fix(livetest): render multi-query bridge calls

scripts/analyze_livetest.py

@@ -43,7 +43,7 @@ def fmt_bridge_seq(calls):
             if isinstance(qs, list):
                 q = "; ".join(str(x) for x in qs)
             else:  # legacy single-query transcripts
-                q = str(args.get("query", "?"))
+                q = str(args["query"] if "query" in args else "?")
             parts.append(f"search('{q[:30]}')")
         elif c["name"] == "tool_describe":
             args = c.get("args") or {}

scripts/tool_search_livetest.py

@@ -287,7 +287,7 @@ def setup_isolated_home(enabled: bool, listing: str = "off",
                 "enabled": "on" if enabled else "off",
                 "threshold_pct": 10,
                 "search_default_limit": 5,
-                "max_search_limit": 20,
+                "max_search_limit": 25,
                 "listing": listing,
                 "listing_max_tokens": listing_max_tokens,
             },

scripts/tool_search_livetest_ue_disc.py

@@ -115,6 +115,15 @@ def score_survey(resp: str, truth: Dict[str, bool]) -> bool:
     return True
 
 
+def _bridge_query_text(call: Dict[str, Any]) -> str:
+    """Render current multi-query calls and legacy saved transcript calls."""
+    args = call.get("args") or {}
+    queries = args.get("queries")
+    if isinstance(queries, list):
+        return "; ".join(str(query) for query in queries)
+    return str(args["query"] if "query" in args else "?")
+
+
 def run_one(scenario, mode, rep, out_dir: Path):
     model = os.environ.get("TS_UE_MODEL", "anthropic/claude-opus-4.8")
     lmax = int(os.environ.get("TS_UE_LISTING_MAX", "30000"))
@@ -203,7 +212,11 @@ def run_one(scenario, mode, rep, out_dir: Path):
         "prompt_tokens_total": sum(u["prompt_tokens"] or 0 for u in usage_log),
         "ue_calls": [c[-60:] for c in ue_calls][:15],
         "write_calls": [c[-60:] for c in write_calls][:10],
-        "bridge_queries": [(b.get("args") or {}).get("query") for b in bridge_call_log if b["name"] == "tool_search"][:10],
+        "bridge_queries": [
+            _bridge_query_text(call)
+            for call in bridge_call_log
+            if call["name"] == "tool_search"
+        ][:10],
         "success": bool(success), "error": error,
         "final_response": base._redact_secrets(final_response)[:400],
     }

scripts/tool_search_livetest_ue_hard.py

@@ -36,6 +36,22 @@ from tool_search_livetest_ue import load_epic_tools, _SANITIZE  # reuse loader
 
 N_REPS = int(os.environ.get("TS_BENCH_REPS", "2"))
 
+
+def _bridge_call_value(call: Dict[str, Any]) -> Any:
+    """Summarize current batch arguments with legacy transcript fallbacks."""
+    args = call.get("args") or {}
+    if call["name"] == "tool_search":
+        queries = args.get("queries")
+        if isinstance(queries, list):
+            return "; ".join(str(query) for query in queries)
+        return args["query"] if "query" in args else None
+    if call["name"] == "tool_describe":
+        names = args.get("names")
+        if isinstance(names, list):
+            return ", ".join(str(name) for name in names)
+    return args.get("name")
+
+
 # ---------------------------------------------------------------------------
 # Type-aware mock world
 # ---------------------------------------------------------------------------
@@ -276,7 +292,9 @@ def run_one(scenario, mode, rep, out_dir: Path):
         "first_correct": first_correct, "final_correct": final_correct,
         "wrong_calls": wrong_calls, "success": bool(success),
         "ue_calls": [c["name"][-70:] for c in ue_calls][:20],
-        "bridge_calls": [(b["name"], (b.get("args") or {}).get("query") or (b.get("args") or {}).get("name")) for b in bridge_call_log][:20],
+        "bridge_calls": [
+            (call["name"], _bridge_call_value(call)) for call in bridge_call_log
+        ][:20],
         "error": error,
         "final_response": base._redact_secrets(final_response)[:300],
     }