agentscope-runtime 1.0.0b1__tar.gz → 1.0.1__tar.gz

{agentscope_runtime-1.0.0b1/src/agentscope_runtime.egg-info → agentscope_runtime-1.0.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: agentscope-runtime
-Version: 1.0.0b1
+Version: 1.0.1
 Summary: A production-ready runtime framework for agent applications, providing secure sandboxed execution environments and scalable deployment solutions with multi-framework support.
 Requires-Python: >=3.10
 Description-Content-Type: text/markdown
@@ -11,7 +11,7 @@ Requires-Dist: uvicorn[standard]>=0.24.0
 Requires-Dist: openai
 Requires-Dist: pydantic>=2.11.7
 Requires-Dist: requests>=2.32.4
-Requires-Dist: agentscope>=1.0.6
+Requires-Dist: agentscope>=1.0.9
 Requires-Dist: docker>=7.1.0
 Requires-Dist: redis>=6.0.0
 Requires-Dist: oss2>=2.19.1
@@ -301,9 +301,11 @@ You’ll see output streamed in **Server-Sent Events (SSE)** format:
 ```bash
 data: {"sequence_number":0,"object":"response","status":"created", ... }
 data: {"sequence_number":1,"object":"response","status":"in_progress", ... }
-data: {"sequence_number":2,"object":"content","status":"in_progress","text":"The" }
-data: {"sequence_number":3,"object":"content","status":"in_progress","text":" capital of France is Paris." }
-data: {"sequence_number":4,"object":"message","status":"completed","text":"The capital of France is Paris." }
+data: {"sequence_number":2,"object":"message","status":"in_progress", ... }
+data: {"sequence_number":3,"object":"content","status":"in_progress","text":"The" }
+data: {"sequence_number":4,"object":"content","status":"in_progress","text":" capital of France is Paris." }
+data: {"sequence_number":5,"object":"message","status":"completed","text":"The capital of France is Paris." }
+data: {"sequence_number":6,"object":"response","status":"completed", ... }
 ```
 
 ### Sandbox Example
@@ -484,6 +486,22 @@ agentscope-registry.ap-southeast-1.cr.aliyuncs.com/agentscope/runtime-sandbox-ba
 
 ---
 
+#### Serverless Sandbox Deployment
+
+AgentScope Runtime also supports serverless deployment, which is suitable for running sandboxes in a serverless environment,
+[Alibaba Cloud Function Compute (FC)](https://help.aliyun.com/zh/functioncompute/fc/) or [Alibaba Cloud AgentRun](https://docs.agent.run/).
+
+First, please refer to the [documentation](https://runtime.agentscope.io/en/sandbox/advanced.html#optional-function-compute-fc-settings) to configure the serverless environment variables.
+Make `CONTAINER_DEPLOYMENT` to `fc` or `agentrun` to enable serverless deployment.
+
+Then, start a sandbox server, use the `--config` option to specify a serverless environment setup:
+
+```bash
+# This command will load the settings defined in the `custom.env` file
+runtime-sandbox-server --config fc.env
+```
+After the server starts, you can access the sandbox server at baseurl `http://localhost:8000` and invoke sandbox tools described above.
+
 ## 📚 Cookbook
 
