How vedantbhatotia/clip-plot2 works
Project type and purpose
• This is a full-stack application that lets users upload or process audio (and possibly video) files, transcribe them, generate embeddings, store and search the results, and display them in a web UI.
• The backend is a Python-based service (likely FastAPI or Flask) orchestrating audio transcription, segmentation, embedding, vector‐search (via Chroma), and RAG. The frontend is a Next.js/React dashboard in TypeScript.
File structure
Top-level:
• docker-compose.yml – orchestrates containers (backend, maybe Chromadb, frontend)
• extract.py – CLI script for batch extraction/transcription
• frontend/ – Next.js web app
• backend/ – Python service with Dockerfile, .env, requirements, services/, data/
• data/ – Chroma DB production files (vector store)
backend/services:
• transcription_service.py – invokes Whisper or another ASR model
• segment_processor_service.py – splits audio into chunks
• embedding_service.py – calls an embedding model
• database_service.py – wraps vector store calls to ChromaDB
• search_service.py & rag_service.py – perform vector search and RAG answer generation
• media_processor.py & clip_builder_service.py – handle media I/O and clip creation
frontend/app:
• page.tsx, dashboard/page.tsx – UI entry points
• components/ui – buttons, cards, etc.
• lib/utils.ts – shared utilities
README
• The README is empty, so we rely solely on the file layout and naming conventions to infer architecture.
Frontend → Backend.API → TranscriptionService → External ASR → TranscriptionService → DatabaseService → ChromaDB
• Blue for Frontend
• Green for Backend services
• Orange for Data store
• Gray for External APIs
• Rectangles for processes/services
• Clouds for external systems