amos f611da1e57 Batch 24: add 23 calibrated tasks written from Bos seed directions
Batch 24 (20261102-C-bos-seed-24) screened all 3078 Bos seed queries.
Only 23 themes were in scope (pure user-space reliability, at most one
task per theme); the rest were single operations or explanations,
installs and networking, system administration, ROS2, board-only or GUI
requests. Two finished drafts were withdrawn as too close to existing
tasks. Seeds supply direction only: all sources, contracts and tests are
new, and no task shares more than 14 characters with its seed text.

Each task passed the local gate and the trace6 Docker gate (5 runs per
sample class). Claude rebuilt the catalog from the calibration receipts:
23/23 USABLE. 16 of 23 are C/C++.

Only the task directories are taken; the batch worktree's lowered disk
floor in rollout/core.py stays out of main.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-02 17:21:50 +08:00

202 lines
6.2 KiB
Python

"""Parse and validate the public batch format before any result is built."""
from dataclasses import dataclass
from typing import Any, Dict, List, Tuple
class InputError(ValueError):
pass
@dataclass(frozen=True)
class Grid:
width: int
height: int
rows: Tuple[Tuple[int, ...], ...]
@dataclass(frozen=True)
class ImageSize:
width: int
height: int
@dataclass(frozen=True)
class Detection:
name: str
box: Tuple[int, int, int, int]
meta: Any
@dataclass(frozen=True)
class Frame:
name: str
grid: Grid
image: ImageSize
detections: Tuple[Detection, ...]
meta: Any
@dataclass(frozen=True)
class Batch:
frames: Tuple[Frame, ...]
meta: Any
def _object(value: Any, path: str) -> Dict[str, Any]:
if type(value) is not dict:
raise InputError(path + " must be an object")
return value
def _array(value: Any, path: str) -> List[Any]:
if type(value) is not list:
raise InputError(path + " must be an array")
return value
def _exact_keys(value: Dict[str, Any], keys: set, path: str) -> None:
if set(value) != keys:
raise InputError(path + " has missing or extra fields")
def _integer(value: Any, low: int, high: int, path: str) -> int:
if type(value) is not int:
raise InputError(path + " must be an integer")
if value < low or value > high:
raise InputError(path + " is outside its range")
return value
def _name(value: Any, path: str) -> str:
if type(value) is not str:
raise InputError(path + " must be a string")
if not value or len(value) > 48:
raise InputError(path + " has invalid length")
return value
def _meta(value: Any, path: str) -> Any:
if type(value) is not dict:
raise InputError(path + " must be an object")
_metadata_tree(value, path, 0)
return value
def _metadata_tree(value: Any, path: str, level: int) -> None:
kind = type(value)
if kind in (dict, list) and level > 16:
raise InputError(path + " is too deeply nested")
if kind is dict:
for key, child in value.items():
if type(key) is not str:
raise InputError(path + " has a non-string key")
_metadata_tree(child, path + "." + key, level + 1)
return
if kind is list:
for index, child in enumerate(value):
_metadata_tree(child, path + "[" + str(index) + "]", level + 1)
return
if kind in (str, int, bool, type(None)):
return
if kind is float:
if value == value and value != float("inf") and value != -float("inf"):
return
raise InputError(path + " contains a non-JSON value")
def _size(value: Any, path: str) -> ImageSize:
obj = _object(value, path)
_exact_keys(obj, {"width", "height"}, path)
width = _integer(obj["width"], 1, 64, path + ".width")
height = _integer(obj["height"], 1, 64, path + ".height")
return ImageSize(width, height)
def _depth(value: Any, path: str) -> Grid:
obj = _object(value, path)
_exact_keys(obj, {"width", "height", "values"}, path)
width = _integer(obj["width"], 1, 64, path + ".width")
height = _integer(obj["height"], 1, 64, path + ".height")
values = _array(obj["values"], path + ".values")
if len(values) != height:
raise InputError(path + ".values has wrong row count")
rows = []
for row_number, raw_row in enumerate(values):
row_path = path + ".values[" + str(row_number) + "]"
raw = _array(raw_row, row_path)
if len(raw) != width:
raise InputError(row_path + " has wrong column count")
row = []
for column, cell in enumerate(raw):
row.append(_integer(cell, 0, 10000,
row_path + "[" + str(column) + "]"))
rows.append(tuple(row))
return Grid(width, height, tuple(rows))
def _box(value: Any, image: ImageSize, path: str) -> Tuple[int, int, int, int]:
raw = _array(value, path)
if len(raw) != 4:
raise InputError(path + " must have four coordinates")
x0 = _integer(raw[0], 0, image.width, path + "[0]")
y0 = _integer(raw[1], 0, image.height, path + "[1]")
x1 = _integer(raw[2], 0, image.width, path + "[2]")
y1 = _integer(raw[3], 0, image.height, path + "[3]")
if x0 >= x1 or y0 >= y1:
raise InputError(path + " has no area")
return x0, y0, x1, y1
def _detection(value: Any, image: ImageSize, path: str) -> Detection:
obj = _object(value, path)
_exact_keys(obj, {"id", "box", "meta"}, path)
name = _name(obj["id"], path + ".id")
box = _box(obj["box"], image, path + ".box")
meta = _meta(obj["meta"], path + ".meta")
return Detection(name, box, meta)
def _detections(value: Any, image: ImageSize, path: str) -> Tuple[Detection, ...]:
raw = _array(value, path)
if len(raw) > 64:
raise InputError(path + " has too many detections")
found = set()
output = []
for index, item in enumerate(raw):
detection = _detection(item, image, path + "[" + str(index) + "]")
if detection.name in found:
raise InputError(path + " repeats an id")
found.add(detection.name)
output.append(detection)
return tuple(output)
def _frame(value: Any, path: str) -> Frame:
obj = _object(value, path)
_exact_keys(obj, {"frame_id", "image", "depth", "detections", "meta"}, path)
name = _name(obj["frame_id"], path + ".frame_id")
image = _size(obj["image"], path + ".image")
grid = _depth(obj["depth"], path + ".depth")
detections = _detections(obj["detections"], image, path + ".detections")
meta = _meta(obj["meta"], path + ".meta")
return Frame(name, grid, image, detections, meta)
def parse_batch(value: Any) -> Batch:
obj = _object(value, "batch")
_exact_keys(obj, {"frames", "meta"}, "batch")
meta = _meta(obj["meta"], "batch.meta")
raw = _array(obj["frames"], "batch.frames")
if len(raw) > 16:
raise InputError("batch.frames has too many frames")
frames = []
found = set()
for index, item in enumerate(raw):
frame = _frame(item, "batch.frames[" + str(index) + "]")
if frame.name in found:
raise InputError("batch.frames repeats a frame_id")
found.add(frame.name)
frames.append(frame)
return Batch(tuple(frames), meta)