59. Running an advisory¶

Once a local development environment is healthy, the canonical check is to produce a synthetic advisory: an advisory generated from non-production data for a chosen location and crop context. It exercises the whole pipeline end to end without touching real farmer data.

59.1 What a run should produce¶

A successful run returns an auditable advisory package. When inspecting it, look for the signals that confirm the pipeline behaved correctly:

• a recommendation for the requested location and context;
• the decision trace linking the recommendation back to its inputs (see Decision trace and confidence);
• a confidence vector, banded for readability;
• any degradations recorded on the advisory, if inputs were missing (see Observability and degradation).

flowchart LR
    REQ[Location + crop context] --> PIPE[Advisory pipeline]
    PIPE --> PKG[Advisory package]
    PKG --> CHK[Recommendation + trace + confidence + degradations]

59.2 Reading the result¶

Compare the output against a known-good expected result for the same synthetic input. Because trace-node identity is deterministic (see Decision trace and confidence), an identical input should produce an identical decision, so any difference is meaningful and worth investigating. For the phases a request passes through, see the advisory lifecycle.

The exact command to invoke a local advisory belongs with the code; follow the project's own developer documentation for the current invocation.