Merge remote-tracking branch 'upstream/main'

Development.md

@@ -5,12 +5,14 @@ Otherwise, you can clone the OpenHands project directly.
 
 ## Start the server for development
 ### 1. Requirements
-* Linux, Mac OS, or [WSL on Windows](https://learn.microsoft.com/en-us/windows/wsl/install)  [ Ubuntu <= 22.04]
+* Linux, Mac OS, or [WSL on Windows](https://learn.microsoft.com/en-us/windows/wsl/install)  [Ubuntu <= 22.04]
 * [Docker](https://docs.docker.com/engine/install/) (For those on MacOS, make sure to allow the default Docker socket to be used from advanced settings!)
 * [Python](https://www.python.org/downloads/) = 3.12
 * [NodeJS](https://nodejs.org/en/download/package-manager) >= 18.17.1
 * [Poetry](https://python-poetry.org/docs/#installing-with-the-official-installer) >= 1.8
-* netcat => sudo apt-get install netcat
+* OS-specific dependencies:
+  - Ubuntu: build-essential => `sudo apt-get install build-essential`
+  - WSL: netcat => `sudo apt-get install netcat`
 
 Make sure you have all these dependencies installed before moving on to `make build`.
 

openhands/controller/agent_controller.py

@@ -254,11 +254,11 @@ async def _handle_observation(self, observation: Observation):
             if self.state.agent_state == AgentState.ERROR:
                 self.state.metrics.merge(self.state.local_metrics)
         elif isinstance(observation, FatalErrorObservation):
-            await self.report_error(
-                'There was a fatal error during agent execution: ' + str(observation)
+            self.state.last_error = (
+                f'There was a fatal error during agent execution: {str(observation)}'
             )
-            await self.set_agent_state_to(AgentState.ERROR)
             self.state.metrics.merge(self.state.local_metrics)
+            await self.set_agent_state_to(AgentState.ERROR)
 
     async def _handle_message_action(self, action: MessageAction):
         """Handles message actions from the event stream.
@@ -406,6 +406,14 @@ async def _step(self) -> None:
             await asyncio.sleep(1)
             return
 
+        # check if agent got stuck before taking any action
+        if self._is_stuck():
+            # This need to go BEFORE report_error to sync metrics
+            self.event_stream.add_event(
+                FatalErrorObservation('Agent got stuck in a loop'), EventSource.USER
+            )
+            return
+
         if self.delegate is not None:
             assert self.delegate != self
             if self.delegate.get_agent_state() == AgentState.PAUSED:
@@ -467,11 +475,6 @@ async def _step(self) -> None:
         await self.update_state_after_step()
         logger.info(action, extra={'msg_type': 'ACTION'})
 
-        if self._is_stuck():
-            # This need to go BEFORE report_error to sync metrics
-            await self.set_agent_state_to(AgentState.ERROR)
-            await self.report_error('Agent got stuck in a loop')
-
     async def _delegate_step(self):
         """Executes a single step of the delegate agent."""
         logger.debug(f'[Agent Controller {self.id}] Delegate not none, awaiting...')

openhands/events/stream.py

@@ -1,5 +1,6 @@
 import asyncio
 import threading
+from dataclasses import dataclass, field
 from datetime import datetime
 from enum import Enum
 from typing import Callable, Iterable
@@ -29,24 +30,17 @@ def session_exists(sid: str, file_store: FileStore) -> bool:
         return False
 
 
+@dataclass
 class EventStream:
     sid: str
     file_store: FileStore
     # For each subscriber ID, there is a stack of callback functions - useful
     # when there are agent delegates
-    _subscribers: dict[str, list[Callable]]
-    _cur_id: int
-    _lock: threading.Lock
-
-    def __init__(self, sid: str, file_store: FileStore):
-        self.sid = sid
-        self.file_store = file_store
-        self._subscribers = {}
-        self._cur_id = 0
-        self._lock = threading.Lock()
-        self._reinitialize_from_file_store()
+    _subscribers: dict[str, list[Callable]] = field(default_factory=dict)
+    _cur_id: int = 0
+    _lock: threading.Lock = field(default_factory=threading.Lock)
 
-    def _reinitialize_from_file_store(self) -> None:
+    def __post_init__(self) -> None:
         try:
             events = self.file_store.list(f'sessions/{self.sid}/events')
         except FileNotFoundError:
@@ -173,4 +167,4 @@ def clear(self):
         self.file_store.delete(f'sessions/{self.sid}')
         self._cur_id = 0
         # self._subscribers = {}
-        self._reinitialize_from_file_store()
+        self.__post_init__()

openhands/runtime/builder/docker.py

@@ -16,7 +16,7 @@ def __init__(self, docker_client: docker.DockerClient):
         self.docker_client = docker_client
 
         version_info = self.docker_client.version()
-        server_version = version_info.get('Version', '')
+        server_version = version_info.get('Version', '').replace('-', '.')
         if tuple(map(int, server_version.split('.'))) < (18, 9):
             raise RuntimeError('Docker server version must be >= 18.09 to use BuildKit')
 
@@ -53,7 +53,7 @@ def build(
         """
         self.docker_client = docker.from_env()
         version_info = self.docker_client.version()
-        server_version = version_info.get('Version', '')
+        server_version = version_info.get('Version', '').replace('-', '.')
         if tuple(map(int, server_version.split('.'))) < (18, 9):
             raise RuntimeError('Docker server version must be >= 18.09 to use BuildKit')