How gurpreetkaurjethra/youtube-video-transcribe-summarizer-llm-app works
This app accepts a YouTube URL, downloads its audio, transcribes it, and summarizes the transcript with a local Llama model; the Streamlit interface also displays the video and result. summary.py provides a separate hard-coded command-line run of the same general workflow. The README describes the intended Haystack, Whisper, Llama 2, and Streamlit capabilities. The imported model_add module is absent from the supplied tree, so its implementation and wiring cannot be shown.
Main entry points: yt_summary.py Streamlit UI · summary.py Scripted run
Components 13
User interface
- Streamlit UI
yt_summary.py - Video and summary
yt_summary.py
Summarization workflow
- Download audio
yt_summary.py - Initialize model
yt_summary.py - Initialize prompt
yt_summary.py - Transcribe pipeline
yt_summary.py - Scripted run
summary.py
AI integrations
- Haystack pipeline
yt_summary.py - Whisper
- PromptNode
- Llama 2 GGUF
Other components
- User
- YouTube
Connections 17
- User → Streamlit UI: enters URL
- Streamlit UI → Download audio: calls
- Download audio → YouTube: downloads audio
- Streamlit UI → Initialize model: initializes
- Initialize model → Llama 2 GGUF: configures model
- Streamlit UI → Initialize prompt: initializes
- Initialize prompt → PromptNode: creates
- Streamlit UI → Transcribe pipeline: calls
- Transcribe pipeline → Haystack pipeline: builds and runs
- Haystack pipeline → Whisper: transcribes
- Haystack pipeline → PromptNode: dispatches transcript
- PromptNode → Llama 2 GGUF: generates summary
- Streamlit UI → Video and summary: displays video and result
- User → Scripted run: runs script
- Scripted run → YouTube: downloads audio
- Scripted run → Haystack pipeline: builds and runs
- Scripted run → Video and summary: prints summary