Operations

Prerequisites

• Python 3.12+
• DuckDB
• Ollama running locally, with phi4-mini and snowflake-arctic-embed2 pulled
• Node (for the 4-Lookup server and web client)

Pipeline scripts

scripts/
  extract_raw.py      <-- parse raw clipboard .txt dumps from 1-Input/
  extract_checked.py  <-- process the reviewed/checked subset
  embed.py             <-- generate vector embeddings via Ollama
  ingest.py             <-- load embedded entries into 3-Database/clipboard.duckdb

Run them in pipeline order (extract → embed → ingest) against files placed in 1-Input/.

Lookup UI

cd 4-Lookup/server
npm install
npm start    # Express + DuckDB API

cd 4-Lookup/web
npm install
npm run dev  # React + Vite client

Lookup API

Endpoint	Purpose
GET /api/stats	Summary stats
GET /api/categories	Available entry categories
GET /api/entry-types	Available entry types
GET /api/entries	List/filter entries
GET /api/entries/:entryId	Single entry detail
GET /api/embeddings/status	Embedding generation status

Configuration

Runtime configuration lives in config.toml at the repo root — model endpoints, paths, and pipeline options.