Show HN: An AI zettelkasten that extracts ideas from articles, videos, and PDFs
10 days ago
- #knowledge-management
- #AI
- #zettelkasten
- Jargon is an AI-managed zettelkasten that summarizes sources, extracts ideas, links related concepts, and collapses duplicates.
- It uses semantic embeddings to surface connections across the library, creating a knowledge base of interlinked ideas.
- Sources like articles, PDFs, and YouTube videos are ingested, summarized, and key ideas are extracted into standalone insight cards.
- Embeddings power similarity search and automatic clustering, grouping similar articles and insights into themes.
- Each insight can spawn research threads that trigger web searches for related articles, which are automatically ingested.
- Jargon supports multiple LLM providers (OpenAI, Anthropic, Gemini, OpenRouter) and uses pgvector for vector similarity search in PostgreSQL.
- It can be run with Docker Compose, using a PostgreSQL database with pgvector and Redis for background jobs.
- Future improvements include replacing Exa with Brave search, parallel searches across multiple LLM-generated queries, and adding visual distinctions between articles and ideas.