Vector pipelines drift fast: re-ingesting a file creates duplicate chunks, and nobody can tell which version is live. This flow keeps insert, upsert, and retrieval in one place.
Vector pipelines drift fast: re-ingesting a file creates duplicate chunks, and nobody can tell which version is live. This flow keeps insert, upsert, and retrieval in one place.
Everything sits in Supabase on pgvector, with upserts keyed to a stable document ID so re-ingesting replaces old chunks instead of piling them up. Every operation is logged with its source ID, so you can always trace what got embedded and when.
How it runs
Used at step 01 to kick off the pipeline.
Write
Used at step 01 to kick off the pipeline.
WebFetch
Used at step 01 to kick off the pipeline.
WebSearch
Used at step 01 to kick off the pipeline.
Upsert by stable document ID so re-ingesting a file replaces its old chunks instead of duplicating them.
pending
Retrieve the nearest chunks for a query using cosine distance and a metadata filter.
pending
Log every insert, upsert, and retrieval with its source ID so the whole flow stays auditable.
pending
Sample output
json
// Sample output
// (generated when the pipeline finishes)
Run insert, upsert, and retrieval over a Supabase pgvector store with stable IDs and a logged audit trail.
Unlock the rest
The full agent definition, install snippet, and starter task are gated for community members.
Members get the full `.md` agent file, the npm / pnpm install one-liners, a starter prompt that we've tuned against real runs, and the open-source repo when this automation ships there. One email, magic link, done.