--config

Why we added this flag: when you index straight from CSVs (without using prepare-datasets), you still need a declarative mapping—this flag points vectorize to that JSON config.

What it does

  • Path to the vectorize config file (see Config Reference).
  • Mutually exclusive with --partition-manifest. Choose one source of truth.
  • Required for direct CSV indexing; optional otherwise.

Typical usage

idxr vectorize index \
  --model "$IDXR_MODEL" \
  --config configs/vectorize_contracts.json \
  --collection ecc-std \
  --persist-dir workdir/chroma

Tips

  • Prefer manifest-driven indexing in production; use config-driven runs for quick experiments or backfills.
  • Keep vectorize and prepare-datasets configs in sync so column renames stay consistent.