Lab 03: retrieve a known document

Prerequisites: Lab 01, a disposable knowledge store, and no sensitive input. Artifact: the retrieved source and an explanation of how it affected the answer.

Exercise

1. Before ingestion, ask which port the fictional Aurora service uses. Record whether there is any relevant source; do not treat a guessed answer as correct.
2. Follow Ingest and retrieve documents to add the example document.
3. Repeat the question with context_limit: 1 and include_sources: true.
4. Inspect source content and metadata. Verify that the returned fact is the one you ingested.
5. Explain the difference between “the model answered 8123” and “the returned source supports 8123.”

Acceptance criteria

• Ingestion reports one document.
• The post-ingestion source contains the fictional service statement.
• You record any unsupported detail added by the model.
• You understand that include_sources: false hides sources; it does not disable retrieval.

Cleanup

The current public API does not delete knowledge documents. Keep the exercise in a disposable store and record its document ID. Do not remove a shared or live store to clean up a single exercise.

Next: Understand RAG and storage or review your learning path.