feat(agent): add task-oriented run logs and execution details

tests/unit_tests/agent/test_run_journal.py

@@ -0,0 +1,58 @@
+"""Agent monitoring snapshots retain useful input without duplicating binary payloads."""
+
+from types import SimpleNamespace
+from unittest.mock import AsyncMock
+
+import pytest
+
+from langbot.pkg.agent.runner.run_journal import AgentRunJournal
+
+
+@pytest.mark.asyncio
+@pytest.mark.parametrize('event_type', ['message.received', 'group.member_joined'])
+async def test_agent_run_snapshot_records_identity_and_redacts_inline_attachments(event_type):
+    journal = AgentRunJournal(SimpleNamespace())
+    ledger = SimpleNamespace(create_run=AsyncMock(return_value={}))
+    journal._run_ledger_store = ledger
+    raw_input = {
+        'text': 'Describe the attached image',
+        'contents': [{'type': 'image_base64', 'image_base64': 'large-image-data'}],
+        'attachments': [{'name': 'notes.txt', 'content': 'large-file-data'}],
+    }
+    event = SimpleNamespace(
+        event_id='event-1',
+        conversation_id='conversation-1',
+        thread_id=None,
+        workspace_id='workspace-1',
+        bot_id='bot-1',
+        event_type=event_type,
+        data={'user_id': 'user-1', 'group_id': 'group-1'},
+        source='platform',
+        delivery=SimpleNamespace(model_dump=lambda **kwargs: {'target_id': 'group-1'}),
+    )
+    binding = SimpleNamespace(
+        binding_id='agent_agent-1_runner-1',
+        agent_id='agent-1',
+        processor_id='agent-1',
+        processor_type='agent',
+    )
+    await journal.create_run(
+        event=event,
+        binding=binding,
+        descriptor=SimpleNamespace(id='runner-1'),
+        context={'run_id': 'run-1', 'input': raw_input},
+        authorization={},
+    )
+    saved = ledger.create_run.call_args.kwargs
+    assert saved['agent_id'] == 'agent-1'
+    assert saved['metadata']['input']['text'] == raw_input['text']
+    assert saved['metadata']['input']['contents'][0]['image_base64'] is None
+    assert saved['metadata']['input']['attachments'][0]['content'] is None
+    assert saved['metadata']['delivery'] == {'target_id': 'group-1'}
+    assert raw_input['contents'][0]['image_base64'] == 'large-image-data'
+    assert raw_input['attachments'][0]['content'] == 'large-file-data'
+
+    if event_type != 'message.received':
+        assert saved['metadata']['input_event'] == event.data
+    else:
+        assert 'input_event' not in saved['metadata']

tests/unit_tests/agent/test_run_ledger_store.py

@@ -472,3 +472,30 @@ async def test_run_lifecycle_retains_milliseconds(store, monkeypatch):
     await store.finalize_run(run_id='run-ms', status='completed')
     saved = await store.get_run('run-ms')
     assert saved['finished_at_ms'] - saved['started_at_ms'] == 275
+
+
+@pytest.mark.asyncio
+async def test_agent_run_pages_include_old_bindings_without_leaking_other_agents(store):
+    for run_id, agent_id, binding, workspace in [
+        ('old', None, 'agent_one_plugin:a/old/default', 'w'),
+        ('debug', None, 'debug:one:plugin:a/new/default', 'w'),
+        ('new', 'one', 'new-binding', 'w'),
+        ('other', 'two', 'agent_two_plugin:a/old/default', 'w'),
+        ('conflicting-id', 'two', 'agent_one_plugin:a/old/default', 'w'),
+        ('other-workspace', 'one', 'agent_one_plugin:a/old/default', 'other'),
+        ('plugin-processor', None, 'event_processor:one', 'w'),
+    ]:
+        await store.create_run(
+            run_id=run_id,
+            event_id=None,
+            agent_id=agent_id,
+            binding_id=binding,
+            workspace_id=workspace,
+            runner_id='runner',
+        )
+    page, cursor, more, total = await store.list_runs(agent_id='one', workspace_id='w', limit=2)
+    assert [run['run_id'] for run in page] == ['new', 'debug']
+    assert more and total == 3
+    page, cursor, more, total = await store.list_runs(agent_id='one', workspace_id='w', before_id=cursor, limit=2)
+    assert [run['run_id'] for run in page] == ['old']
+    assert not more and total == 3

tests/unit_tests/api/service/test_agent_service.py

@@ -905,3 +905,44 @@ async def test_processor_trace_rejects_other_workspace_or_instance(monkeypatch,
     with pytest.raises(ValueError, match='Processor run not found'):
         await service.get_processor_run_events(WORKSPACE_UUID, 'one', 'run')
     store.page_run_events.assert_not_called()
+
+
+async def test_agent_runs_use_workspace_and_stable_agent_identity(monkeypatch):
+    service = AgentService(_make_app())
+    service.get_agent = AsyncMock(return_value={'kind': 'agent'})
+    service.ap.persistence_mgr.get_db_engine = Mock()
+    store = SimpleNamespace(list_runs=AsyncMock(return_value=([], None, False, 0)))
+    monkeypatch.setattr('langbot.pkg.agent.runner.run_ledger_store.RunLedgerStore', Mock(return_value=store))
+    assert (await service.get_processor_runs(WORKSPACE_UUID, 'one', before_id=12))['total'] == 0
+    store.list_runs.assert_awaited_once_with(workspace_id=WORKSPACE_UUID, agent_id='one', before_id=12)
+
+
+@pytest.mark.parametrize(
+    'binding,agent_id,workspace,allowed',
+    [
+        ('agent_one_plugin:a/old/default', None, WORKSPACE_UUID, True),
+        ('debug:one:plugin:a/new/default', None, WORKSPACE_UUID, True),
+        ('any-new-binding', 'one', WORKSPACE_UUID, True),
+        ('agent_two_plugin:a/old/default', None, WORKSPACE_UUID, False),
+        ('debug:two:plugin:a/new/default', None, WORKSPACE_UUID, False),
+        ('agent_one_plugin:a/old/default', 'two', WORKSPACE_UUID, False),
+        ('agent_one_plugin:a/old/default', None, 'other-workspace', False),
+        ('event_processor:one', None, WORKSPACE_UUID, False),
+    ],
+)
+async def test_agent_trace_authorizes_history_across_runner_changes(monkeypatch, binding, agent_id, workspace, allowed):
+    service = AgentService(_make_app())
+    service.get_agent = AsyncMock(return_value={'kind': 'agent'})
+    service.ap.persistence_mgr.get_db_engine = Mock()
+    store = SimpleNamespace(
+        get_run=AsyncMock(return_value={'workspace_id': workspace, 'binding_id': binding, 'agent_id': agent_id}),
+        page_run_events=AsyncMock(return_value=([], None, None, False)),
+    )
+    monkeypatch.setattr('langbot.pkg.agent.runner.run_ledger_store.RunLedgerStore', Mock(return_value=store))
+    if allowed:
+        await service.get_processor_run_events(WORKSPACE_UUID, 'one', 'run', after_sequence=100)
+        store.page_run_events.assert_awaited_once_with(run_id='run', after_sequence=100, limit=100)
+    else:
+        with pytest.raises(ValueError, match='Processor run not found'):
+            await service.get_processor_run_events(WORKSPACE_UUID, 'one', 'run')
+        store.page_run_events.assert_not_called()