[Catalog](https://skills.polygres.com/index.md) / [polygres-cli](https://skills.polygres.com/polygres-cli/SKILL.md) Published skills release: 0.9.0. [Compatibility and source](https://skills.polygres.com/manifest.json). Scripts and templates mentioned here require the installed package; reading this page does not install them. # Retrieval configuration ## Contents - [Graph](#graph) - [Managed embeddings](#managed-embeddings) - [Existing vector configurations](#existing-vector-configurations) - [Text](#text) - [General readiness](#general-readiness) ## Graph ```bash polygres graph discover polygres graph config export polygres graph config apply --file graph.json polygres graph build polygres graph status ``` `graph discover` is read-only. `graph config apply` accepts a raw configuration or an export wrapper with a non-null `configuration`. Review the full document before applying. `graph build` invokes the backend build and can mutate index state. For a temporary table created solely to validate pgGraph, do not enable Row Level Security or `FORCE ROW LEVEL SECURITY` unless the test explicitly targets RLS behavior. Before reporting a pgGraph build failure, verify every registered test table has both `relrowsecurity = false` and `relforcerowsecurity = false`. If either flag caused the failure, classify it as an incompatible test fixture, not a pgGraph product failure. Never disable RLS on a pre-existing user table without explicit approval. ## Managed embeddings To make source text searchable by meaning, use `polygres embeddings` to select a source and model, preview generation, and start processing. Then use the returned managed table to create a Context collection. Read [embeddings.md](https://skills.polygres.com/polygres-cli/references/embeddings.md) for setup, costs, processing, and removal, and [context.md](https://skills.polygres.com/polygres-cli/references/context.md#choose-retrieval) for text queries. ## Existing vector configurations ```bash polygres vector configs list polygres vector configs delete [--yes] polygres vector reindex ``` Use these commands to inspect or maintain previously registered configurations. For new semantic retrieval, start with `polygres context capabilities`, then use `polygres context collections create` with a native `pgcontext.vector` column or a compatible `public.vector` column that will be converted in place. The older `polygres vector configs create` path is retired and returns `VECTOR_CREATION_RETIRED`. Existing configurations remain listable, reindexable, and removable. Legacy retrieval additionally requires a persisted enabled registration that is effectively Ready. HNSW configurations require their exact physical index to be Ready; an existing `index_kind: none` configuration can be Ready for exact scan without HNSW. Never synthesize a configuration from an unregistered physical index or claim the retired API can register or re-enable it. Delete accepts a configuration ID and requires approval. When migrating a registered pgvector column, preflight the intended Context request, show the in-place type conversion and dependent-index impact, delete the exact legacy registration with approval, then create and verify the native collection. Do not use `context init` for this outcome; that command is the guided onboarding front end for choosing one eligible persisted Legacy source and submitting the same ordinary native collection-create request. Candidate discovery requires the certified compatibility extension and a Ready physical index; it cannot adopt a physical-only pgvector index. It does not activate an internal same-column binding or preserve the old pgvector index. ## Text ```bash polygres text configs list polygres text configs get polygres text configs create-fuzzy \ --table \ --text-column polygres text configs create-tsvector \ --table
\ --tsvector-column polygres text configs create-tsvector \ --table
\ --text-column \ --generated-column \ --yes polygres text configs update [options] polygres text configs diagnostics polygres text configs reindex polygres text configs delete [--yes] ``` Existing-column TSVector mode registers a compatible `tsvector` column without creating another column. Generated-column mode sends one request to the existing text configuration endpoint, which creates the stored generated column, creates and verifies its GIN index, and saves the configuration. It does not create or apply a migration. PostgreSQL keeps the generated value current when its source text changes. Generated-column mode requires explicit approval before `--yes`. If it fails, Polygres tries to remove the new column, index, and configuration. Stop and inspect the table and saved configurations when cleanup is reported as incomplete. Do not blindly retry with another generated-column name. Fuzzy mode uses an existing text-like column and creates its managed trigram index. Repeat `--row-id-column` for a compound primary or unique non-null key. Use `--default-limit` and `--max-limit` to bound results, repeat `--metadata-column` for returned properties, and repeat `--filter-column` for allowed exact-match filters. Ensure the default limit does not exceed the maximum. Use `get` for one saved definition and `diagnostics` to compare saved state with the physical index. A configuration is usable only when `index_status` is `ready`. After correcting a failed or stale target, obtain approval before `reindex`, then confirm diagnostics are healthy. Deleting a configuration removes its managed index, but does not drop a generated source-table column. ## General readiness ```bash polygres ready ``` This reports graph, vector, and hybrid readiness. It does not report text readiness in the current launch surface. Check text separately. Polygres AI Context has its own capability, collection, point, operation, and retrieval lifecycle. Use it for new semantic retrieval setup. Read [`context.md`](https://skills.polygres.com/polygres-cli/references/context.md) from this skill when the request mentions Context, pgContext, AI Search, collections, Joint retrieval, or point reconciliation. An existing pgvector configuration is never itself a pgContext collection, even when its physical column is eligible for in-place migration.