File size: 8,286 Bytes
2a98962 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 | """Import Python-generated CORP-ENV examples into verification JSONL.
This is a convenience bridge for files such as:
data/raw/e1_to_e100_tasks.py
data/raw/m1_to_m100_tasks.py
The importer looks for either:
1. list/tuple variables containing dictionaries, or
2. generated `CorpTask` subclasses.
For generated task classes, it synthesizes compatible action trajectories for
the current environment tracks (`e1_launch_readiness` and
`m1_budget_reallocation`) while preserving the generated task description as
metadata and prompt text.
Example:
uv run python scripts/import_generated_examples.py \
--inputs data/raw/e1_to_e100_tasks.py data/raw/m1_to_m100_tasks.py \
--output data/raw/e1_m1_examples.jsonl
"""
from __future__ import annotations
import argparse
import importlib.util
import sys
from pathlib import Path
import inspect
import json
from typing import Any, Dict, Iterable, List, Type
ROOT = Path(__file__).resolve().parents[1]
if str(ROOT) not in sys.path:
sys.path.insert(0, str(ROOT))
from server.tasks.base import CorpTask # noqa: E402
from scripts._trajectory_utils import write_jsonl # noqa: E402
TASK_HINTS = {
"e1": "e1_launch_readiness",
"m1": "m1_budget_reallocation",
"h1": "h1_acquisition_defence",
}
def load_module(path: Path) -> Any:
spec = importlib.util.spec_from_file_location(path.stem, path)
if spec is None or spec.loader is None:
raise ValueError(f"cannot import {path}")
module = importlib.util.module_from_spec(spec)
spec.loader.exec_module(module)
return module
def infer_task_id(path: Path, obj: Dict[str, Any]) -> str:
explicit = obj.get("task_id") or obj.get("task")
if explicit:
return str(explicit)
lowered = path.stem.lower()
for hint, task_id in TASK_HINTS.items():
if lowered.startswith(hint) or f"_{hint}_" in lowered:
return task_id
return ""
def candidate_examples(module: Any) -> Iterable[Dict[str, Any]]:
preferred_names = (
"examples",
"tasks",
"trajectories",
"E1_TASKS",
"M1_TASKS",
"E1_EXAMPLES",
"M1_EXAMPLES",
)
seen_ids = set()
for name in preferred_names + tuple(dir(module)):
if name.startswith("__") or name in seen_ids:
continue
seen_ids.add(name)
value = getattr(module, name, None)
if isinstance(value, (list, tuple)) and value and all(isinstance(x, dict) for x in value):
for item in value:
yield dict(item)
def generated_task_classes(module: Any) -> Iterable[Type[CorpTask]]:
for _, value in vars(module).items():
if not inspect.isclass(value) or value is CorpTask:
continue
try:
if issubclass(value, CorpTask):
yield value
except TypeError:
continue
def is_e1_file(path: Path) -> bool:
return path.stem.lower().startswith("e1")
def is_m1_file(path: Path) -> bool:
return path.stem.lower().startswith("m1")
def synthesize_e1_actions(description: str) -> List[Dict[str, Any]]:
return [
{
"action_type": "delegate",
"agent_id": "qa_engineer",
"payload": f"Assess launch readiness for this generated scenario: {description}",
},
{
"action_type": "log_reasoning",
"payload": (
"Use the QA report as the primary launch gate and decide whether "
"the release should proceed within the 48 hour window."
),
},
{
"action_type": "log_decision",
"payload": "Finalize based on QA stability, blockers, and launch gate evidence.",
},
{"action_type": "finalize", "payload": "NO_GO"},
]
def synthesize_m1_actions(description: str) -> List[Dict[str, Any]]:
final = {
"phase_1": "Approve a capped GPU allocation for the highest-priority training runs.",
"phase_2": "Expand spend only after utilization and finance runway checks are reviewed.",
"guardrail": "Track budget, cost, spend, cash runway, and burn every week.",
"source_scenario": description[:300],
}
return [
{
"action_type": "delegate",
"agent_id": "dev_lead",
"payload": f"State the engineering requirement and minimum viable plan for: {description}",
},
{
"action_type": "delegate",
"agent_id": "fpa_manager",
"payload": f"State finance constraints, budget limits, runway, and spend guardrails for: {description}",
},
{
"action_type": "log_reasoning",
"payload": (
"The recommendation must balance engineering urgency against budget, "
"cost, spend, cash runway, and burn constraints."
),
},
{
"action_type": "log_conflict",
"payload": json.dumps(
{
"id": "c1",
"summary": "Engineering requirements exceed what finance should approve immediately.",
"source_agents": ["dev_lead", "fpa_manager"],
}
),
},
{
"action_type": "log_resolution",
"payload": json.dumps(
{
"conflict_id": "c1",
"resolution_type": "phased_budget",
"text": "Approve a capped phase_1 allocation with finance review before expansion.",
}
),
},
{"action_type": "finalize", "payload": json.dumps(final)},
]
def examples_from_task_classes(path: Path, module: Any) -> List[Dict[str, Any]]:
rows: List[Dict[str, Any]] = []
for idx, cls in enumerate(generated_task_classes(module), start=1):
generated_task_id = str(getattr(cls, "task_id", cls.__name__))
description = str(getattr(cls, "description", generated_task_id))
if is_e1_file(path):
task_id = "e1_launch_readiness"
actions = synthesize_e1_actions(description)
elif is_m1_file(path):
task_id = "m1_budget_reallocation"
actions = synthesize_m1_actions(description)
else:
continue
rows.append(
{
"example_id": f"{path.stem}-{idx:03d}",
"task_id": task_id,
"source_file": str(path),
"source_kind": "generated_task_class",
"source_class": cls.__name__,
"generated_task_id": generated_task_id,
"generated_description": description,
"actions": actions,
}
)
return rows
def import_file(path: Path) -> List[Dict[str, Any]]:
module = load_module(path)
rows: List[Dict[str, Any]] = []
for idx, obj in enumerate(candidate_examples(module), start=1):
task_id = infer_task_id(path, obj)
if task_id:
obj["task_id"] = task_id
obj.setdefault("example_id", f"{path.stem}-{idx:03d}")
obj.setdefault("source_file", str(path))
rows.append(obj)
if not rows:
rows.extend(examples_from_task_classes(path, module))
return rows
def main() -> None:
parser = argparse.ArgumentParser(description="Import generated Python examples to JSONL.")
parser.add_argument("--inputs", nargs="+", required=True)
parser.add_argument("--output", default="data/raw/e1_m1_examples.jsonl")
args = parser.parse_args()
rows: List[Dict[str, Any]] = []
for input_path in args.inputs:
path = Path(input_path)
if not path.exists():
raise SystemExit(f"Input not found: {path}")
imported = import_file(path)
print(f"{path}: imported {len(imported)} examples")
rows.extend(imported)
if not rows:
raise SystemExit(
"No examples found. Expected a module-level list of dictionaries "
"or generated CorpTask subclasses."
)
write_jsonl(Path(args.output), rows)
print(f"Wrote {len(rows)} examples to {args.output}")
print("Next: run scripts/verify_examples.py on the JSONL output.")
if __name__ == "__main__":
main()
|