 - **[📖 Cookbook](https://runtime.agentscope.io/en/intro.html)**: Comprehensive tutorials
@@ -534,6 +552,40 @@ response = client.responses.create(
 print(response)
 ```
 
+Besides, `DeployManager` also supports serverless deployments, such as deploying your agent app
+to [ModelStudio](https://bailian.console.aliyun.com/?admin=1&tab=doc#/doc/?type=app&url=2983030)
+or [AgentRun](https://docs.agent.run/).
+
+```python
+from agentscope_runtime.engine.deployers import ModelStudioDeployManager
+# Create deployment manager
+deployer = ModelstudioDeployManager(
+    oss_config=OSSConfig(
+        access_key_id=os.environ.get("ALIBABA_CLOUD_ACCESS_KEY_ID"),
+        access_key_secret=os.environ.get("ALIBABA_CLOUD_ACCESS_KEY_SECRET"),
+    ),
+    modelstudio_config=ModelstudioConfig(
+        workspace_id=os.environ.get("MODELSTUDIO_WORKSPACE_ID"),
+        access_key_id=os.environ.get("ALIBABA_CLOUD_ACCESS_KEY_ID"),
+        access_key_secret=os.environ.get("ALIBABA_CLOUD_ACCESS_KEY_SECRET"),
+        dashscope_api_key=os.environ.get("DASHSCOPE_API_KEY"),
+    ),
+)
+
+# Deploy to ModelStudio
+result = await app.deploy(
+    deployer,
+    deploy_name="agent-app-example",
+    telemetry_enabled=True,
+    requirements=["agentscope", "fastapi", "uvicorn"],
+    environment={
+        "PYTHONPATH": "/app",
+        "DASHSCOPE_API_KEY": os.environ.get("DASHSCOPE_API_KEY"),
+    },
+)
+```
+
+For more advanced serverless deployment guides, please refer to the [documentation](https://runtime.agentscope.io/en/advanced_deployment.html#method-4-modelstudio-deployment).
 
 ---
 
@@ -584,7 +636,7 @@ limitations under the License.
 
 ## Contributors ✨
 <!-- ALL-CONTRIBUTORS-BADGE:START - Do not remove or modify this section -->
-[![All Contributors](https://img.shields.io/badge/all_contributors-23-orange.svg?style=flat-square)](#contributors-)
+[![All Contributors](https://img.shields.io/badge/all_contributors-25-orange.svg?style=flat-square)](#contributors-)
 <!-- ALL-CONTRIBUTORS-BADGE:END -->
 
 Thanks goes to these wonderful people ([emoji key](https://allcontributors.org/emoji-key/)):
@@ -624,6 +676,8 @@ Thanks goes to these wonderful people ([emoji key](https://allcontributors.org/e
     <tr>
       <td align="center" valign="top" width="14.28%"><a href="https://github.com/iSample"><img src="https://avatars.githubusercontent.com/u/12894421?v=4?s=100" width="100px;" alt="iSample"/><br /><sub><b>iSample</b></sub></a><br /><a href="https://github.com/agentscope-ai/agentscope-runtime/commits?author=iSample" title="Code">💻</a> <a href="https://github.com/agentscope-ai/agentscope-runtime/commits?author=iSample" title="Documentation">📖</a></td>
       <td align="center" valign="top" width="14.28%"><a href="https://github.com/XiuShenAl"><img src="https://avatars.githubusercontent.com/u/242360128?v=4?s=100" width="100px;" alt="XiuShenAl"/><br /><sub><b>XiuShenAl</b></sub></a><br /><a href="https://github.com/agentscope-ai/agentscope-runtime/commits?author=XiuShenAl" title="Code">💻</a> <a href="https://github.com/agentscope-ai/agentscope-runtime/commits?author=XiuShenAl" title="Documentation">📖</a></td>
+      <td align="center" valign="top" width="14.28%"><a href="https://github.com/k-farruh"><img src="https://avatars.githubusercontent.com/u/33511681?v=4?s=100" width="100px;" alt="Farruh Kushnazarov"/><br /><sub><b>Farruh Kushnazarov</b></sub></a><br /><a href="https://github.com/agentscope-ai/agentscope-runtime/commits?author=k-farruh" title="Documentation">📖</a></td>
+      <td align="center" valign="top" width="14.28%"><a href="https://github.com/fengxsong"><img src="https://avatars.githubusercontent.com/u/7008971?v=4?s=100" width="100px;" alt="fengxsong"/><br /><sub><b>fengxsong</b></sub></a><br /><a href="https://github.com/agentscope-ai/agentscope-runtime/issues?q=author%3Afengxsong" title="Bug reports">🐛</a></td>
     </tr>
   </tbody>
   <tfoot>

{agentscope_runtime-1.0.0b1 → agentscope_runtime-1.0.1}/README.md

@@ -236,9 +236,11 @@ You’ll see output streamed in **Server-Sent Events (SSE)** format:
 ```bash
 data: {"sequence_number":0,"object":"response","status":"created", ... }
 data: {"sequence_number":1,"object":"response","status":"in_progress", ... }
-data: {"sequence_number":2,"object":"content","status":"in_progress","text":"The" }
-data: {"sequence_number":3,"object":"content","status":"in_progress","text":" capital of France is Paris." }
-data: {"sequence_number":4,"object":"message","status":"completed","text":"The capital of France is Paris." }
+data: {"sequence_number":2,"object":"message","status":"in_progress", ... }
+data: {"sequence_number":3,"object":"content","status":"in_progress","text":"The" }
+data: {"sequence_number":4,"object":"content","status":"in_progress","text":" capital of France is Paris." }
+data: {"sequence_number":5,"object":"message","status":"completed","text":"The capital of France is Paris." }
+data: {"sequence_number":6,"object":"response","status":"completed", ... }
 ```
 
 ### Sandbox Example
@@ -419,6 +421,22 @@ agentscope-registry.ap-southeast-1.cr.aliyuncs.com/agentscope/runtime-sandbox-ba
 
 ---
 
+#### Serverless Sandbox Deployment
+
+AgentScope Runtime also supports serverless deployment, which is suitable for running sandboxes in a serverless environment,
+[Alibaba Cloud Function Compute (FC)](https://help.aliyun.com/zh/functioncompute/fc/) or [Alibaba Cloud AgentRun](https://docs.agent.run/).
+
+First, please refer to the [documentation](https://runtime.agentscope.io/en/sandbox/advanced.html#optional-function-compute-fc-settings) to configure the serverless environment variables.
+Make `CONTAINER_DEPLOYMENT` to `fc` or `agentrun` to enable serverless deployment.
+
+Then, start a sandbox server, use the `--config` option to specify a serverless environment setup:
+
+```bash
+# This command will load the settings defined in the `custom.env` file
+runtime-sandbox-server --config fc.env
+```
+After the server starts, you can access the sandbox server at baseurl `http://localhost:8000` and invoke sandbox tools described above.
+
 ## 📚 Cookbook
 
 - **[📖 Cookbook](https://runtime.agentscope.io/en/intro.html)**: Comprehensive tutorials
@@ -469,6 +487,40 @@ response = client.responses.create(
 print(response)
 ```
 
