AI-Docs

AI-Docs is an advanced AI-powered chatbot that uses Retrieval-Augmented Generation (RAG) to answer user queries based on stored PDF documents. The chatbot integrates FastAPI for the backend, React (Vite) for the frontend, and is powered by Anthropic Claude.

Features

• ChatGPT-style conversation with past chat recall.
• PDF Processing: Extracts and indexes data from uploaded PDF files.
• Real-time message storage: Saves chat history using SQLite.
• RAG-based answering: Uses ChromaDB and embedding models for accurate, document-grounded responses.
• Multi-document retrieval: Queries across multiple PDFs simultaneously using MMR search.
• Dynamic chat list: Automatically updates the sidebar with active chats.
• Chat title editing: Rename conversations directly from the sidebar.
• Delete chat support: Deletes conversations dynamically.
• Configurable settings: Switch between Claude Haiku, Sonnet, and Opus models at runtime.
• Dark UI theme for a better user experience.
• Responsive Web Design: Mobile-friendly and accessible from all screen sizes.
• Dockerized Deployment: Easily run with Docker or Docker Compose.

Tech Stack

• Frontend: React 19 (Vite), TypeScript, TailwindCSS
• Backend: FastAPI, SQLite, ChromaDB, LangChain
• AI Model: Anthropic Claude (Haiku / Sonnet / Opus, configurable via settings)
• Embeddings: ChromaDB DefaultEmbeddingFunction (local, sentence-transformers — no API key required)

---

How It Works

User Question
     │
     ▼
LangChain RAG Chain
     │
     ├─► ChromaDB (MMR Retrieval)
     │        └─ Searches across all indexed PDF chunks
     │
     ├─► Retrieved Context (top-k relevant chunks)
     │
     └─► Anthropic Claude (claude-opus-4-6 / sonnet / haiku)
              └─ Generates grounded answer from context
                       │
                       ▼
               Answer + Chat History (SQLite)

1. PDFs are uploaded and chunked via pdfplumber + RecursiveCharacterTextSplitter
2. Chunks are embedded locally and stored in ChromaDB (persistent vector database)
3. On each query, MMR (Maximal Marginal Relevance) retrieval fetches the most relevant and diverse chunks
4. LangChain passes the retrieved context + conversation history to Claude
5. Claude generates a document-grounded answer — no hallucination from outside the provided PDFs

---

Project Structure

ai-docs/
├── backend/
│   ├── src/
│   │   ├── api.py            # FastAPI routes
│   │   ├── chat_manager.py   # Chat history management
│   │   ├── embedding.py      # ChromaDB embedding storage
│   │   ├── file_manager.py   # PDF upload/delete operations
│   │   ├── preprocessing.py  # PDF text extraction & chunking
│   │   ├── retrieval.py      # LangChain RAG chain + Claude integration
│   │   └── settings.py       # Runtime settings management
│   ├── Dockerfile
│   └── requirements.txt
├── frontend/
│   ├── src/
│   │   ├── components/       # React UI components
│   │   ├── hooks/            # Custom React hooks
│   │   ├── styles/           # Global styles
│   │   └── App.tsx
│   ├── Dockerfile
│   └── package.json
├── docker-compose.build.yml
├── docker-compose.image.yml
└── deploy.sh

---

Installation & Setup (Local)

Clone the Repository

git clone https://github.com/erenisci/ai-docs
cd ai-docs

Backend Setup

Navigate to the backend directory and install dependencies:

cd backend
python -m venv venv

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

pip install -r requirements.txt

Create a .env file in the backend directory:

ANTHROPIC_API_KEY=your_anthropic_api_key_here

Frontend Setup

Navigate to the frontend directory and install dependencies:

cd frontend
npm install

---

Run the Project

Start Backend Server

cd backend
uvicorn src.api:app --host 127.0.0.1 --port 8000 --reload

Start Frontend

cd frontend
npm run dev

The frontend will be available at http://localhost:5173

---

Docker Setup

Option 1: Build Locally with Docker Compose

docker-compose -f docker-compose.build.yml up --build -d

Access the App

• Frontend: http://localhost:5173
• Backend: http://localhost:8000

---

Option 2: Use Prebuilt Docker Images

a. Pull from Docker Hub

docker pull erenisci/ai-docs:backend
docker pull erenisci/ai-docs:frontend

b. Run Using Docker Compose (Recommended)

docker-compose -f docker-compose.image.yml up -d

Alternatively: Run Containers Manually

docker run -d -p 8000:8000 erenisci/ai-docs:backend
docker run -d -p 5173:3000 erenisci/ai-docs:frontend

---

API Endpoints

Method	Endpoint	Description
POST	/ask/	Sends a query to the chatbot.
GET	/get-chats/	Retrieves all stored chat sessions.
GET	/get-chat-history/{chat_id}	Fetches messages from a specific chat.
POST	/update-chat-title/{chat_id}/{title}	Updates the title of a specific chat.
DELETE	/delete-chat/{chat_id}	Deletes a specific chat.
GET	/list-pdfs/	Lists all stored PDFs.
POST	/upload-pdf/	Uploads a PDF file for processing.
POST	/process-pdfs/	Processes all uploaded PDFs.
DELETE	/delete-pdf/	Deletes a specific PDF file.
GET	/get-settings/	Returns the current AI settings.
POST	/update-settings/	Updates AI model and runtime settings.

---

Usage Guide

1. Start a new chat by clicking "New Chat".
2. Upload PDFs via the PDF manager to provide document-based answers.
3. Process PDFs to chunk and index them into ChromaDB.
4. Ask questions in the message box — Claude will answer using only the document content.
5. Switch models from the settings panel (Haiku for speed, Opus for depth).
6. Review past conversations in the sidebar.
7. Rename or delete chats as needed.

---

Contribution

Contributions are welcome! If you'd like to improve the project, feel free to open an issue or submit a pull request.

---

License

This project is licensed under the MIT License.