SILVER/NODEDC_MISSION_CORE
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

refactor(lab): admit versioned graph ledgers

Browse Source
This commit is contained in:
DCCONSTRUCTIONS
2026-08-26 13:10:11 +03:00
parent 252cbc8495
commit b688347778
1 changed files with 60 additions and 20 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/k1link/perception/m48s_replay_timeline.py
+60 -20
View File
@@ -38,13 +38,13 @@ from .threat_timeline import (
)
FRAME_EVIDENCE_SCHEMA: Final = "missioncore.m48s-reference-graph-frame-evidence/v0"
M48R3_FRAME_EVIDENCE_SCHEMA: Final = (
"missioncore.m48s-reference-graph-frame-evidence/v1"
)
EXPECTED_FRAME_COUNT: Final = 4_489
CAMERA_ACCUMULATION_WINDOW_SECONDS: Final = 2.0
CAMERA_ACCUMULATION_POINT_LIMIT: Final = 20_000
CAMERA_POINT_OVERLAY_SCHEMA: Final = "missioncore.m48s-camera-point-overlay/v1"
_FRAME_EVIDENCE_SCHEMA_MARKER: Final = (
b'"schema_version":"missioncore.m48s-reference-graph-frame-evidence/v0"'
)
_SOURCE_ENVELOPE_MARKER: Final = b'"source_envelope":'
_JSON_DECODER: Final = json.JSONDecoder()
@@ -61,16 +61,37 @@ class _LedgerIndex:
class M48sReplayTimeline:
"""Read source-indexed chunks while preserving latest-wins world-state gaps."""
def __init__(self, *, repository_root: Path, result_root: Path, result_id: str) -> None:
def __init__(
self,
*,
repository_root: Path,
result_root: Path,
result_id: str,
frames_name: str = "reference-graph-replay-frames.jsonl",
worker_result_name: str = "reference-graph-replay-worker-result.json",
frame_evidence_schema: str = FRAME_EVIDENCE_SCHEMA,
camera_endpoint_root: str = (
"/api/v1/laboratory/m48s/fixed-class-detector"
),
) -> None:
self.repository_root = repository_root.resolve(strict=True)
self.result_root = result_root.resolve(strict=True)
self.result_id = result_id
self.frames_path = (
self.result_root / "reference-graph-replay-frames.jsonl"
).resolve(strict=True)
self.worker_path = (
self.result_root / "reference-graph-replay-worker-result.json"
).resolve(strict=True)
if (
Path(frames_name).name != frames_name
or Path(worker_result_name).name != worker_result_name
or frame_evidence_schema not in {
FRAME_EVIDENCE_SCHEMA,
M48R3_FRAME_EVIDENCE_SCHEMA,
}
or not camera_endpoint_root.startswith("/api/v1/laboratory/")
or camera_endpoint_root.endswith("/")
):
raise M48sReplayTimelineError("M4.8S timeline binding is invalid")
self.frame_evidence_schema = frame_evidence_schema
self.camera_endpoint_root = camera_endpoint_root
self.frames_path = (self.result_root / frames_name).resolve(strict=True)
self.worker_path = (self.result_root / worker_result_name).resolve(strict=True)
if (
self.frames_path.parent != self.result_root
or self.worker_path.parent != self.result_root
@@ -96,7 +117,12 @@ class M48sReplayTimeline:
raise M48sReplayTimelineError("M4.8S source clock is not monotonic")
worker = _object(json.loads(self.worker_path.read_text("utf-8")), "worker result")
self.outcomes = _terminal_outcomes(worker)
self.index = _index_ledger(self.frames_path, self.source_times_ns, self.outcomes)
self.index = _index_ledger(
self.frames_path,
self.source_times_ns,
self.outcomes,
frame_evidence_schema=self.frame_evidence_schema,
)
self._cache_lock = Lock()
self._chunk_json_cache: OrderedDict[tuple[int, int], bytes] = OrderedDict()
self._camera_point_json_cache: OrderedDict[int, bytes] = OrderedDict()
@@ -432,8 +458,8 @@ class M48sReplayTimeline:
"camera_proposals": camera_proposals,
"decision_counts": _decision_counts(assessments),
"camera_url": (
"/api/v1/laboratory/m48s/fixed-class-detector/"
f"{self.result_id}/timeline/frames/{sequence}/camera"
f"{self.camera_endpoint_root}/{self.result_id}"
f"/timeline/frames/{sequence}/camera"
),
"ground_truth": False,
"authority": "replay-simulated",
@@ -447,7 +473,10 @@ class M48sReplayTimeline:
stream.seek(offset)
line = stream.readline()
value = json.loads(line)
if not isinstance(value, dict) or value.get("schema_version") != FRAME_EVIDENCE_SCHEMA:
if (
not isinstance(value, dict)
or value.get("schema_version") != self.frame_evidence_schema
):
raise M48sReplayTimelineError("M4.8S frame row is invalid")
return value
@@ -456,6 +485,8 @@ def _index_ledger(
path: Path,
source_times_ns: tuple[int, ...],
outcomes: dict[int, str],
*,
frame_evidence_schema: str,
) -> _LedgerIndex:
offsets: dict[int, int] = {}
with path.open("rb") as stream:
@@ -464,7 +495,10 @@ def _index_ledger(
line = stream.readline()
if not line:
break
envelope = _ledger_source_envelope(line)
envelope = _ledger_source_envelope(
line,
frame_evidence_schema=frame_evidence_schema,
)
timestamps = _object(envelope.get("timestamps"), "source timestamps")
sequence = envelope.get("sequence")
if (
@@ -483,7 +517,11 @@ def _index_ledger(
return _LedgerIndex(offsets)
def _ledger_source_envelope(line: bytes) -> dict[str, object]:
def _ledger_source_envelope(
line: bytes,
*,
frame_evidence_schema: str,
) -> dict[str, object]:
"""Validate a ledger row while decoding only its small trailing envelope.
The full row can exceed 100 KiB because it contains the delivered world
@@ -492,10 +530,10 @@ def _ledger_source_envelope(line: bytes) -> dict[str, object]:
the GIL for many seconds.
"""
if (
line.count(_FRAME_EVIDENCE_SCHEMA_MARKER) != 1
or line.count(_SOURCE_ENVELOPE_MARKER) != 1
):
schema_marker = (
f'"schema_version":"{frame_evidence_schema}"'.encode("ascii")
)
if line.count(schema_marker) != 1 or line.count(_SOURCE_ENVELOPE_MARKER) != 1:
raise M48sReplayTimelineError("M4.8S ledger schema changed")
start = line.find(_SOURCE_ENVELOPE_MARKER) + len(_SOURCE_ENVELOPE_MARKER)
try:
@@ -569,6 +607,8 @@ __all__ = [
"CAMERA_ACCUMULATION_POINT_LIMIT",
"CAMERA_ACCUMULATION_WINDOW_SECONDS",
"CAMERA_POINT_OVERLAY_SCHEMA",
"FRAME_EVIDENCE_SCHEMA",
"M48R3_FRAME_EVIDENCE_SCHEMA",
"M48sReplayTimeline",
"M48sReplayTimelineError",
]