session logging

agent/config.py

@@ -19,6 +19,8 @@ class Config(BaseModel):
 
     model_name: str
     mcpServers: dict[str, MCPServerConfig] = {}
+    save_sessions: bool = True
+    session_dataset_repo: str = "smolagents/hf-agent-sessions"
 
 
 def substitute_env_vars(obj: Any) -> Any:

agent/core/agent_loop.py

@@ -414,6 +414,16 @@ class Handlers:
     @staticmethod
     async def shutdown(session: Session) -> bool:
         """Handle shutdown (like shutdown in codex.rs:1329)"""
+        # Save session trajectory if enabled
+        if session.config.save_sessions:
+            print("💾 Saving session trajectory...")
+            repo_id = session.config.session_dataset_repo
+            url = await session.push_to_dataset(repo_id)
+            if url:
+                print(f"✅ Session saved to: {url}")
+            else:
+                print("❌ Failed to save session")
+
         session.is_running = False
         await session.send_event(Event(event_type="shutdown"))
         return True

agent/core/session.py

@@ -1,6 +1,8 @@
 import asyncio
+import json
 import uuid
 from dataclasses import dataclass
+from datetime import datetime
 from enum import Enum
 from typing import Any, Optional
 
@@ -55,11 +57,84 @@ class Session:
         self.current_task: asyncio.Task | None = None
         self.pending_approval: Optional[dict[str, Any]] = None
 
+        # Session trajectory logging
+        self.logged_events: list[dict] = []
+        self.session_start_time = datetime.now().isoformat()
+
     async def send_event(self, event: Event) -> None:
-        """Send event back to client"""
+        """Send event back to client and log to trajectory"""
         await self.event_queue.put(event)
 
+        # Log event to trajectory
+        self.logged_events.append(
+            {
+                "timestamp": datetime.now().isoformat(),
+                "event_type": event.event_type,
+                "data": event.data,
+            }
+        )
+
     def interrupt(self) -> None:
         """Interrupt current running task"""
         if self.current_task and not self.current_task.done():
             self.current_task.cancel()
+
+    def get_trajectory(self) -> dict:
+        """Serialize complete session trajectory for logging"""
+        return {
+            "session_id": self.session_id,
+            "session_start_time": self.session_start_time,
+            "session_end_time": datetime.now().isoformat(),
+            "model_name": self.config.model_name,
+            "messages": [msg.model_dump() for msg in self.context_manager.items],
+            "events": self.logged_events,
+        }
+
+    async def push_to_dataset(self, repo_id: str) -> Optional[str]:
+        """
+        Push session trajectory to Hugging Face dataset
+
+        Args:
+            repo_id: HuggingFace dataset repo ID (e.g. 'username/dataset-name')
+
+        Returns:
+            URL to the uploaded file if successful, None otherwise
+        """
+        try:
+            import os
+
+            from datasets import Dataset
+
+            # Get trajectory data
+            trajectory = self.get_trajectory()
+
+            # Convert to dataset row format
+            row = {
+                "session_id": trajectory["session_id"],
+                "session_start_time": trajectory["session_start_time"],
+                "session_end_time": trajectory["session_end_time"],
+                "model_name": trajectory["model_name"],
+                "messages": json.dumps(trajectory["messages"]),
+                "events": json.dumps(trajectory["events"]),
+            }
+
+            # Try to load existing dataset and append
+            try:
+                from datasets import load_dataset
+
+                existing_dataset = load_dataset(repo_id, split="train")
+                new_dataset = Dataset.from_dict(
+                    {k: list(existing_dataset[k]) + [v] for k, v in row.items()}
+                )
+            except Exception:
+                # Dataset doesn't exist yet, create new one
+                new_dataset = Dataset.from_dict({k: [v] for k, v in row.items()})
+
+            # Push to hub
+            new_dataset.push_to_hub(repo_id, private=True, token=os.getenv("HF_TOKEN"))
+
+            return f"https://huggingface.co/datasets/{repo_id}"
+
+        except Exception as e:
+            print(f"Failed to push session to dataset: {e}")
+            return None

agent/main.py

@@ -415,8 +415,8 @@ async def main():
     )
     await submission_queue.put(shutdown_submission)
 
-    # Wait for tasks to complete
-    await asyncio.wait_for(agent_task, timeout=2.0)
+    # Wait for tasks to complete (longer timeout to allow for session save)
+    await asyncio.wait_for(agent_task, timeout=30.0)
     listener_task.cancel()
 
     print("✨ Goodbye!\n")

configs/main_agent_config.json

@@ -1,5 +1,7 @@
 {
-  "model_name": "anthropic/claude-sonnet-4-5-20250929",
+  "model_name": "anthropic/claude-opus-4-5-20251101",
+  "save_sessions": true,
+  "session_dataset_repo": "smolagents/hf-agent-sessions",
   "mcpServers": {
     "hf-mcp-server": {
       "transport": "http",

pyproject.toml

@@ -5,6 +5,7 @@ description = "Add your description here"
 readme = "README.md"
 requires-python = ">=3.12"
 dependencies = [
+    "datasets>=4.4.1",
     # Core dependencies (always required)
     "pydantic>=2.12.3",
     "python-dotenv>=1.2.1",
@@ -22,6 +23,7 @@ agent = [
     "thefuzz>=0.22.1",
     "nbconvert>=7.16.6",
     "nbformat>=5.10.4",
+    "datasets>=4.3.0",  # For session logging to HF datasets
 ]
 
 # Evaluation/benchmarking dependencies

uv.lock

@@ -902,12 +902,14 @@ name = "hf-agent"
 version = "0.1.0"
 source = { virtual = "." }
 dependencies = [
+    { name = "datasets" },
     { name = "pydantic" },
     { name = "python-dotenv" },
 ]
 
 [package.optional-dependencies]
 agent = [
+    { name = "datasets" },
     { name = "fastmcp" },
     { name = "huggingface-hub" },
     { name = "litellm" },
@@ -946,6 +948,8 @@ eval = [
 
 [package.metadata]
 requires-dist = [
+    { name = "datasets", specifier = ">=4.4.1" },
+    { name = "datasets", marker = "extra == 'agent'", specifier = ">=4.3.0" },
     { name = "datasets", marker = "extra == 'eval'", specifier = ">=4.3.0" },
     { name = "fastmcp", marker = "extra == 'agent'", specifier = ">=2.4.0" },
     { name = "hf-agent", extras = ["agent", "eval", "dev"], marker = "extra == 'all'" },