Building a YouTube Transcript Chatbot with RAG

Grounded / Real Inflated / Uruttu

98% real

2% uruttu

article Original Content

🚀 Excited to share one of my recent AI projects — YouTube Transcript RAG Chatbot!
Over the past few weeks, I built a full-stack Retrieval-Augmented Generation (RAG) application that enables users to chat with the content of any YouTube video using natural language.
🔹 What it does
- Extracts YouTube transcripts automatically
- Performs recursive text chunking
- Generates semantic embeddings
- Creates a dedicated FAISS vector database for each YouTube video
- Retrieves the most relevant transcript chunks
- Uses Ollama (Llama 3.1 8B) to generate context-aware responses based only on the retrieved transcript
🔹 Engineering Highlights
- FastAPI backend with REST APIs
- SQLite + SQLAlchemy for user and conversation management
- LangChain-powered modular RAG pipeline
- Separate FAISS vector indexes for every YouTube video using the Video ID
- Dynamic loading of vector databases to avoid duplicate embedding generation
- HTML, CSS, and JavaScript frontend
- Multi-user architecture with isolated conversations
🛠 Tech Stack
Python • FastAPI • LangChain • Ollama • Llama 3.1 8B • Hugging Face Embeddings • FAISS • SQLAlchemy • SQLite • HTML • CSS • JavaScript
This project gave me practical experience in building production-style AI systems, including Retrieval-Augmented Generation (RAG), semantic search, vector databases, local LLM deployment, backend API development, and scalable application architecture.
GitHub link :- https://lnkd.in/dpuaQPnq
I'd love to hear your thoughts or suggestions!

verified Validated Content

Assuming this is your own project and the implementation matches what you've described, the post is technically accurate.

Fact check

Project description

"YouTube Transcript RAG Chatbot"

✅ Accurate.
This is a standard RAG application pattern.

Features

"Extracts YouTube transcripts automatically."

✅ Accurate, assuming you're using a transcript API or library.

"Performs recursive text chunking."

✅ Accurate.
LangChain's RecursiveCharacterTextSplitter is commonly used.

"Generates semantic embeddings."

✅ Accurate.

"Creates a dedicated FAISS vector database for each YouTube video."

✅ Accurate.
This is a valid architecture and helps isolate retrieval per video.

"Retrieves the most relevant transcript chunks."

✅ Accurate.

"Uses Ollama (Llama 3.1 8B) to generate context-aware responses based only on the retrieved transcript."

✅ Accurate, provided your prompt instructs the model to answer only from retrieved context.

Engineering highlights

"FastAPI backend with REST APIs."

✅ Accurate.

"SQLite + SQLAlchemy."

✅ Accurate.

"LangChain-powered modular RAG pipeline."

✅ Accurate.

"Separate FAISS vector indexes for every YouTube video using the Video ID."

✅ Accurate.
This is a sensible design choice.

"Dynamic loading of vector databases to avoid duplicate embedding generation."

✅ Accurate.
Reusing previously created indexes is a common optimization.

"HTML, CSS, JavaScript frontend."

✅ Accurate.

"Multi-user architecture with isolated conversations."

✅ Accurate, assuming conversations are scoped per user in your database.

Tech stack

Everything listed is consistent with the described architecture.

Final paragraph

"This project gave me practical experience in building production-style AI systems..."

✅ Accurate.
That's a personal statement and a fair characterization.

One small wording suggestion:

Instead of "production-style AI systems," consider "production-oriented AI systems." It sounds slightly more precise, especially if the app hasn't been deployed in production.

GitHub link

🟡 I can't verify the LinkedIn short link (lnkd.in) or the repository contents from the text alone. If the repository matches the described implementation, then this is fine.

Overall verdict

Accuracy: 9.8/10

The architecture is technically sound and reflects common best practices for a transcript-based RAG application. The only thing I can't independently verify is that the GitHub repository implements exactly what you've described.

Real vs Fluff

🟢 98% Real | 🟡 2% Fluff

Real (98%)

RAG architecture.
FAISS indexing.
Recursive chunking.
Embeddings.
Ollama.
FastAPI.
LangChain.
SQLite.
Multi-user design.
Dynamic vector loading.

Fluff (2%)

"Production-style AI systems" is slightly promotional but entirely reasonable for a portfolio project.

Overall, this is a strong, technically grounded project showcase.