fix: close legacy quality and manifest compatibility gaps

Co-Authored-By: Codex <noreply@openai.com>

Author: Pigbibi

docs/advisory_contract.md

@@ -270,8 +270,9 @@ policy
 generated_at
 ```
 
-`source_artifacts` also records each existing input's SHA-256, point-in-time
-`as_of`/`generated_at`/`expires_at`, and schema/header metadata. Scheduled
+`source_artifacts_metadata` (schema version `1`) records each existing input's
+SHA-256, point-in-time `as_of`/`generated_at`/`expires_at`, and schema/header
+metadata while preserving the legacy `source_artifacts` string map. Scheduled
 workflows pass the checked-out commit SHA for each upstream repository in
 `upstream_repositories`; missing or stale context is excluded from scoring and
 reported in `summary.data_quality_warnings`.

src/quant_advisor_research/advisory_report.py

@@ -410,6 +410,7 @@ def load_market_confirmation(path: str | Path | None, as_of: dt.date) -> dict[st
         current = confirmations.get(symbol)
         if current and current.as_of and row_as_of and row_as_of < current.as_of:
             continue
+        quality_metadata_present = "confirmation_quality" in row and "warnings" in row
         confirmations[symbol] = MarketConfirmation(
             symbol=symbol,
             as_of=row_as_of,
@@ -424,7 +425,9 @@ def load_market_confirmation(path: str | Path | None, as_of: dt.date) -> dict[st
             market_score=as_float(row.get("market_score")) if str(row.get("market_score", "")).strip() else None,
             data_source=str(row.get("data_source", "")),
             price_observation_count=int(as_float(row.get("price_observation_count"))),
-            confirmation_quality=str(row.get("confirmation_quality", "")),
+            confirmation_quality=(str(row.get("confirmation_quality", "")).strip() or "missing_quality_metadata")
+            if quality_metadata_present
+            else "missing_quality_metadata",
             warnings=str(row.get("warnings", "")),
         )
     return confirmations
@@ -1021,7 +1024,7 @@ def market_confirmation_score(market: MarketConfirmation | None) -> float | None
         return None
     if market.data_source == "theme_momentum_fallback":
         return None
-    if market.confirmation_quality == "anomalous" or "extreme_return_63d" in market.warnings or "stale_price" in market.warnings:
+    if market.confirmation_quality != "price_observed":
         return None
     if market.market_score is not None:
         return round(clamp(market.market_score, 0, 1), 3)

src/quant_advisor_research/artifacts.py

@@ -62,7 +62,7 @@ def write_report_manifest(
     if run_attempt:
         version_parts.append(f"attempt-{run_attempt}")
 
-    source_artifacts = {}
+    source_artifact_metadata = {}
     for name, raw_path in (input_paths or {}).items():
         if raw_path is None:
             continue
@@ -90,7 +90,7 @@ def write_report_manifest(
                 )
             except (OSError, json.JSONDecodeError):
                 metadata["schema"] = "invalid_json"
-        source_artifacts[name] = metadata
+        source_artifact_metadata[name] = metadata
 
     payload = {
         "manifest_type": "model_recommendation_report",
@@ -109,7 +109,11 @@ def write_report_manifest(
             "github_run_id": run_id or "",
             "github_run_attempt": run_attempt or "",
         },
-        "source_artifacts": source_artifacts or dict(report.get("source_artifacts") or {}),
+        "source_artifacts": dict(report.get("source_artifacts") or {}),
+        "source_artifacts_metadata": {
+            "schema_version": "1",
+            "items": source_artifact_metadata,
+        },
         "upstream_repositories": dict(upstream_repo_shas or {}),
         "summary": dict(report.get("summary") or {}),
         "artifacts": {

tests/test_advisory_report.py

@@ -52,13 +52,37 @@ def test_expired_ai_signal_is_excluded_from_long_context_and_warns(tmp_path: Pat
     assert report["summary"]["long_context_available"] is False
 
 
+def test_legacy_market_confirmation_without_quality_metadata_cannot_score(tmp_path: Path) -> None:
+    market_path = tmp_path / "legacy_market.csv"
+    market_path.write_text(
+        "symbol,as_of,return_5d,return_20d,return_63d,relative_return_20d,relative_return_63d,volume_zscore,drawdown_63d,volatility_21d,market_score,data_source,price_observation_count\n"
+        "MU,2026-05-30,0.03,0.12,0.28,0.07,0.13,1.2,-0.06,0.33,0.99,yahoo_chart,120\n",
+        encoding="utf-8",
+    )
+    report = build_advisory_report(
+        as_of="2026-05-30",
+        cadence="weekly",
+        political_events_path=ROOT / "examples/political_events.example.csv",
+        political_watchlist_path=ROOT / "examples/political_watchlist.example.csv",
+        market_confirmation_path=market_path,
+    )
+
+    assert report["final_decisions"]["recommendations"] == []
+
+
 def test_manifest_records_input_hash_and_upstream_metadata(tmp_path: Path) -> None:
     report_path = tmp_path / "report.json"
     markdown_path = tmp_path / "report.md"
     report_path.write_text("{}\n", encoding="utf-8")
     markdown_path.write_text("# report\n", encoding="utf-8")
     input_path = ROOT / "examples/political_events.example.csv"
-    report = {"as_of": "2026-05-30", "cadence": "weekly", "schema_version": "5", "mode": "model_recommendations"}
+    report = {
+        "as_of": "2026-05-30",
+        "cadence": "weekly",
+        "schema_version": "5",
+        "mode": "model_recommendations",
+        "source_artifacts": {"political_events": str(input_path)},
+    }
 
     manifest_path = write_report_manifest(
         report=report,
@@ -71,8 +95,11 @@ def test_manifest_records_input_hash_and_upstream_metadata(tmp_path: Path) -> No
 
     manifest = json.loads(manifest_path.read_text(encoding="utf-8"))
     assert manifest["upstream_repositories"]["QuantStrategyLab/PoliticalEventTrackingResearch"] == "abc123"
-    assert manifest["source_artifacts"]["political_events"]["sha256"]
-    assert manifest["source_artifacts"]["political_events"]["schema"].startswith("event_id,")
+    assert manifest["source_artifacts"]["political_events"] == str(input_path)
+    metadata = manifest["source_artifacts_metadata"]
+    assert metadata["schema_version"] == "1"
+    assert metadata["items"]["political_events"]["sha256"]
+    assert metadata["items"]["political_events"]["schema"].startswith("event_id,")
 
 
 def test_low_confidence_events_remain_verify_source_until_verified() -> None: