Pre-trained video RAG system

Ask questions across your indexed video knowledge base.

A production-focused V1 experience for searching pre-transcribed, chunked, embedded, and timestamped videos. No uploads, no dataset expansion, just fast grounded answers with sources.

Loading index...
0Videos
0Chunks
V1No uploads
QueryRetrieveAnswer

Live assistant

Grounded answers with video timestamps

Interactive flow

How the RAG pipeline works

Tech stack explorer

Click a layer to inspect the system design

RAG pipeline visualizer

Click each node to see what happens at runtime

Select a node to inspect the runtime answer.

Video knowledge base

Explore indexed videos and transcript chunks

Open on YouTube
Select a source or video to load the YouTube timestamp here.

Live query playground

Compare retrieval context with the final answer

Prompt Context

Run a question to see the context sent to the model.

Response Metrics

Confidence - Latency - DeepSeek -

Disclaimer

Training source and usage notes

The videos trained for this project are from the CodeWithHarry YouTube Channel, and the videos are available on YouTube for free.

This project is an educational RAG showcase and is not affiliated with, sponsored by, or officially endorsed by CodeWithHarry or YouTube.

All answers should remain grounded in the retrieved transcript chunks. If a topic is not present in the indexed knowledge base, the assistant must say so clearly.

Video ownership, copyright, branding, and monetization rights remain with their original creators and platforms. Source links are preserved so users can watch the original videos.