--collection¶

Why we added this flag: Chroma stores documents in named collections; idxr needs to know which collection to upsert into (or prefix when partitioning).

What it does¶

• Identifies the target collection.
• Required for persistent client runs; optional for partition-aware Cloud runs where each partition uses its own collection name derived from metadata.
• Used when generating resume state filenames (<collection>_resume_state.json).

Typical usage¶

idxr vectorize index \
  --model "$IDXR_MODEL" \
  --partition-manifest workdir/partitions/manifest.json \
  --partition-out-dir workdir/chroma_partitions \
  --client-type cloud \
  --collection ecc-std

Tips¶

• Namespace collections by environment (e.g., kb-prod, kb-staging) to simplify cross-env testing.
• When using partition-specific collections, include the base collection name in observability dashboards so metrics aggregate correctly.