+Besides, `DeployManager` also supports serverless deployments, such as deploying your agent app
+to [ModelStudio](https://bailian.console.aliyun.com/?admin=1&tab=doc#/doc/?type=app&url=2983030)
+or [AgentRun](https://docs.agent.run/).
+
+```python
+from agentscope_runtime.engine.deployers import ModelStudioDeployManager
+# Create deployment manager
+deployer = ModelstudioDeployManager(
+    oss_config=OSSConfig(
+        access_key_id=os.environ.get("ALIBABA_CLOUD_ACCESS_KEY_ID"),
+        access_key_secret=os.environ.get("ALIBABA_CLOUD_ACCESS_KEY_SECRET"),
+    ),
+    modelstudio_config=ModelstudioConfig(
+        workspace_id=os.environ.get("MODELSTUDIO_WORKSPACE_ID"),
+        access_key_id=os.environ.get("ALIBABA_CLOUD_ACCESS_KEY_ID"),
+        access_key_secret=os.environ.get("ALIBABA_CLOUD_ACCESS_KEY_SECRET"),
+        dashscope_api_key=os.environ.get("DASHSCOPE_API_KEY"),
+    ),
+)
+
+# Deploy to ModelStudio
+result = await app.deploy(
+    deployer,
+    deploy_name="agent-app-example",
+    telemetry_enabled=True,
+    requirements=["agentscope", "fastapi", "uvicorn"],
+    environment={
+        "PYTHONPATH": "/app",
+        "DASHSCOPE_API_KEY": os.environ.get("DASHSCOPE_API_KEY"),
+    },
+)
+```
+
+For more advanced serverless deployment guides, please refer to the [documentation](https://runtime.agentscope.io/en/advanced_deployment.html#method-4-modelstudio-deployment).
 
 ---
 
@@ -519,7 +571,7 @@ limitations under the License.
 
 ## Contributors ✨
 <!-- ALL-CONTRIBUTORS-BADGE:START - Do not remove or modify this section -->
-[![All Contributors](https://img.shields.io/badge/all_contributors-23-orange.svg?style=flat-square)](#contributors-)
+[![All Contributors](https://img.shields.io/badge/all_contributors-25-orange.svg?style=flat-square)](#contributors-)
 <!-- ALL-CONTRIBUTORS-BADGE:END -->
 
