How abhishekadbhute-create/yt_chatbot works
This is a Streamlit YouTube-video RAG chatbot. Users load a video by URL or provide transcript text; the app fetches or accepts transcript content, builds a FAISS index, then rewrites questions, retrieves relevant chunks, and generates English answers. Video and chat state are maintained in the Streamlit session. The implementation is concentrated in app.py; the graph separates its documented stages and integrations. Relationships are evidenced by app.py or the README; no other runtime source files are present.
Main entry points: app.py Streamlit App
Components 13
Transcript Ingestion
- Streamlit App
app.py - YouTube Transcripts
- Manual Transcript
RAG Retrieval
- Text Chunking
app.py - Text Embeddings
app.py - FAISS Vector Store
app.py
Answer Generation
- Query Rewriting
app.py - Similarity Retrieval
app.py - Answer Prompt
app.py - Hugging Face LLM
Session Management
- Chat History
app.py - Active Video State
app.py
Other components
- Video Viewer
Connections 15
- Video Viewer → Streamlit App: loads and asks
README.md - Streamlit App → YouTube Transcripts: fetches transcript
app.py - YouTube Transcripts → Streamlit App: returns captions
app.py - Streamlit App → Manual Transcript: accepts text
app.py - Streamlit App → Text Chunking: chunks transcript
app.py - Streamlit App → Text Embeddings: embeds chunks
app.py - Streamlit App → FAISS Vector Store: builds index
app.py - Streamlit App → Query Rewriting: rewrites question
app.py - Query Rewriting → Hugging Face LLM: invokes model
app.py - Streamlit App → Similarity Retrieval: searches index
app.py - Similarity Retrieval → FAISS Vector Store: retrieves chunks
app.py - Streamlit App → Answer Prompt: supplies context
app.py - Answer Prompt → Hugging Face LLM: generates answer
app.py - Streamlit App → Chat History: resets and appends
app.py - Streamlit App → Active Video State: sets active video
app.py