Update app.py

app.py

@@ -150,11 +150,13 @@ def eval_latest(run_dir_text):
         f"--output_dir='{eval_out_dir}'"
     )
     rc, tail = _run(cmd, elog)
-        # --- begin optional patch: extract metrics from log tail ---
-    import re, ast
+
+    # --- optional patch: parse the printed dict and write metrics.json ---
+    import re, ast, json, pathlib
     metrics_txt = "(metrics.json not found)"
     p = pathlib.Path(eval_out_dir) / "metrics.json"
-    # try to parse the last Python-dict-looking block from the log tail
+
+    # Try to parse the last dict-like summary from the log tail
     m = re.findall(r"\{[^}]+pc_success[^}]+\}", tail, flags=re.S)
     if m:
         try:
@@ -169,19 +171,18 @@ def eval_latest(run_dir_text):
             metrics_txt = f"Success rate: {out['success_rate']}\nAvg max overlap: {out['avg_max_overlap']}"
         except Exception:
             pass
-    # --- end optional patch ---
-    
-    metrics_txt = "(metrics.json not found)"
-    p = pathlib.Path(eval_out_dir) / "metrics.json"
-    if p.exists():
+    elif p.exists():
+        # Fallback: if a previous metrics.json exists, show it
         try:
-            m = json.loads(p.read_text())
-            metrics_txt = f"Success rate: {m.get('success_rate')}\nAvg max overlap: {m.get('avg_max_overlap')}"
+            d = json.loads(p.read_text())
+            metrics_txt = f"Success rate: {d.get('success_rate')}\nAvg max overlap: {d.get('avg_max_overlap')}"
         except Exception:
             metrics_txt = "(could not parse metrics.json)"
+    # --- end patch ---
+
     msg = f"Evaluated run at: {run_dir}\nEval exited rc={rc}\n\n=== eval.log tail ===\n{tail}"
     return msg, run_dir, tail_file(elog), metrics_txt
-
+    
 # ---------- Maintenance (list / delete runs) ----------
 def list_runs():
     root = pathlib.Path(RUN_ROOT)