 Thanks goes to these wonderful people ([emoji key](https://allcontributors.org/emoji-key/)):
@@ -559,6 +611,8 @@ Thanks goes to these wonderful people ([emoji key](https://allcontributors.org/e
     <tr>
       <td align="center" valign="top" width="14.28%"><a href="https://github.com/iSample"><img src="https://avatars.githubusercontent.com/u/12894421?v=4?s=100" width="100px;" alt="iSample"/><br /><sub><b>iSample</b></sub></a><br /><a href="https://github.com/agentscope-ai/agentscope-runtime/commits?author=iSample" title="Code">💻</a> <a href="https://github.com/agentscope-ai/agentscope-runtime/commits?author=iSample" title="Documentation">📖</a></td>
       <td align="center" valign="top" width="14.28%"><a href="https://github.com/XiuShenAl"><img src="https://avatars.githubusercontent.com/u/242360128?v=4?s=100" width="100px;" alt="XiuShenAl"/><br /><sub><b>XiuShenAl</b></sub></a><br /><a href="https://github.com/agentscope-ai/agentscope-runtime/commits?author=XiuShenAl" title="Code">💻</a> <a href="https://github.com/agentscope-ai/agentscope-runtime/commits?author=XiuShenAl" title="Documentation">📖</a></td>
+      <td align="center" valign="top" width="14.28%"><a href="https://github.com/k-farruh"><img src="https://avatars.githubusercontent.com/u/33511681?v=4?s=100" width="100px;" alt="Farruh Kushnazarov"/><br /><sub><b>Farruh Kushnazarov</b></sub></a><br /><a href="https://github.com/agentscope-ai/agentscope-runtime/commits?author=k-farruh" title="Documentation">📖</a></td>
+      <td align="center" valign="top" width="14.28%"><a href="https://github.com/fengxsong"><img src="https://avatars.githubusercontent.com/u/7008971?v=4?s=100" width="100px;" alt="fengxsong"/><br /><sub><b>fengxsong</b></sub></a><br /><a href="https://github.com/agentscope-ai/agentscope-runtime/issues?q=author%3Afengxsong" title="Bug reports">🐛</a></td>
     </tr>
   </tbody>
   <tfoot>

{agentscope_runtime-1.0.0b1 → agentscope_runtime-1.0.1}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "agentscope-runtime"
-version = "1.0.0b1"
+version = "1.0.1"
 description = "A production-ready runtime framework for agent applications, providing secure sandboxed execution environments and scalable deployment solutions with multi-framework support."
 readme = "README.md"
 requires-python = ">=3.10"
@@ -11,7 +11,7 @@ dependencies = [
     "openai",
     "pydantic>=2.11.7",
     "requests>=2.32.4",
-    "agentscope>=1.0.6",
+    "agentscope>=1.0.9",
     "docker>=7.1.0",
     "redis>=6.0.0",
     "oss2>=2.19.1",
@@ -84,4 +84,4 @@ ext = [
     "wheel",
     "alibabacloud-credentials",
     "PyYAML",
-]
+]

{agentscope_runtime-1.0.0b1 → agentscope_runtime-1.0.1}/src/agentscope_runtime/adapters/agentscope/message.py

@@ -1,5 +1,5 @@
 # -*- coding: utf-8 -*-
-# pylint:disable=too-many-branches,too-many-statements
+# pylint:disable=too-many-branches,too-many-statements,protected-access
 # TODO: support file block
 import json
 
@@ -7,6 +7,7 @@ from collections import OrderedDict
 from typing import Union, List
 from urllib.parse import urlparse
 
+from mcp.types import CallToolResult
 from agentscope.message import (
     Msg,
     ToolUseBlock,
@@ -19,6 +20,7 @@ from agentscope.message import (
     URLSource,
     Base64Source,
 )
+from agentscope.mcp._client_base import MCPClientBase
 
 from ...engine.schemas.agent_schemas import (
     Message,
@@ -338,6 +340,16 @@ def message_to_agentscope_msg(
         A single Msg object or a list of Msg objects.
     """
 
+    def _try_loads(v, default, keep_original=False):
+        if isinstance(v, (dict, list)):
+            return v
+        if isinstance(v, str) and v.strip():
+            try:
+                return json.loads(v)
+            except Exception:
+                return v if keep_original else default
+        return default
+
     def _convert_one(message: Message) -> Msg:
         # Normalize role
         if message.role == "tool":
@@ -365,12 +377,23 @@ def message_to_agentscope_msg(
             MessageType.FUNCTION_CALL,
         ):
             # convert PLUGIN_CALL, FUNCTION_CALL to ToolUseBlock
+            tool_args = None
+            for cnt in reversed(message.content):
+                if hasattr(cnt, "data"):
+                    v = cnt.data.get("arguments")
+                    if isinstance(v, (dict, list)) or (
+                        isinstance(v, str) and v.strip()
+                    ):
+                        tool_args = _try_loads(v, {}, keep_original=False)
+                        break
+            if tool_args is None:
+                tool_args = {}
             result["content"] = [
                 ToolUseBlock(
                     type="tool_use",
                     id=message.content[0].data["call_id"],
                     name=message.content[0].data.get("name"),
-                    input=json.loads(message.content[0].data["arguments"]),
+                    input=tool_args,
                 ),
             ]
         elif message.type in (
@@ -379,7 +402,20 @@ def message_to_agentscope_msg(
         ):
             # convert PLUGIN_CALL_OUTPUT, FUNCTION_CALL_OUTPUT to
             # ToolResultBlock
-            blk = json.loads(message.content[0].data["output"])
+            out = None
+            raw_output = ""
+            for cnt in reversed(message.content):
+                if hasattr(cnt, "data"):
+                    v = cnt.data.get("output")
+                    if isinstance(v, (dict, list)) or (
+                        isinstance(v, str) and v.strip()
+                    ):
+                        raw_output = v
+                        out = _try_loads(v, "", keep_original=True)
+                        break
+            if out is None:
+                out = ""
+            blk = out
 
             def is_valid_block(obj):
                 return any(
@@ -389,12 +425,19 @@ def message_to_agentscope_msg(
 
             if isinstance(blk, list):
                 if not all(is_valid_block(item) for item in blk):
-                    blk = message.content[0].data["output"]
+                    try:
+                        # Try to convert to MCP CallToolResult then to blocks
+                        blk = CallToolResult.model_validate(blk)
+                        blk = MCPClientBase._convert_mcp_content_to_as_blocks(
+                            blk.content,
+                        )
+                    except Exception:
+                        blk = raw_output
             elif isinstance(blk, dict):
                 if not is_valid_block(blk):
-                    blk = message.content[0].data["output"]
+                    blk = raw_output
             else:
-                blk = message.content[0].data["output"]
+                blk = raw_output
 
             result["content"] = [
                 ToolResultBlock(

{agentscope_runtime-1.0.0b1 → agentscope_runtime-1.0.1}/src/agentscope_runtime/adapters/agentscope/stream.py

@@ -14,6 +14,8 @@ from ...engine.schemas.agent_schemas import (
     ImageContent,
     AudioContent,
     DataContent,
+    McpCall,
+    McpCallOutput,
     FunctionCall,
     FunctionCallOutput,
     MessageType,
@@ -78,8 +80,6 @@ async def adapt_agentscope_message_stream(
             # Note: Tool use content only happens in the last of messages
             tool_start = False
 
-            tool_use_messages_dict = {}
-
             # Cache msg id
             msg_id = msg.id
 
@@ -251,24 +251,65 @@ async def adapt_agentscope_message_stream(
                     elif element.get("type") == "tool_use":  # Tool use
                         call_id = element.get("id")
 
+                        if element.get("tool_type", "plugin") == "mcp":
+                            msg_type = MessageType.MCP_TOOL_CALL
+                            fc_cls = McpCall
+                            fc_kwargs = {
+                                "server_label": element.get("server_label"),
+                            }
+                        else:
+                            msg_type = MessageType.PLUGIN_CALL
+                            fc_cls = FunctionCall
+                            fc_kwargs = {}
+
                         if last:
-                            plugin_call_message = tool_use_messages_dict[
-                                call_id
-                            ]
-                            json_str = json.dumps(element.get("input"))
+                            plugin_call_message = tool_use_messages_dict.get(
+                                call_id,
+                            )
+
+                            if plugin_call_message is None:
+                                # Only one tool use message yields, we fake
+                                #  Build a new tool call message
+                                plugin_call_message = Message(
+                                    type=msg_type,
+                                    role="assistant",
+                                )
+
+                                data_delta_content = DataContent(
+                                    index=0,
+                                    data=fc_cls(
+                                        call_id=element.get("id"),
+                                        name=element.get("name"),
+                                        arguments="",
+                                        **fc_kwargs,
+                                    ).model_dump(),
+                                    delta=False,
+                                )
+
+                                plugin_call_message = _update_obj_attrs(
+                                    plugin_call_message,
+                                    metadata=metadata,
+                                    usage=usage,
+                                )
+                                yield plugin_call_message.in_progress()
+                                yield data_delta_content.in_progress()
+
+                            json_str = json.dumps(
+                                element.get("input"),
+                                ensure_ascii=False,
+                            )
                             data_delta_content = DataContent(
-                                index=index,
-                                data=FunctionCall(
+                                index=None,  # Will be set by `add_content`
+                                data=fc_cls(
                                     call_id=element.get("id"),
                                     name=element.get("name"),
                                     arguments=json_str,
+                                    **fc_kwargs,
                                 ).model_dump(),
-                                delta=True,
+                                delta=False,
                             )
-                            data_delta_content = (
-                                plugin_call_message.add_delta_content(
-                                    new_content=data_delta_content,
-                                )
+                            plugin_call_message.add_content(
+                                new_content=data_delta_content,
                             )
                             yield data_delta_content.completed()
                             plugin_call_message = _update_obj_attrs(
@@ -284,18 +325,19 @@ async def adapt_agentscope_message_stream(
                             else:
                                 # Build a new tool call message
                                 plugin_call_message = Message(
-                                    type=MessageType.PLUGIN_CALL,
+                                    type=msg_type,
                                     role="assistant",
                                 )
 
                                 data_delta_content = DataContent(
-                                    index=index,
-                                    data=FunctionCall(
+                                    index=0,
+                                    data=fc_cls(
                                         call_id=element.get("id"),
                                         name=element.get("name"),
                                         arguments="",
+                                        **fc_kwargs,
                                     ).model_dump(),
-                                    delta=True,
+                                    delta=False,
                                 )
 
                                 plugin_call_message = _update_obj_attrs(
@@ -304,32 +346,46 @@ async def adapt_agentscope_message_stream(
                                     usage=usage,
                                 )
                                 yield plugin_call_message.in_progress()
-                                data_delta_content = (
-                                    plugin_call_message.add_delta_content(
-                                        new_content=data_delta_content,
-                                    )
-                                )
-                                yield data_delta_content
+                                yield data_delta_content.in_progress()
 
                                 tool_use_messages_dict[
                                     call_id
                                 ] = plugin_call_message
 
                     elif element.get("type") == "tool_result":  # Tool result
+                        call_id = element.get("id")
+
+                        plugin_call_message = tool_use_messages_dict.get(
+                            call_id,
+                        )
+                        # Determine the output message type and class to use
+                        # for the tool result message based on the type of
+                        # the original tool call message.
+                        msg_type = MessageType.PLUGIN_CALL_OUTPUT
+                        fc_cls = FunctionCallOutput
+
+                        if plugin_call_message:
+                            if (
+                                plugin_call_message.type
+                                == MessageType.MCP_TOOL_CALL
+                            ):
+                                msg_type = MessageType.MCP_TOOL_CALL_OUTPUT
+                                fc_cls = McpCallOutput
+
                         json_str = json.dumps(
                             element.get("output"),
                             ensure_ascii=False,
                         )
                         data_delta_content = DataContent(
                             index=index,
-                            data=FunctionCallOutput(
+                            data=fc_cls(
                                 call_id=element.get("id"),
                                 name=element.get("name"),
                                 output=json_str,
                             ).model_dump(),
                         )
                         plugin_output_message = Message(
-                            type=MessageType.PLUGIN_CALL_OUTPUT,
+                            type=msg_type,
                             role="tool",
                             content=[data_delta_content],
                         )

{agentscope_runtime-1.0.0b1 → agentscope_runtime-1.0.1}/src/agentscope_runtime/adapters/agentscope/tool/tool.py

@@ -9,9 +9,7 @@ from typing import (
 )
 
 from agentscope.tool import Toolkit, ToolResponse
-from agentscope.tool._registered_tool_function import (
-    RegisteredToolFunction,
-)
+from agentscope.tool._types import RegisteredToolFunction
 
 from agentscope_runtime.tools.base import Tool
 

{agentscope_runtime-1.0.0b1 → agentscope_runtime-1.0.1}/src/agentscope_runtime/engine/app/agent_app.py

@@ -1,6 +1,7 @@
 # -*- coding: utf-8 -*-
 import logging
 import types
+import platform
 import subprocess
 import shlex
 from typing import Optional, Callable, List
@@ -203,7 +204,13 @@ class AgentApp(BaseApp):
                     "[AgentApp] Note: First WebUI launch may take extra time "
                     "as dependencies are installed.",
                 )
-                with subprocess.Popen(shlex.split(cmd)):
+
+                cmd_kwarg = {}
+                if platform.system() == "Windows":
+                    cmd_kwarg.update({"shell": True})
+                else:
+                    cmd = shlex.split(cmd)
+                with subprocess.Popen(cmd, **cmd_kwarg):
                     uvicorn.run(
                         fastapi_app,
                         host=host,

