A self-improving RLM agent for coding workflows and long-running autonomous tasks.
Rust
import copy
import json
import unittest
from check_factory_resumption import MARKER, extract, validate
class FactoryReceiptGuards(unittest.TestCase):
def probe(self):
return {'paused': {'state': 'paused'}, 'done': {'state': 'done', 'nodes': [
{'id': 'work', 'status': 'done', 'entries_used': 1, 'attempts': 2},
{'id': 'verified', 'status': 'done', 'entries_used': 1, 'attempts': 1}]}}
def test_rejects_missing_resume_extra_entry_or_extra_attempt(self):
validate(self.probe())
for key, value in [('entries_used', 2), ('attempts', 1), ('attempts', 3), ('status', 'error')]:
probe = self.probe()
probe['done']['nodes'][0][key] = value
with self.assertRaises(AssertionError):
validate(probe)
def test_accepts_explicit_wire_camel_case_without_ignoring_duplicates(self):
probe = self.probe()
for row in probe['done']['nodes']:
row['entriesUsed'] = row.pop('entries_used')
validate(probe)
probe['done']['nodes'][0]['entries_used'] = 1
with self.assertRaises(AssertionError):
validate(probe)
def test_rejects_missing_or_historical_tool_receipt(self):
probe = self.probe()
tool = {'role': 'tool', 'tool_call_id': 'factory_probe',
'content': MARKER + json.dumps(probe)}
rows = [{}, {'messages': [tool]}]
self.assertEqual(extract(rows), probe)
for changed in [[{'messages': [tool]}, {'messages': []}],
[{}, {'messages': [tool, copy.deepcopy(tool)]}],
[{}, {'messages': [{**tool, 'tool_call_id': 'historical'}]}]]:
with self.assertRaises(AssertionError):
extract(changed)
if __name__ == '__main__':
unittest.main()