feat(detection-core): migrate benchmark readiness and admission consumers

benchmarks/unseen/calibrate.py

@@ -12,9 +12,9 @@ import statistics
 from datetime import date
 from pathlib import Path
 
-from benchmarks.ax.batch import write_json
 from benchmarks.unseen.replay import canonical, load_slice, run_replays, sample_changes
 from benchmarks.unseen.score import coverage, load_records
+from benchmarks.valid.artifact_contract import write_benchmark_artifact
 
 
 def stage_times(record):
@@ -120,7 +120,7 @@ def main():
     result.update(
         image=args.image, deadline_seconds=args.deadline, sample_digest=hashlib.sha256(canonical([c["id"] for c in changes])).hexdigest()
     )
-    write_json(args.record, result)
+    write_benchmark_artifact(args.record, "deadline-calibration", result, replace=True)
     return 0 if result["status"] == "ok" else 2
 
 

benchmarks/unseen/replay.py

@@ -14,6 +14,7 @@ from pathlib import Path
 
 from benchmarks.ax.batch import AXLane, DockerLane, Item, Workload, dispatch, write_json
 from benchmarks.unseen.score import coverage
+from benchmarks.valid.artifact_contract import read_benchmark_artifact, write_benchmark_sequence
 
 
 def make_item(change, arm, image, *, deadline_scale=1.0, lane="ax", deadline=None):
@@ -180,7 +181,7 @@ def main():
     changes = load_slice(args.manifest, args.slice, args.private_manifest)
     factor = args.deadline_scale
     if args.calibration:
-        calibration = json.loads(args.calibration.read_text())
+        calibration = read_benchmark_artifact(args.calibration, "deadline-calibration")
         if calibration.get("status") != "ok" or calibration.get("image") != args.image:
             parser.error("calibration must be successful and match the replay image")
         factor = calibration["quick_total"]["factor"]
@@ -213,9 +214,11 @@ def run_replays(changes, args, out, lane, *, deadline_scale=1.0, deadline=None):
     if out.resolve().is_relative_to(Path(__file__).resolve().parents[2]):
         raise ValueError("raw replay output must be outside the repository")
     items = [make_item(c, args.arm, args.image, deadline_scale=deadline_scale, lane=lane, deadline=deadline) for c in changes]
-    write_json(
+    write_benchmark_sequence(
         out / "inputs.json",
+        "benchmark-arm-inputs",
         [{"item_id": item.id, "change": change, "input_digest": item.input_digest} for item, change in zip(items, changes, strict=True)],
+        replace=True,
     )
     maximum = max(
         max(json.loads(i.inputs["SPEC_URL"])["deadline_seconds"], json.loads(i.inputs["SPEC_URL"])["base_deadline_seconds"]) for i in items
@@ -240,7 +243,7 @@ def run_replays(changes, args, out, lane, *, deadline_scale=1.0, deadline=None):
             return record
 
         executors[name] = execute
-    return dispatch(items, workload, out, executors, lanes=lanes)
+    return dispatch(items, workload, out, executors, lanes=lanes, record_kind="benchmark-arm-record")
 
 
 if __name__ == "__main__":

benchmarks/unseen/score.py

@@ -9,7 +9,13 @@ import math
 import random
 from collections import Counter, defaultdict
 from pathlib import Path
+from typing import Any, cast
 
+from benchmarks.valid.artifact_contract import (
+    read_benchmark_artifact,
+    read_benchmark_sequence,
+    write_benchmark_artifact,
+)
 from openultrasast.model.taxonomy import load_families
 from openultrasast.push.report import CAPABILITY_VETOES
 
@@ -324,13 +330,17 @@ def gate(a, b, *, slice_id=None, informed=False, binary=True, threshold=None, th
     }
 
 
-def load_records(path):
+def load_records(path: Path) -> list[dict[str, Any]]:
+    """Load only canonical benchmark v2 records; producer handoffs are explicit adapters."""
+    path = Path(path)
     if path.is_file():
-        return json.loads(path.read_text())
-    rows = []
-    for entry in json.loads((path / "inputs.json").read_text()):
+        return [cast(dict[str, Any], row) for row in read_benchmark_sequence(path, "benchmark-arm-records")]
+    rows: list[dict[str, Any]] = []
+    entries = read_benchmark_sequence(path / "inputs.json", "benchmark-arm-inputs")
+    for raw_entry in entries:
+        entry = cast(dict[str, Any], raw_entry)
         result = path / (hashlib.sha256(entry["item_id"].encode()).hexdigest() + ".json")
-        record = json.loads(result.read_text()) if result.exists() else {}
+        record: dict[str, Any] = cast(dict[str, Any], read_benchmark_artifact(result, "benchmark-arm-record")) if result.exists() else {}
         if record.get("input_digest") != entry["input_digest"]:
             record = {}
         rows.append({**entry["change"], **record, "id": entry["change"]["id"]})
@@ -356,7 +366,7 @@ def main():
     result["unseen"] = gate(load_records(args.against) if args.against else [], current, slice_id=args.slice_id, informed=args.informed)
     if native and result["flagged"]:
         result["unseen"] = {"gate": "not_run", "reason": "native_agreement"}
-    args.out.write_text(json.dumps(result, sort_keys=True, indent=2) + "\n")
+    write_benchmark_artifact(args.out, "unseen-score", result, replace=True)
 
 
 if __name__ == "__main__":