{agentscope_runtime-1.0.0b1 → agentscope_runtime-1.0.1}/src/agentscope_runtime/engine/deployers/agentrun_deployer.py

@@ -1022,7 +1022,7 @@ ls -lh /output/{zip_filename}
                     artifact_type="Code",
                     cpu=self.agentrun_config.cpu,
                     memory=self.agentrun_config.memory,
-                    port=8080,
+                    port=8090,
                     code_configuration=CodeConfig(
                         command=["python3", "/code/deploy_starter/main.py"],
                         oss_bucket_name=oss_bucket_name,
@@ -1159,7 +1159,7 @@ ls -lh /output/{zip_filename}
                 artifact_type="Code",
                 cpu=self.agentrun_config.cpu,
                 memory=self.agentrun_config.memory,
-                port=8080,
+                port=8090,
                 code_configuration=CodeConfig(
                     command=["python3", "/code/deploy_starter/main.py"],
                     oss_bucket_name=oss_bucket_name,

{agentscope_runtime-1.0.0b1 → agentscope_runtime-1.0.1}/src/agentscope_runtime/engine/deployers/utils/service_utils/fastapi_factory.py

@@ -693,9 +693,32 @@ class FastAPIAppFactory:
         parsing."""
         is_async_gen = inspect.isasyncgenfunction(handler)
 
+        # NOTE:
+        # -----
+        # FastAPI >= 0.123.5 uses Dependant.is_coroutine_callable, which in
+        # turn unwraps callables via inspect.unwrap() and then inspects the
+        # unwrapped target to decide whether it is a coroutine function /
+        # generator / async generator.
+        #
+        # If we decorate an async-generator handler with
+        # functools.wraps(handler), FastAPI will unwrap back to the original
+        # async-generator function and *misclassify* the endpoint as
+        # non-coroutine. It will then call our async wrapper *without awaiting
+        # it*, and later try to JSON-encode the resulting coroutine object,
+        # causing errors like:
+        # TypeError("'coroutine' object is not iterable")
+        #
+        # To avoid that, we deliberately do NOT use functools.wraps() here.
+        # Instead, we manually copy the key metadata (name, qualname, doc,
+        # module, and signature) from the original handler, but we do NOT set
+        # __wrapped__. This ensures:
+        #   * FastAPI sees the wrapper itself as the callable (an async def),
+        #     so Dependant.is_coroutine_callable is True, and it is properly
+        #     awaited.
+        #   * FastAPI still sees the correct signature for parameter parsing.
+
         if is_async_gen:
 
-            @functools.wraps(handler)
             async def wrapped_handler(*args, **kwargs):
                 async def generate():
                     try:
@@ -720,12 +743,8 @@ class FastAPIAppFactory:
                     media_type="text/event-stream",
                 )
 
-            wrapped_handler.__signature__ = inspect.signature(handler)
-            return wrapped_handler
-
         else:
 
-            @functools.wraps(handler)
             def wrapped_handler(*args, **kwargs):
                 def generate():
                     try:
@@ -748,8 +767,34 @@ class FastAPIAppFactory:
                     media_type="text/event-stream",
                 )
 
-            wrapped_handler.__signature__ = inspect.signature(handler)
-            return wrapped_handler
+        # Manually propagate essential metadata without creating a __wrapped__
+        # chain that would confuse FastAPI's unwrap logic.
+        wrapped_handler.__name__ = getattr(
+            handler,
+            "__name__",
+            wrapped_handler.__name__,
+        )
+        wrapped_handler.__qualname__ = getattr(
+            handler,
+            "__qualname__",
+            wrapped_handler.__qualname__,
+        )
+        wrapped_handler.__doc__ = getattr(
+            handler,
+            "__doc__",
+            wrapped_handler.__doc__,
+        )
+        wrapped_handler.__module__ = getattr(
+            handler,
+            "__module__",
+            wrapped_handler.__module__,
+        )
+        wrapped_handler.__signature__ = inspect.signature(handler)
+
+        # Make sure FastAPI doesn't see any stale __wrapped__ pointing back to
+        # the original async-generator; if present, remove it.
+
+        return wrapped_handler
 
     @staticmethod
     def _add_custom_endpoints(app: FastAPI):