โ† Back to walkthrough missions

Mission #01 ยท Intermediate

Observe prompt & model assets

Tell BHANUJ AIGP about a prompt and model your app used. Your app still writes prompts and runs models itself.

Asset RegistryRuntime observationEvidence lineage
Website Academy is a guided journey. It never calls your localhost services or reads their state. Mark each step complete yourself; Studio validates the runtime once you choose to open it.
  1. STEP 01

    Start AIGP locally

    Start the local stack from your AIGP checkout.

    Run in your AI Governance Platform checkout
    ./servers.sh

    Expected outcome

    The launcher finishes and Studio is available locally.

  2. STEP 02

    Open local Studio

    Sign in to Studio with the local development credentials from your Keycloak environment.

    Expected outcome

    You reach the Studio home screen and can open the Assets workspace.

  3. STEP 03

    Prepare your app credentials

    Set the local API address and a bearer token from your sign-in provider.

    Run in your AI Governance Platform checkout
    export AI_GOVERNANCE_API_URL=http://localhost:8000
    export AI_GOVERNANCE_TOKEN=replace-with-a-bearer-token

    Expected outcome

    Your app now has what it needs to call the observation API.

  4. STEP 04

    Record a prompt your app used

    Send the stable identity of the prompt used by your app.

    Run in your AI Governance Platform checkout
    curl -X POST "$AI_GOVERNANCE_API_URL/api/v1/prompts/observations" \
      -H "Authorization: Bearer $AI_GOVERNANCE_TOKEN" \
      -H 'Content-Type: application/json' \
      -d '{
        "name": "support-assistant",
        "version": "v7",
        "source_system": "support-api",
        "source_reference": "evaluation-run-123",
        "content_hash": "sha256:replace-with-the-runtime-hash",
        "variables": ["question", "context"]
      }'

    Expected outcome

    The API accepts the observation. Repeating the identical request is idempotent.

  5. STEP 05

    Check record

    Find the new prompt in the Prompt Catalog and see where it came from.

    Expected outcome

    The prompt is marked Observed. Its text is not shown when you supplied only a hash.