fix: several issues related to scalability (#1619)

Co-authored-by: Xingyao Wang <[email protected]>
Co-authored-by: openhands <[email protected]>

Author: 3 people

openhands-sdk/openhands/sdk/agent/base.py

@@ -503,5 +503,5 @@ def tools_map(self) -> dict[str, ToolDefinition]:
             RuntimeError: If the agent has not been initialized.
         """
         if not self._initialized:
-            raise RuntimeError("Agent not initialized; call initialize() before use")
+            raise RuntimeError("Agent not initialized; call _initialize() before use")
         return self._tools

openhands-sdk/openhands/sdk/conversation/conversation.py

@@ -74,6 +74,7 @@ def __new__(
             type[ConversationVisualizerBase] | ConversationVisualizerBase | None
         ) = DefaultConversationVisualizer,
         secrets: dict[str, SecretValue] | dict[str, str] | None = None,
+        delete_on_close: bool = False,
     ) -> "LocalConversation": ...
 
     @overload
@@ -96,6 +97,7 @@ def __new__(
             type[ConversationVisualizerBase] | ConversationVisualizerBase | None
         ) = DefaultConversationVisualizer,
         secrets: dict[str, SecretValue] | dict[str, str] | None = None,
+        delete_on_close: bool = False,
     ) -> "RemoteConversation": ...
 
     def __new__(
@@ -118,6 +120,7 @@ def __new__(
             type[ConversationVisualizerBase] | ConversationVisualizerBase | None
         ) = DefaultConversationVisualizer,
         secrets: dict[str, SecretValue] | dict[str, str] | None = None,
+        delete_on_close: bool = False,
     ) -> BaseConversation:
         from openhands.sdk.conversation.impl.local_conversation import LocalConversation
         from openhands.sdk.conversation.impl.remote_conversation import (
@@ -143,6 +146,7 @@ def __new__(
                 visualizer=visualizer,
                 workspace=workspace,
                 secrets=secrets,
+                delete_on_close=delete_on_close,
             )
 
         return LocalConversation(
@@ -159,4 +163,5 @@ def __new__(
             workspace=workspace,
             persistence_dir=persistence_dir,
             secrets=secrets,
+            delete_on_close=delete_on_close,
         )

openhands-sdk/openhands/sdk/conversation/impl/local_conversation.py

@@ -65,6 +65,7 @@ class LocalConversation(BaseConversation):
     llm_registry: LLMRegistry
     _cleanup_initiated: bool
     _hook_processor: HookEventProcessor | None
+    delete_on_close: bool = True
     # Plugin lazy loading state
     _plugin_specs: list[PluginSource] | None
     _resolved_plugins: list[ResolvedPluginSource] | None
@@ -90,6 +91,7 @@ def __init__(
             type[ConversationVisualizerBase] | ConversationVisualizerBase | None
         ) = DefaultConversationVisualizer,
         secrets: Mapping[str, SecretValue] | None = None,
+        delete_on_close: bool = True,
         cipher: Cipher | None = None,
         **_: object,
     ):
@@ -242,6 +244,7 @@ def _default_callback(e):
 
         atexit.register(self.close)
         self._start_observability_span(str(desired_id))
+        self.delete_on_close = delete_on_close
 
     @property
     def id(self) -> ConversationID:
@@ -708,20 +711,23 @@ def close(self) -> None:
         except AttributeError:
             # Object may be partially constructed; span fields may be missing.
             pass
-        try:
-            tools_map = self.agent.tools_map
-        except (AttributeError, RuntimeError):
-            # Agent not initialized or partially constructed
-            return
-        for tool in tools_map.values():
+        if self.delete_on_close:
             try:
-                executable_tool = tool.as_executable()
-                executable_tool.executor.close()
-            except NotImplementedError:
-                # Tool has no executor, skip it without erroring
-                continue
-            except Exception as e:
-                logger.warning(f"Error closing executor for tool '{tool.name}': {e}")
+                tools_map = self.agent.tools_map
+            except (AttributeError, RuntimeError):
+                # Agent not initialized or partially constructed
+                return
+            for tool in tools_map.values():
+                try:
+                    executable_tool = tool.as_executable()
+                    executable_tool.executor.close()
+                except NotImplementedError:
+                    # Tool has no executor, skip it without erroring
+                    continue
+                except Exception as e:
+                    logger.warning(
+                        f"Error closing executor for tool '{tool.name}': {e}"
+                    )
 
     def ask_agent(self, question: str) -> str:
         """Ask the agent a simple, stateless question and get a direct LLM response.

