Fix coverage crash on cost reports in reports/

code_tests/unit_tests/test_agents_and_tools/test_source_archive/test_reports.py

@@ -3,6 +3,10 @@
 import json
 
 from forecasting_tools.agents_and_tools.source_archive.config import ArchiveConfig
+from forecasting_tools.agents_and_tools.source_archive.cost import (
+    RunCost,
+    write_cost_report,
+)
 from forecasting_tools.agents_and_tools.source_archive.models import (
     StoredCapture,
     url_hash,
@@ -87,3 +91,22 @@ def test_captured_status_wins_over_failure(tmp_path):
         config,
     )
     assert read_outcomes(store, config)["https://a.test"] == "stored"
+
+
+def test_read_outcomes_ignores_cost_reports(tmp_path):
+    """Cost reports live under reports/ too (``<run_id>_cost.json``, a JSON
+    dict, not a list of rows) — read_outcomes must skip them, not crash."""
+    store = LocalBlobStore(tmp_path)
+    config = ArchiveConfig(s3_prefix="t")
+    write_run_report(
+        store,
+        "r1",
+        PipelineSummary(
+            outcomes=[CaptureOutcome(url="https://a.test", status="stored")]
+        ),
+        config,
+    )
+    write_cost_report(store, "r1", RunCost(run_id="r1", archived=1), config)
+
+    out = read_outcomes(store, config)
+    assert out == {"https://a.test": "stored"}

forecasting_tools/agents_and_tools/source_archive/reports.py

@@ -73,12 +73,16 @@ def read_outcomes(store: BlobStore, config: ArchiveConfig) -> dict[str, str]:
     prefix = config.s3_prefix.rstrip("/")
     out: dict[str, str] = {}
     for key in store.list_keys(f"{prefix}/reports/"):
-        if not key.endswith(".json"):
+        # reports/ also holds per-run cost breakdowns (``<run_id>_cost.json``,
+        # a dict) — only run reports (a list of outcome rows) belong here.
+        if not key.endswith(".json") or key.endswith("_cost.json"):
             continue
         try:
             rows = json.loads(store.get(key).decode("utf-8"))
         except (UnicodeDecodeError, ValueError):
             continue
+        if not isinstance(rows, list):
+            continue
         for r in rows:
             url = canonicalize_url(r.get("url", ""))
             status = r.get("status", "")