All Projects
In Progress
Tablo News
An AI powered local news search and archive interface for Seattle area TV broadcasts, transcribed with Whisper, indexed with ChromaDB, and searchable via natural language RAG queries.
Overview
An automated pipeline that records Seattle local news from a Tablo DVR, transcribes with WhisperX, identifies speakers via voice fingerprinting, summarizes with Ollama, and indexes in ChromaDB for semantic search.
Frontend
Next.js 16 · React 19 · TypeScript · Cloudinary
- Featured story display with AI-generated summaries
- Paginated broadcast archive grouped by date
- Natural language search via RAG-powered semantic search
- Speaker filtering: query by specific anchors/reporters
- Full transcript viewer with timestamps and speaker ID
- Broadcast screenshots with AI-scored captions
- Speaker analytics: speaking time and segment counts
- Coverage stats dashboard
Backend Pipeline
Python · WhisperX · ChromaDB · Ollama · SQLite / Turso
- WhisperX transcription pipeline for broadcast audio
- ChromaDB vector database for semantic indexing
- Ollama for local LLM summarization
- Voice fingerprint database for speaker identification
- SQLite (dev) / Turso (production) for structured data
Coverage: KING 5 News · KOMO 4 News · KIRO 7 News · Local Programming
Built With
Next.jsPythonWhisperXChromaDBOllamaTurso