Skip to Content
We are live but in Staging 🎉
CLI GuideVector and Search Commands

Vector and Search Commands

This domain covers vector engine setup, collection management, and vector search.

Vector Store Commands

Command group: dodil k3 vector-store

Persistent flag:

  • --bucket, -b
SubcommandArgsImportant flags
createnone--mode, -m (auto/external)
getnone--bucket
deletenone--bucket
list-instancesnonenone

Example:

dodil k3 vector-store create --bucket kb-dev --mode auto dodil k3 vector-store get --bucket kb-dev -o json

Vector Collection Commands

Command group: dodil k3 vector-collection

Persistent flag:

  • --bucket, -b
SubcommandArgsImportant flags
add<name>--description, --pipeline, --dimensions, --metric
get<id>--bucket
listnone--bucket
delete<id>--bucket

Example:

dodil k3 vector-collection add contracts-v1 \ --bucket kb-dev \ --pipeline object_embedding_index \ --dimensions 1536 \ --metric cosine

Search Command

Command: dodil k3 search [query]

FlagDefaultPurpose
--bucket, -bnoneBucket to search
--collection, -cnoneRestrict to one collection
--top-k, -k10Max results
--min-score, -s0.0Score threshold

Example:

dodil k3 search "contract renewal terms" \ --bucket kb-dev \ --collection contracts-v1 \ --top-k 8 \ --min-score 0.2

Important Notes

  1. CLI does not expose first-class external vector write commands (InsertVectors, UpsertVectors, DeleteVectors).
  2. Use direct API for external/manual vector ingestion flows.
  3. Search behavior may go through compatibility mapping in some adapter layers; canonical API route remains /:bucket/vector/search.

Common Use Cases

  1. Quickly bootstrap vector retrieval in a new bucket.
  2. Validate search quality and thresholds from terminal.
  3. Manage collection lifecycle during model/dimension changes.

Next: Table and Engine Commands