Project Overview
Kova is an intelligent web highlighter and research workspace built to curate noise and synthesize signal across long-form reading.
A Plasmo Chrome MV3 side-panel extension captures color-coded highlights directly on any webpage and scores page relevance against your research intent before you commit time to reading. Highlights sync into a Next.js workspace backed by FastAPI, PostgreSQL, and pgvector.
Each topic folder forms a strictly isolated RAG boundary so AI chat answers only from the passages you captured in that topic, with swappable OpenAI, Anthropic, and local Ollama providers.
Engineering decision
A Plasmo Chrome MV3 side-panel extension and Next.js workspace share a FastAPI + pgvector backend behind a Strategy pattern for OpenAI, Anthropic, and local Ollama models.
Outcome
In private beta with 100+ highlights processed across early users; combines pre-read page relevancy scoring (0–100), auto-categorisation, and strictly topic-isolated RAG chat with inline citations.
Key Features
- Pre-read page relevance scoring against active research topics
- Topic-scoped RAG chat grounded strictly in your saved highlights
- Chrome MV3 side-panel extension with color-coded web annotations
- Swappable OpenAI, Anthropic, and local Ollama providers via FastAPI + pgvector
