20260720 Corpus Explorer S1 Etl

loom-builderpromptmarkdown
Path: loom-builder/prompts/20260720-corpus-explorer-s1-etl.mdRelative: prompts/20260720-corpus-explorer-s1-etl.mdSubproject: loom-builderCategory: dispatch-prompts
Lines: 134Bytes: 7,649Imports: 0Exports: 0
loom-builderpromptmarkdownprompts

Content Preview

  1. # Corpus Explorer — S1 Data Plane (ETL → DuckDB → Parquet)
  3. ## User intent (restated)
  5. Mike wants his Claude conversation history (917 exported conversations, two
  6. accounts, already parsed to staging JSONL) turned into a queryable analytics
  7. dataset: a DuckDB database with proper tables, enrichment (categories/tags/
  8. languages), and a Parquet export the Astro site (S2, not your job) will read.
  9. **Success metric: all 12 spec §15 assertions pass green against the built
  10. DuckDB file, on the FULL corpus