Merge: WebResearchEnv evaluate() with full agent loop + tools

environments/web_research_env.py

@@ -425,8 +425,16 @@ class WebResearchEnv(HermesAgentBaseEnv):
     # ------------------------------------------------------------------
 
     async def evaluate(self, *args, **kwargs) -> None:
-        """Run evaluation on the held-out split using the agent loop."""
+        """Run evaluation on the held-out split using the full agent loop with tools.
+
+        Each eval item runs through the same agent loop as training —
+        the model can use web_search, web_extract, etc. to research answers.
+        This measures actual agentic research capability, not just knowledge.
+        """
         import time
+        import uuid
+        from environments.agent_loop import HermesAgentLoop
+        from environments.tool_context import ToolContext
 
         items = self._eval_items
         if not items:
@@ -436,43 +444,75 @@ class WebResearchEnv(HermesAgentBaseEnv):
         eval_size = min(self.config.eval_size, len(items))
         eval_items = items[:eval_size]
 
-        logger.info(f"Running eval on {len(eval_items)} questions...")
+        logger.info(f"Running eval on {len(eval_items)} questions (with agent loop + tools)...")
         start_time = time.time()
         samples = []
 
-        for item in eval_items:
+        # Resolve tools once for all eval items
+        tools, valid_names = self._resolve_tools_for_group()
+
+        for i, item in enumerate(eval_items):
+            task_id = str(uuid.uuid4())
+            logger.info(f"Eval [{i+1}/{len(eval_items)}]: {item['question'][:80]}...")
+
             try:
-                # Use the base env's agent loop for eval (same as training)
+                # Build messages
-                prompt = self.format_prompt(item)
+                messages: List[Dict[str, Any]] = []
-                completion = await self.server.chat_completion(
+                if self.config.system_prompt:
-                    messages=[
+                    messages.append({"role": "system", "content": self.config.system_prompt})
-                        {"role": "system", "content": self.config.system_prompt or ""},
+                messages.append({"role": "user", "content": self.format_prompt(item)})
-                        {"role": "user", "content": prompt},
+
-                    ],
+                # Run the full agent loop with tools
-                    n=1,
+                agent = HermesAgentLoop(
+                    server=self.server,
+                    tool_schemas=tools,
+                    valid_tool_names=valid_names,
+                    max_turns=self.config.max_agent_turns,
+                    task_id=task_id,
+                    temperature=0.0,  # Deterministic for eval
                     max_tokens=self.config.max_token_length,
-                    temperature=0.0,
+                    extra_body=self.config.extra_body,
-                    split="eval",
                 )
+                result = await agent.run(messages)
 
-                response_content = (
+                # Extract final response and compute reward
-                    completion.choices[0].message.content if completion.choices else ""
+                ctx = ToolContext(task_id)
-                )
+                try:
+                    reward = await self.compute_reward(item, result, ctx)
+                finally:
+                    ctx.cleanup()
 
-                # Score the response
+                # Extract final response for logging
+                final_response = ""
+                tool_call_count = 0
+                for msg in reversed(result.messages):
+                    if msg.get("role") == "assistant" and msg.get("content") and not final_response:
+                        final_response = msg["content"]
+                    if msg.get("role") == "assistant" and msg.get("tool_calls"):
+                        tool_call_count += len(msg["tool_calls"])
+
+                # Score correctness separately for the metric
                 correctness = await self._llm_judge(
                     question=item["question"],
                     expected=item["answer"],
-                    model_answer=response_content,
+                    model_answer=final_response,
                 )
 
                 samples.append({
                     "prompt": item["question"],
-                    "response": response_content,
+                    "response": final_response[:500],
                     "expected": item["answer"],
                     "correctness": correctness,
+                    "reward": reward,
+                    "tool_calls": tool_call_count,
+                    "turns": result.turns_used,
                 })
 
+                logger.info(
+                    f"  → correctness={correctness:.2f}, reward={reward:.3f}, "
+                    f"tools={tool_call_count}, turns={result.turns_used}"
+                )
+
             except Exception as e:
                 logger.error(f"Eval error on item: {e}")
                 samples.append({
@@ -480,20 +520,33 @@ class WebResearchEnv(HermesAgentBaseEnv):
                     "response": f"ERROR: {e}",
                     "expected": item["answer"],
                     "correctness": 0.0,
+                    "reward": 0.0,
+                    "tool_calls": 0,
+                    "turns": 0,
                 })
 
         end_time = time.time()
 
-        # Compute metrics
+        # Compute aggregate metrics
         correctness_scores = [s["correctness"] for s in samples]
+        rewards = [s["reward"] for s in samples]
+        tool_counts = [s["tool_calls"] for s in samples]
+        n = len(samples)
+
         eval_metrics = {
-            "eval/mean_correctness": (
+            "eval/mean_correctness": sum(correctness_scores) / n if n else 0.0,
-                sum(correctness_scores) / len(correctness_scores)
+            "eval/mean_reward": sum(rewards) / n if n else 0.0,
-                if correctness_scores else 0.0
+            "eval/mean_tool_calls": sum(tool_counts) / n if n else 0.0,
-            ),
+            "eval/tool_usage_rate": sum(1 for t in tool_counts if t > 0) / n if n else 0.0,
-            "eval/n_items": len(samples),
+            "eval/n_items": n,
         }
 
+        logger.info(
+            f"Eval complete — correctness={eval_metrics['eval/mean_correctness']:.3f}, "
+            f"reward={eval_metrics['eval/mean_reward']:.3f}, "
+            f"tool_usage={eval_metrics['eval/tool_usage_rate']:.0%}"
+        )
+
         await self.evaluate_log(
             metrics=eval_metrics,
             samples=samples,