openhands-sdk/openhands/sdk/conversation/impl/remote_conversation.py

@@ -555,6 +555,7 @@ class RemoteConversation(BaseConversation):
     _client: httpx.Client
     _hook_processor: HookEventProcessor | None
     _cleanup_initiated: bool
+    delete_on_close: bool = False
 
     def __init__(
         self,
@@ -573,6 +574,7 @@ def __init__(
             type[ConversationVisualizerBase] | ConversationVisualizerBase | None
         ) = DefaultConversationVisualizer,
         secrets: Mapping[str, SecretValue] | None = None,
+        delete_on_close: bool = False,
         **_: object,
     ) -> None:
         """Remote conversation proxy that talks to an agent server.
@@ -765,6 +767,7 @@ def __init__(
             )
             self._hook_processor = HookEventProcessor(hook_manager=hook_manager)
             self._hook_processor.run_session_start()
+        self.delete_on_close = delete_on_close
 
     def _create_llm_completion_log_callback(self) -> ConversationCallbackType:
         """Create a callback that writes LLM completion logs to client filesystem."""
@@ -1134,6 +1137,13 @@ def close(self) -> None:
             pass
 
         self._end_observability_span()
+        if self.delete_on_close:
+            try:
+                # trigger server-side delete_conversation to release resources
+                # like tmux sessions
+                _send_request(self._client, "DELETE", f"/api/conversations/{self.id}")
+            except Exception:
+                pass
 
     def __del__(self) -> None:
         try:

openhands-sdk/openhands/sdk/workspace/remote/base.py

@@ -50,12 +50,17 @@ def client(self) -> httpx.Client:
         if client is None:
             # Configure reasonable timeouts for HTTP requests
             # - connect: 10 seconds to establish connection
-            # - read: 60 seconds to read response (for LLM operations)
+            # - read: 600 seconds (10 minutes) to read response (for LLM operations)
             # - write: 10 seconds to send request
             # - pool: 10 seconds to get connection from pool
-            timeout = httpx.Timeout(connect=10.0, read=60.0, write=10.0, pool=10.0)
+            timeout = httpx.Timeout(
+                connect=10.0, read=self.read_timeout, write=10.0, pool=10.0
+            )
             client = httpx.Client(
-                base_url=self.host, timeout=timeout, headers=self._headers
+                base_url=self.host,
+                timeout=timeout,
+                headers=self._headers,
+                limits=httpx.Limits(max_connections=self.max_connections),
             )
             self._client = client
         return client

openhands-sdk/openhands/sdk/workspace/remote/remote_workspace_mixin.py

@@ -25,6 +25,15 @@ class RemoteWorkspaceMixin(BaseModel):
     working_dir: str = Field(
         description="The working directory for agent operations and tool execution."
     )
+    read_timeout: float = Field(
+        default=600.0,
+        description="Timeout in seconds for reading operations of httpx.Client.",
+    )
+    max_connections: int | None = Field(
+        default=None,
+        description="Maximum number of connections for httpx.Client. "
+        "None means no limit, useful for running many conversations in parallel.",
+    )
 
     def model_post_init(self, context: Any) -> None:
         # Set up remote host

openhands-workspace/openhands/workspace/docker/workspace.py

@@ -235,6 +235,8 @@ def _start_container(self, image: str, context: Any) -> None:
             "--platform",
             self.platform,
             "--rm",
+            "--ulimit",
+            "nofile=65536:65536",  # prevent "too many open files" errors
             "--name",
             f"agent-server-{uuid.uuid4()}",
             *flags,