Log task planner conversations locally

runtime/ai_planner.py

@@ -1,5 +1,6 @@
 from __future__ import annotations
 
+from collections.abc import Callable
 from typing import TYPE_CHECKING, Any
 
 from core.errors import TaskFailedError
@@ -23,9 +24,11 @@ class AIPlanner(Planner):
         *,
         client: ToolCallingClient | None = None,
         config: PlannerConfig | None = None,
+        event_logger: Callable[[dict[str, Any]], None] | None = None,
     ) -> None:
         self.config = config or load_config()
         self.client = client or build_client(self.config)
+        self.event_logger = event_logger
 
     def plan(
         self,
@@ -49,13 +52,24 @@ class AIPlanner(Planner):
                 "corrected action or finish the task if it cannot proceed.\n"
                 f"Previous failure: {context.step_results[-1].error or 'unknown error'}"
             )
-        decision = self.client.decide(
-            system_prompt=PLANNER_SYSTEM_PROMPT,
-            user_prompt=user_prompt,
-            screenshot=screenshot,
-            tools=ALL_TOOL_SPECS,
-            timeout=self.config.timeout,
-        )
+        self._log({"type": "llm_request", "task_id": context.task_id, "goal": goal,
+                   "system_prompt": PLANNER_SYSTEM_PROMPT, "user_prompt": user_prompt,
+                   "has_screenshot": screenshot is not None})
+        try:
+            decision = self.client.decide(
+                system_prompt=PLANNER_SYSTEM_PROMPT,
+                user_prompt=user_prompt,
+                screenshot=screenshot,
+                tools=ALL_TOOL_SPECS,
+                timeout=self.config.timeout,
+            )
+        except Exception as exc:
+            self._log({"type": "agent_error", "task_id": context.task_id, "error": str(exc)})
+            raise
+        self._log({"type": "llm_response", "task_id": context.task_id,
+                   "content": decision.text_output, "thinking": decision.thinking,
+                   "tool_name": decision.tool_name, "arguments": decision.arguments,
+                   "purpose": decision.purpose, "expected_outcome": decision.expected_outcome})
 
         if decision.tool_name == FINISH_TASK_TOOL:
             if decision.arguments.get("success"):
@@ -85,6 +99,13 @@ class AIPlanner(Planner):
         # (mapped to an empty plan above), never via this hook.
         return False
 
+    def _log(self, event: dict[str, Any]) -> None:
+        if self.event_logger is not None:
+            try:
+                self.event_logger(event)
+            except Exception:
+                pass
+
 
 def _history_summary(world: "WorldState | None") -> list[dict[str, Any]]:
     if world is None: