knowledge-admin

To connect and configure a pgvector backend for your knowledge base, you register the connection to automatically run Postgres and pgvector version validation, detect PgBouncer edge cases, apply Alembic migrations, and seed required configuration.

Monitoring pgvector health and capacity for knowledge storage involves checking schema versioning, pgvector version, document and chunk counts, space counts, and the last error per connection to quickly identify broken environments.

To export documents and chunks from a specific knowledge space, you select a connection and export the data, optionally including chunks, storing results into a workspace export directory for backups or downstream processing.

To install Marker OCR parser models for document ingestion, you run the install-parser action, which idempotently installs large parser models and provides installation progress with safe no-ops when already installed.

Yes, knowledge-admin works with PgBouncer connection pooling by detecting PgBouncer edge cases during connection validation, ensuring safe database onboarding and troubleshooting for your knowledge base.

Your pgvector database migrations might be failing during onboarding due to unsupported Postgres or pgvector versions, or PgBouncer edge cases; the system validates these and applies Alembic migrations to seed configuration.