File size: 3,768 Bytes
a333428 e84d389 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 | ---
title: Video Rag
emoji: 🎥
colorFrom: blue
colorTo: indigo
sdk: docker
app_port: 7860
pinned: false
---
# Video RAG System Project
This FastAPI-based Video RAG (Retrieval-Augmented Generation) system provides endpoints to:
1. **Register & Authenticate** users
2. **Transcribe** YouTube or uploaded videos
3. **Query** the RAG system
4. **Manage** sessions (list, view, delete)
---
## Endpoint Flow
```mermaid
graph TD
A[POST /register] --> B[POST /token]
B --> C[POST /transcribe]
B --> D[POST /upload]
C --> E[Start RAG session]
D --> E
E --> F[POST /query]
E --> G[GET /sessions]
G --> H[GET /sessions/{session_id}]
H --> F
G --> I[DELETE /sessions/{session_id}]
```
1. **User Registration & Login**
- **POST /register**: Create a new user.
- **POST /token**: Obtain JWT access token.
2. **Video Transcription**
- **POST /transcribe** (YouTube URL): Transcribe via Google GenAI → split & store chunks → initialize chat history → return `session_id`.
- **POST /upload** (Multipart Form Video): Upload & transcribe file → split & store chunks → initialize chat history → return `session_id`.
3. **Query RAG System**
- **POST /query** with `{ session_id, query }`:
• Rebuild FAISS retriever from MongoDB chunks
• Invoke ConversationalRetrievalChain
• Append messages to chat history
• Return `{ answer, session_id, source_documents }`
4. **Session Management**
- **GET /sessions**: List all sessions for current user.
- **GET /sessions/{session_id}**: Get full transcription & Q&A history.
- **DELETE /sessions/{session_id}**: Remove metadata, chunks, chat history, and video files.
---
## README.md
```markdown
# Video RAG System
## Overview
A FastAPI application that:
- Authenticates users (JWT)
- Transcribes videos (YouTube or upload) via Google GenAI
- Stores transcription chunks in MongoDB
- Builds a FAISS retriever on demand
- Provides a conversational retrieval endpoint
- Manages sessions and associated data
## API Endpoints
| Method | Path | Auth Required | Description |
|--------|----------------------------|---------------|-----------------------------------------------|
| POST | /register | No | Create a new user |
| POST | /token | No | Login and return JWT token |
| POST | /transcribe | Yes | Transcribe YouTube video and init session |
| POST | /upload | Yes | Upload & transcribe video file |
| POST | /query | Yes | Run Q&A against a session |
| GET | /sessions | Yes | List all user sessions |
| GET | /sessions/{session_id} | Yes | Get session transcription & chat history |
| DELETE | /sessions/{session_id} | Yes | Delete session & all associated data |
## Usage
1. Clone repo & install dependencies:
```bash
pip install -r requirements.txt
```
2. Create `.env` with your credentials (MongoDB, JWT secret, API keys).
3. Run the app:
```bash
uvicorn app.main:app --reload
```
4. Interact via HTTP clients (curl, Postman) following the flow above.
## Folder Structure
```
rag_system/
├── app/
│ ├── main.py
│ ├── config.py
│ ├── dependencies.py
│ ├── models/
│ ├── db/
│ ├── services/
│ ├── routes/
│ └── utils/
├── temp_videos/
├── .env
├── requirements.txt
└── README.md
```
```
```
|