fix: resolve mypy type errors in generate command and test generation

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: hidai25

evalview/commands/generate_cmd.py

@@ -110,7 +110,7 @@ def generate(
 
         generator = AgentTestGenerator(
             adapter=adapter,
-            endpoint=endpoint,
+            endpoint=endpoint or "",
             adapter_type=resolved_adapter,
             include_tools=included,
             exclude_tools=excluded,
@@ -121,7 +121,7 @@ def generate(
     else:
         result = run_generation(
             adapter=adapter,
-            endpoint=endpoint,
+            endpoint=endpoint or "",
             adapter_type=resolved_adapter,
             budget=budget,
             seed_prompts=seed_prompts,
@@ -140,7 +140,7 @@ def generate(
 
     generator = AgentTestGenerator(
         adapter=adapter,
-        endpoint=endpoint,
+        endpoint=endpoint or "",
         adapter_type=resolved_adapter,
         include_tools=included,
         exclude_tools=excluded,

evalview/test_generation.py

@@ -173,6 +173,8 @@ async def generate(
             seen_queries.add(query)
 
             try:
+                if self.adapter is None:
+                    raise RuntimeError("No adapter configured for live probing")
                 trace = await self.adapter.execute(query)
             except Exception as exc:
                 failures.append(f"{query[:80]}: {exc}")
@@ -465,7 +467,7 @@ def _build_test_case(self, probe: ProbeResult, clustered: Dict[str, ProbeResult]
             expected.output = ExpectedOutput(not_contains=["error", "traceback"])
 
         confidence = self._confidence_for_probe(probe)
-        dangerous_tools = self._infer_forbidden_tools(probe, clustered.values())
+        dangerous_tools = self._infer_forbidden_tools(probe, list(clustered.values()))
         if dangerous_tools:
             expected.forbidden_tools = dangerous_tools