GitHub: 0xarchit/ChatDoc
GitHub: 0xarchit/DocumentChat-Simple-Rag-VectorDB-Project

• ChatDoc
  • Overview
  • Preview
  • Architecture
  • Features
  • Tech Stack
  • Getting Started
  • API Reference
  • Future Goals
• AI/ML/GenAI Mentor Chatbot
  • Running in Google Colab
  • Preview
  • Basic Working
  • Usage
  • Clearing Data

GitHub: 0xarchit/ChatDoc
Live Demo: https://chatdoc.0xarchit.is-a.dev

A unified retrieval-augmented generation (RAG) document API and web interface, powered by FastAPI, React, Vite, Milvus, and MistralAI.

ChatDoc is a web application enabling users to upload documents, extract and chunk text, store embeddings in Milvus, and query with state-of-the-art LLMs. It provides both a REST API and a web-based interface for seamless integration.

flowchart TB
  subgraph Frontend
    UI[React & Vite] -->|REST API| API(FastAPI)
  end
  subgraph Backend
    API --> Extract[Text Extraction]
    Extract --> Chunk[Text Chunking]
    Chunk --> Embed[MistralAI Embedding]
    Embed --> Store[Milvus Vector Store]
    API --> Retrieve[Retrieval]
    Retrieve --> LLM[ChatOpenAI]
    LLM --> Store
  end
  Store -.->|Query Results| API

• Upload PDF, TXT, CSV, XLSX, PPTX, DOCX files via API or web form
• Automatic text extraction and chunking (500 tokens, 50 overlap)
• Embedding with MistralAI Embeddings & storage in Milvus (Zilliz)
• Retrieval and response generation via OpenAI-compatible LLM
• Real-time, responsive React UI with upload, history, and settings
• Per-request overrides for API keys, endpoints, and collections
• Admin endpoints for deleting uploads or clearing the vector store

• Backend: FastAPI, Python, PyPDF2, python-pptx, python-docx, Pandas, Milvus
• Frontend: React, Vite, TypeScript, Tailwind CSS
• Embeddings: MistralAI
• Vector Database: Milvus (Zilliz Cloud)
• LLM: OpenAI-compatible ChatOpenAI via LangChain

• Node.js >= 16 and npm/yarn
• Python >= 3.9
• Docker (optional)
• Milvus or Zilliz Cloud credentials
• MistralAI & OpenAI API keys

cd Backend
copy .env.example .env
# Edit .env and set:
# MISTRAL_API_KEY, ZILLIZ_URI, ZILLIZ_TOKEN, HF_TOKEN (optional), COLLECTION_NAME
pip install -r requirements.txt
uvicorn main:app --reload

cd Frontend
npm install
npm run dev

# Build and run backend container
docker build -t chatdocapi-backend .
docker run --rm -p 8080:8080 \
  -e MISTRAL_API_KEY=$env:MISTRAL_API_KEY \
  -e ZILLIZ_URI=$env:ZILLIZ_URI \
  -e ZILLIZ_TOKEN=$env:ZILLIZ_TOKEN \
  -e ZILLIZ_COLLECTION_NAME=$env:ZILLIZ_COLLECTION_NAME \
  chatdocapi-backend

• Description: Upload document and store embeddings.
• Content-Type: multipart/form-data
• Fields:
  • file (required)
  • mistral_api_key, zilliz_uri, zilliz_token, collection_name (optional)
• Responses:
  • 200: { "upload_id": "<uuid>" }
  • 400: errors (no file, extraction failure)
  • 413: file too large

• Description: Retrieve and answer based on stored chunks.
• Content-Type: application/json

• Body:

  {
    "question": "string",
    "upload_id": "string",
    ...overrides
  }

• Responses:
  • 200: { "answer": "<generated answer>" }
  • 400: invalid body
  • 500: generation error

• Description: Remove all vectors for a given upload.
• Params: upload_id path, overrides as query params
• Response: { "status": "deleted" }

• Description: Clear entire vector store.
• Query: password (native admin) or per-request overrides
• Response: { "status": "all_deleted" }

• Streaming responses from the model to improve perceived latency and UX.
• Better OCR and robust file parsing for scanned PDFs and more file formats.
• Pluggable support for multiple vector stores (Milvus, FAISS, Pinecone, etc.).
• Increase upload and context limits (larger files, fewer artificial word/chunk restrictions).
• Personalization with login/signup, per-user profiles, metadata, and tags.
• Expand supported AI models/providers and allow per-request model selection.

GitHub: 0xarchit/DocumentChat-Simple-Rag-VectorDB-Project

This vibe coded project provides an AI/ML/GenAI Mentor Chatbot built with Gradio, LangChain, ChromaDB, and Google Gemini. You can run the notebook in Google Colab to interact with the chatbot.

1. Open the notebook:

   • Import the notebook into your Google Drive or GitHub repository.
   • Open Google Colab and select File > Open notebook.
   • Under the Google Drive, GitHub, or Upload tab, locate your notebook (main.ipynb) and open it.

2. Install dependencies:

   • Colab will automatically install required packages when running the first cell.

3. Run all cells:

   • In Google Colab, select Runtime > Run all to execute every cell sequentially.
   • This will set up the environment, process documents, initialize the chatbot, and launch the Gradio interface.

Alternatively, you can use the public Colab link:

Open in Google Colab

This project combines several modern AI tools to create an interactive mentor chatbot:

• Gradio UI: Provides a simple, user-friendly web interface for uploading documents and chatting with the AI mentor.
• ChromaDB (Vector Database): Stores document chunks as vector embeddings, enabling fast and relevant retrieval of information based on your questions.
• all-MiniLM-L6-v2 Embedding Model: Converts text from your documents and queries into numerical vectors, allowing semantic search and matching.
• Google Gemini (Generative AI): Generates detailed, structured answers using both retrieved document context and its own AI/ML/GenAI knowledge.

How it works:

1. Upload your PDF, PPT, or TXT files using the Gradio interface.
2. The files are split into chunks and embedded using the all-MiniLM-L6-v2 model.
3. ChromaDB stores these embeddings for efficient retrieval.
4. When you ask a question, the system finds the most relevant document chunks and sends them, along with your question, to Gemini.
5. Gemini generates a comprehensive answer, referencing your materials and providing real-world examples.

• Upload your PDF, PPT, or TXT files using the notebook’s upload section.
• After processing, ask questions about AI, ML, or GenAI topics.
• The chatbot will provide structured, detailed responses with real-world examples based on provided material only.

• To clear all uploaded documents and chat history, run the Clear All Data button in the interface.