Guides

Run Evaluate your pipeline before changing retrieval settings. After the quickstart, use the remaining guides for corpus discovery, database setup, live operations, deployment, and extension.

Evaluate your pipelineScore retrieval layer by layer, grade generated answers, compare two runs, and check the grader against human judgment.

Run a corpus campaignFind papers by topic or DOI list, resolve their rights, download the open-access PDFs, and screen them against stated criteria.

Configure Postgres backendDocker, conda-forge, a system PostgreSQL or Postgres.app, or the optional Cloud SQL development helper.

Operate a live corpusBack up, restore, snapshot, delete, and re-embed a corpus that keeps changing under a running service.

Deploy on Google CloudProvision Cloud SQL and Cloud Run from the included Terraform and verify the running service.

Extend the kitAdd a parser, corpus collector, reranker, model provider, or authentication backend at one of the five supported boundaries.

Begin with Bring your own domain to ingest your first corpus. Use the reference section when a guide names a command, field, or contract without listing every detail.