feat(hunt): feature table, producers, labels, metric, artifact, the DSPy hunt program, runner and CLI

hunter-learning-loop groups 1-3 (first increment). The feature table is the
contract: one row per touched function with diff, source, semantic and
taxonomy columns, provenance per producer, labels by the benchmark's rule.
Producers read the repository at head only (a planted advisory id and
commit text never reach a feature). The hunt program is a DSPy ReAct loop
over a signature whose inputs are the unmasked columns in a fixed order;
answers outside the closed family list or citing unseen lines score as
unknown. The runner writes records the benchmark scorer reads, resumable
by change, program id and mask, and refuses the held-out pilot unless the
run is a report run. Research extra [hunt] = dspy, pandas; the core
install is unchanged. 43 tests on a scripted model, no network.

Co-Authored-By: Claude Fable 5.1 <[email protected]>

tests/test_hunt_artifact.py

@@ -0,0 +1,49 @@
+"""Checks for stable run records and budget accounting."""
+
+import json
+
+import pytest
+
+from openultrasast.hunt.artifact import RunArtifact, UsageLedger
+
+
+def artifact():
+    return RunArtifact(program_id="program", schema_version=1, mask=("source",), data_digests={"train": "abc"},
+                       configurations=[{"seed": 0}], usage={"usd": 0.1, "calls": 1, "replay_hits": 0, "wall_seconds": 2},
+                       versions={"dspy": "3.4.0", "litellm": "1", "openultrasast": "2.0.1"}, created="2026-10-13T00:00:00Z")
+
+
+def test_artifact_round_trip_and_tamper(tmp_path):
+    run = artifact()
+    assert run.artifact_id == artifact().artifact_id
+    path = run.write(tmp_path)
+    assert RunArtifact.read(path) == run
+    payload = json.loads(path.read_text())
+    payload["program_id"] = "tampered"
+    path.write_text(json.dumps(payload))
+    with pytest.raises(ValueError, match="id"):
+        RunArtifact.read(path)
+
+
+def test_ledger_and_budget_prevent_artifact(tmp_path):
+    run = artifact()
+    ledger = UsageLedger(artifact=run)
+    ledger.add({"usd": 0.25, "calls": 1, "wall_seconds": 2})
+    ledger.add({"usd": 0, "replay_hits": 1, "wall_seconds": 1})
+    assert ledger.usage == {"usd": 0.25, "calls": 1, "replay_hits": 1, "wall_seconds": 3}
+    assert ledger.exhausted({"usd": 0.25, "calls": 10})
+    assert ledger.exhausted({"usd": 10, "calls": 1})
+    assert ledger.finish({"usd": 0.25}) is None
+    assert not list(tmp_path.iterdir())
+    result = ledger.finish({"usd": 1, "calls": 2})
+    assert result is run
+    assert result.usage == ledger.usage
+    assert run.finish({"usd": 0.25}) is None
+    assert run.write(tmp_path, budget={"usd": 0.25}) is None
+    assert not list(tmp_path.iterdir())
+
+
+@pytest.mark.parametrize("value", [-1, float("nan"), float("inf")])
+def test_invalid_usage_rejected(value):
+    with pytest.raises(ValueError):
+        UsageLedger().add({"usd": value})