A self-improving RLM agent for coding workflows and long-running autonomous tasks.
Rust
update saved-catalog parity check for the removed allMessagesText field (#3682)
kt2a7126c811f4
diff --git a/scripts/review_saved_catalog_parity.py b/scripts/review_saved_catalog_parity.pyindex 8b48d602b8..eb580884d9 100644--- a/scripts/review_saved_catalog_parity.py+++ b/scripts/review_saved_catalog_parity.py@@ -23,8 +23,7 @@ PAGE = 200BASE_MS = 1704067200000CWD = '/tmp/catalog-paging-fixture'TARGET = 'SAVED_CATALOG_OLD_NAME_SEARCH_TARGET'-TRANSCRIPT_TARGET = 'SAVED_CATALOG_OLD_ASSISTANT_WIRE_TARGET'-FIELDS = ('path', 'id', 'cwd', 'name', 'messageCount', 'firstMessage', 'allMessagesText')+FIELDS = ('path', 'id', 'cwd', 'name', 'messageCount', 'firstMessage')def iso(milliseconds):@@ -40,7 +39,7 @@ def seed(root):ident = f'catalog-{index:03}'stamp = BASE_MS + index * 10000user = f'Catalog user {index:03}'- assistant = TRANSCRIPT_TARGET if index == 0 else f'Catalog assistant {index:03}'+ assistant = f'Catalog assistant {index:03}'name = TARGET if index == 0 else f'Catalog row {index:03}'entries = [{'type': 'session', 'version': 3, 'id': ident, 'timestamp': iso(stamp),@@ -63,7 +62,7 @@ def seed(root):os.utime(path, ns=(ns, ns))assert path.stat().st_mtime_ns == ns, 'fixture filesystem lost deterministic mtime'expected.append(dict(zip(FIELDS, (path.name, ident, CWD, name,- 2, user, user + ' ' + assistant))))+ 2, user))))return list(reversed(expected))@@ -84,7 +83,7 @@ def normalize(rows, root):def same_rows(actual, expected):assert sorted(actual, key=lambda row: row['id']) == sorted(- expected, key=lambda row: row['id']), 'catalog stable fields differ (including allMessagesText)'+ expected, key=lambda row: row['id']), 'catalog stable fields differ'def ts_ignores_paging(call):@@ -196,8 +195,7 @@ def observe(binary, kind, root, observed, ts_cursor=None):'mtime_step_ms': 10000, 'session_archive_max_sessions': 'off','session_archive_max_age_days': 'off','expected_newest_first_ids': [row['id'] for row in expected],- 'old_search_target_id': expected[-1]['id'], 'name_search_target': TARGET,- 'assistant_wire_target': TRANSCRIPT_TARGET}+ 'old_search_target_id': expected[-1]['id'], 'name_search_target': TARGET}observed['queries'] = {}path = root / 'd.sock'scope = OwnedProcesses(binary, root, path)@@ -237,7 +235,6 @@ def observe(binary, kind, root, observed, ts_cursor=None):assert joined == expected, 'pages did not cover full catalog exactly once'assert TARGET not in json.dumps(first['final_data']['sessions'])assert rest['final_data']['sessions'][-1]['name'] == TARGET- assert TRANSCRIPT_TARGET in rest['final_data']['sessions'][-1]['allMessagesText']else:ts_ignores_paging(rest)boundary = expected[PAGE - 1]@@ -253,7 +250,6 @@ def observe(binary, kind, root, observed, ts_cursor=None):for rows in (observed['baseline_rows'], after['final_data']['sessions']):target, = [row for row in rows if row['id'] == expected[-1]['id']]assert target['name'] == TARGET- assert TRANSCRIPT_TARGET in target['allMessagesText']scope.discover()assert set(scope.targets) == set(scope.direct), 'catalog unexpectedly spawned a worker/kernel'observed['behavior_verified'] = Truediff --git a/scripts/test_review_saved_catalog_parity.py b/scripts/test_review_saved_catalog_parity.pyindex 6ac688c640..9d8a732c50 100644--- a/scripts/test_review_saved_catalog_parity.py+++ b/scripts/test_review_saved_catalog_parity.py@@ -4,8 +4,8 @@ from pathlib import Pathimport tempfileimport unittest-from review_saved_catalog_parity import (COUNT, PAGE, TARGET, TRANSCRIPT_TARGET,- normalize, same_rows, seed, ts_ignores_paging, validate_query)+from review_saved_catalog_parity import (COUNT, PAGE, TARGET, normalize,+ same_rows, seed, ts_ignores_paging, validate_query)class SavedCatalogGuards(unittest.TestCase):@@ -57,16 +57,14 @@ class SavedCatalogGuards(unittest.TestCase):self.assertEqual(entries[-1]['message']['role'], 'assistant')self.assertNotIn(TARGET, json.dumps(self.expected[:PAGE]))self.assertEqual(self.expected[-1]['name'], TARGET)- self.assertIn(TRANSCRIPT_TARGET, self.expected[-1]['allMessagesText'])- self.assertNotIn(TRANSCRIPT_TARGET, self.expected[-1]['firstMessage'])self.assertEqual(self.check(self.call()), self.expected)- def test_rejects_missing_duplicate_wrong_or_transcript_truncated_rows(self):+ def test_rejects_missing_duplicate_or_wrong_rows(self):for mutate in (lambda rows: rows.pop(),lambda rows: rows.append(copy.deepcopy(rows[0])),lambda rows: rows[0].update(name='wrong'),- lambda rows: rows[-1].update(allMessagesText=rows[-1]['firstMessage']),+ lambda rows: rows[-1].update(firstMessage='wrong'),):call = self.call()mutate(call['final_data']['sessions'])@@ -120,7 +118,7 @@ class SavedCatalogGuards(unittest.TestCase):call['raw_sent_hex'] = [b'{"id":"elsewhere","command":{}}\n'.hex()]elif mode == 'wrong_stream':call['frames'][0] = copy.deepcopy(call['frames'][0])- call['frames'][0]['session']['allMessagesText'] = 'lost transcript'+ call['frames'][0]['session']['firstMessage'] = 'wrong'elif mode == 'duplicate_stream':call['frames'].insert(0, copy.deepcopy(call['frames'][0]))elif mode == 'wrong_order':