Focus investigation context and recall archived evidence on demand

tests/test_investigation_focus.py

@@ -0,0 +1,59 @@
+import copy
+import json
+from jsonschema import Draft202012Validator
+from test_investigation_consolidation import populated
+from viano_bug_analysis.analysis.investigation_view import working_view
+from viano_bug_analysis.analysis.investigation_focus import focused_view
+from viano_bug_analysis.analysis.investigation_blocks import restore_blocks, block_view
+from viano_bug_analysis.analysis.investigation_prompt import build_prompt, output_schema
+from viano_bug_analysis.analysis.model_context import model_json
+
+
+def fixture():
+    c = populated()
+    c['prompt_version'] = 'investigation_project_v8'
+    c['recalled_evidence'] = []
+    c['budget']['accounts']['H1']['actions'] = {}
+    return c
+
+
+def test_index_omits_old_payloads_but_preserves_predicates_and_scope():
+    c = fixture(); before = copy.deepcopy(c)
+    v = focused_view(c)
+    assert set(v['evidence']) == {'E000001'}
+    assert len(v['evidence_index']['rows']) == 2
+    row = v['evidence_index']['rows'][1]
+    assert row[0:2] == ['E000002','H2']
+    assert row[3][0]['query']['start_line'] == 41
+    assert row[3][0]['result_scope'] == {'coverage':'selected_lines','truncated':False}
+    assert 'code line 120' not in model_json(v)
+    assert c == before and focused_view(c) == v
+    assert set(v['hypotheses']['H2']) == {'proposal','confidence','disposition'}
+    assert set(v['hypotheses']['H2']['proposal']) == {'title','mechanism'}
+    assert len(model_json(v)) < len(model_json(block_view(c)))
+
+
+def test_recall_exact_contents_including_cross_hypothesis_counterevidence():
+    c = fixture(); c['recalled_evidence'] = ['E000002']
+    v = restore_blocks(focused_view(c))
+    assert v['evidence']['E000002']['content'] == working_view(c)['evidence']['E000002']['content']
+    c['recalled_evidence'] = []
+    assert 'E000002' not in focused_view(c)['evidence']
+
+
+def test_unassessed_other_hypothesis_result_stays_visible():
+    c = fixture()
+    c['budget']['accounts']['H2'] = {'actions':{'A1':{'status':'result_stored'}}}
+    assert set(focused_view(c)['evidence']) == {'E000001','E000002'}
+
+
+def test_recall_schema_is_bounded_and_legacy_prompt_is_unchanged():
+    c = fixture(); new = output_schema(c)
+    schema = new['properties']['next']
+    validator = Draft202012Validator(schema)
+    assert validator.is_valid({'kind':'recall','evidence_refs':['E000002']})
+    for refs in ([], ['E000002']*2, ['E000001','E000002','E000003','E000004','E000005']):
+        assert not validator.is_valid({'kind':'recall','evidence_refs':refs})
+    assert json.loads(build_prompt(c)['messages'][1]['content'])['input'] == focused_view(c)
+    c['prompt_version'] = 'investigation_project_v7'
+    assert not Draft202012Validator(output_schema(c)['properties']['next']).is_valid({'kind':'recall','evidence_refs':['E000001']})

tests/test_project_research.py

@@ -189,7 +189,8 @@ def test_case_builder_provider_outage_blocks_without_sealing(case_setup):
         assert reader.manifest["state"]=="BLOCKED" and reader.manifest["artifacts"].get("case") is None
 
 
-def test_model_loop_uses_project_tools_and_compact_previous_result(tool_case):
+@pytest.mark.parametrize("recall", [False, True, "unknown"])
+def test_model_loop_uses_project_tools_and_compact_previous_result(tool_case, recall):
     store=tool_case[0]
     with PackageReader(tool_case[1]) as reader:project=reader.manifest["project"]
     service=ProjectResearch(project,database=FakeDB([{"row":{"id":1}}]))
@@ -203,16 +204,37 @@ def test_model_loop_uses_project_tools_and_compact_previous_result(tool_case):
             if previous:
                 assert context["evidence"][eid]["content"]["items"]
             assert "inputs" not in context["case"]
-            assert request["prompt_version"]=="investigation_project_v7"
+            assert request["prompt_version"]=="investigation_project_v8"
             reply={"assessment":{"semantic_score":50,"reason":"Scoped fact","evidence_refs":[eid]} if previous else None,
                    "update":{"confidence":"plausible","evidence_refs":[eid] if previous else [],"resolve_questions":[],"add_questions":[],"note":"Query typed identity"},
                    "next":{"kind":"finish","reason":"Check complete"} if previous else {"kind":"tool","operation":"project.query","arguments":rows()}}
+            if recall:
+                if previous:
+                    reply['next'] = {'kind':'recall','evidence_refs':['E999999' if recall == 'unknown' else eid]}
+                elif any(e['record']['hypothesis_id'] == context['selected_hypothesis'] for e in context.get('evidence', {}).values()):
+                    recalled = next(eid for eid,e in context['evidence'].items() if e['record']['hypothesis_id'] == context['selected_hypothesis'])
+                    assert context['evidence'][recalled]['content']['items']
+                    reply['update']['evidence_refs'] = [recalled]
+                    reply['next'] = {'kind':'finish','reason':'Recalled exact evidence; uncertainty remains'}
             return ProviderResponse(reply,"fixture",100,10)
     adapter=Adapter()
     loop=Investigator(store,gateway(adapter),router=InvestigationToolRouter(project_research=service))
+    if recall is True:
+        partial=asyncio.run(loop.run("case-A","stable-package",max_steps=4))
+        assert partial.status == 'YIELDED'
+        loop=Investigator(store,gateway(adapter),router=InvestigationToolRouter(project_research=service))
     result=asyncio.run(loop.run("case-A","stable-package",max_steps=20))
-    assert result.status=="COMPLETED" and len(adapter.contexts)==6
+    if recall == 'unknown':
+        assert result.status == 'BLOCKED' and result.code == 'INVESTIGATION_EVIDENCE_UNKNOWN'
+    else:
+        assert result.status=="COMPLETED" and len(adapter.contexts)==(9 if recall else 6)
     assert len(EvidenceStore(store).snapshot("case-A","stable-package").record_refs)==3
+    accounts=BudgetManager(store).snapshot("case-A","stable-package").state['accounts']
+    assert sum(len(a['actions']) for a in accounts.values()) == 3
+    if recall == 'unknown':
+        assert any(a['status'] != 'settled' for a in accounts['H1']['actions'].values())
+    else:
+        assert all(a['status'] == 'settled' for account in accounts.values() for a in account['actions'].values())
 
 
 def test_provider_outage_blocks_without_ending_hypothesis(tool_case):