SATE-Laravel
← All pipelines

Governance & Learning

sate emit-learning-candidates

Opt-in

Opt-in: extracts STABLE_TESTHASH_PASS candidates from an execution trace for downstream review.

What it does

Reads an execution trace (.ai/execution-trace.ndjson), promotes test hashes that have ≥N PASS-only outcomes with zero filesystem deltas into candidates, and writes a JSON file ready for review-queue. Write-once per run-id — never overwrites. Requires SATE_LEARNING_OPT_IN=1 to do anything.

Who runs this

  • →CI when explicitly enabled to feed the learning pipeline

How to invoke

SATE_LEARNING_OPT_IN=1 sate emit-learning-candidates --run-id r1 --project-root /path/to/laravel-app

Flags

--run-id <id>

Required.

--project-root <path>

Resolves trace to <root>/.ai/execution-trace.ndjson.

--trace <path>

Override trace path.

--out <path>

Override output path (default co-located with trace).

--required-n <N>default: 5

Minimum PASS count to promote.

--quiet

Suppress opt-out notice + write-once skip notice.

What it produces

  • ·<trace-dir>/learning-candidates.<runId>.json — write-once payload

What it does NOT do

  • ✖Does NOT run without SATE_LEARNING_OPT_IN=1
  • ✖Does NOT overwrite an existing output file for the same run-id
  • ✖Does NOT promote candidates with any non-PASS outcomes or non-zero filesystem deltas

Status

Engine tests

17

Status

Opt-in

Category

Governance & Learning

Related pipelines

sate trace-report

Reads execution-trace NDJSON, emits deterministic anomaly summary: outcome distribution, per-prefix quantiles, top repeated hashes, duration outliers.

sate review-queue

Shows the REVIEW-class learning candidates a human needs to approve, reject, or defer.

sate submit-decision

Append a human decision (APPROVE / REJECT / DEFER) for a candidate to the append-only review ledger.

sate learning

Multi-repo learning loop that extracts patterns into the canonical knowledge base.