant dashboard: http://localhost:6333/dashboard

4. Start the frontend:

   cd frontend
   npm install
   npm run dev

   • App: http://localhost:5173

---

Option B — Manual / Local Dev

Backend

# 1. Start Qdrant via Docker
docker run -p 6333:6333 -p 6334:6334 qdrant/qdrant

# 2. Set up Python env
cd backend
python -m venv venv

# Windows
venv\Scripts\activate
# macOS/Linux
source venv/bin/activate

pip install -r requirements.txt

# 3. Create .env
copy .env.example .env
# Edit .env and add your GEMINI_API_KEY

# 4. Run backend
uvicorn main:app --reload --port 8000

Frontend

cd frontend
npm install
npm run dev

---

🔑 Environment Variables

Variable	Default	Description
GEMINI_API_KEY	(required)	Google Gemini API key
QDRANT_HOST	localhost	Qdrant host
QDRANT_PORT	6333	Qdrant port
COLLECTION_NAME	knowledge_base	Qdrant collection name
CHUNK_SIZE	800	Characters per chunk
CHUNK_OVERLAP	150	Character overlap between chunks
TOP_K	5	Number of chunks to retrieve per query
EMBEDDING_MODEL	models/text-embedding-004	Gemini embedding model
CHAT_MODEL	gemini-1.5-flash	Gemini chat model

---

📡 API Endpoints

Method	Endpoint	Description
POST	/upload-pdf	Upload and process a PDF
POST	/ask	Ask a question (RAG)
GET	/documents	List all uploaded documents
DELETE	/document/{doc_id}	Delete a document's vectors
GET	/health	Health check

Interactive API docs: http://localhost:8000/docs

---

🧪 Usage

1. Open http://localhost:5173
2. Drag & drop a PDF onto the upload zone (or click to browse)
3. Click Upload & Process — the PDF is chunked, embedded, stored in Qdrant
4. Type your question in the chat box and press Enter
5. The AI retrieves relevant chunks and returns a grounded answer with source citations
6. Manage your documents in the sidebar — delete when no longer needed

---

🐳 Docker Notes

• Qdrant data is persisted in a Docker volume (qdrant_storage) — your documents survive restarts.
• When using Docker Compose, QDRANT_HOST is automatically set to qdrant (the service name). In local dev, it remains localhost.