bohr-agent-sdk 0.1.102__py3-none-any.whl → 0.1.104__py3-none-any.whl

bohr_agent_sdk-0.1.104.dist-info/METADATA

@@ -0,0 +1,293 @@
+Metadata-Version: 2.4
+Name: bohr-agent-sdk
+Version: 0.1.104
+Summary: SDK for scientific agents
+Home-page: https://github.com/dptech-corp/bohr-agent-sdk/
+Author: DP Technology
+Maintainer-email: liupeng <liupeng@dp.tech>, zjgemi <liuxzj@dp.tech>
+License: MIT
+Project-URL: Homepage, https://github.com/dptech-corp/bohr-agent-sdk
+Project-URL: repository, https://github.com/dptech-corp/bohr-agent-sdk
+Project-URL: Bug Reports, https://github.com/dptech-corp/bohr-agent-sdk/issues
+Keywords: agent SDK,AI for science
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+Requires-Dist: click>=8.0.0
+Requires-Dist: mcp>=1.17.0
+Requires-Dist: python-dotenv>=1.0.0
+Requires-Dist: typing-extensions>=4.8.0
+Requires-Dist: dpdispatcher>=0.6.8
+Requires-Dist: lbg>=1.2.29
+Requires-Dist: jsonpickle>=3.0.3
+Requires-Dist: psutil>=5.9.6
+Requires-Dist: paho-mqtt>=2.1.0
+Requires-Dist: redis>=6.2.0
+Requires-Dist: twine>=6.1.0
+Requires-Dist: build>=1.2.2.post1
+Requires-Dist: cloudpickle==2.2.0
+Requires-Dist: watchdog>=6.0.0
+Requires-Dist: fastapi>=0.116.0
+Requires-Dist: bohrium-open-sdk
+Provides-Extra: device
+Requires-Dist: pywinauto-recorder>=0.1.0; extra == "device"
+Provides-Extra: cloud
+Requires-Dist: paho-mqtt>=1.6.1; extra == "cloud"
+Requires-Dist: redis>=5.0.1; extra == "cloud"
+Requires-Dist: aiohttp>=3.9.1; extra == "cloud"
+Provides-Extra: dev
+Requires-Dist: pytest>=7.4.0; extra == "dev"
+Requires-Dist: pytest-asyncio>=0.23.0; extra == "dev"
+Requires-Dist: pytest-cov>=4.1.0; extra == "dev"
+Requires-Dist: black>=23.11.0; extra == "dev"
+Requires-Dist: isort>=5.12.0; extra == "dev"
+Requires-Dist: mypy>=1.7.0; extra == "dev"
+Requires-Dist: pylint>=3.0.0; extra == "dev"
+Requires-Dist: google-adk; extra == "dev"
+Provides-Extra: docs
+Requires-Dist: sphinx>=7.2.0; extra == "docs"
+Requires-Dist: sphinx-rtd-theme>=1.3.0; extra == "docs"
+Provides-Extra: all
+Requires-Dist: bohr-agent-sdk[bohrium,cloud,dev,device,dispatcher,docs]; extra == "all"
+Dynamic: home-page
+Dynamic: requires-python
+
+# Bohrium Science Agent SDK
+
+[English](README.md) | [简体中文](README_CN.md)
+
+**Transform Scientific Software into AI Assistants — 3 Steps to Intelligent Transformation**
+
+## 📖 Introduction
+
+The Bohrium platform introduces the **bohr-agent-sdk Scientific Agent Development Kit**, enabling AI systems to truly execute professional scientific tasks and helping developers quickly build their own specialized research agents. Through a three-step process — **Invoking MCP Tools, Orchestrating Agent Workflows, and Deploying Services** — any scientific software can be rapidly transformed into an AI assistant.
+
+## ✨ Core Features
+
+### 🎯 Intelligent Task Management: Simplified Development, Standardized Output
+With a decorator pattern, just a few annotations can quickly transform scientific computing programs into MCP standard services. Built-in application templates turn scattered research code into standardized, reusable intelligent components.
+
+### 🔧 Multi-Backend Framework Support
+Supports mainstream Agent open frameworks including Google ADK, Langraph, and Camel, providing flexible choices for developers familiar with different technology stacks.
+
+### ☁️ Flexible Deployment: Local Development, Cloud Production
+Dual-mode architecture supports seamless transition between development and production. Local environments enable rapid iteration and feature validation, while Bohrium's cloud GPU clusters handle production-grade computing tasks. The SDK automatically manages the complete workflow of task scheduling, status monitoring, and result collection, with built-in file transfer mechanisms for handling large-scale data uploads and downloads. Developers focus on core algorithm implementation while infrastructure management is fully automated.
+
+### 🖼️ Visual Interactive Interface: Professional Presentation, Intuitive Operation
+Based on the modern React framework, deploy fully-featured web applications with one click. Built-in 3D molecular visualization engine supports multiple structure formats and rendering modes for interactive molecular structure display. Real-time data synchronization ensures instant computing status updates, while multi-session management supports parallel task processing. Integrated with enterprise-grade features including file management, project switching, and permission control. Transform command-line tools into professional visual applications, significantly enhancing user experience and tool usability.
+
+## 🖼️ Interface Showcase
+
+### Scientific Computing Master Console
+<div align="center">
+
+![SCIMaster](image/SCIMaster.PNG)
+
+*Powerful scientific computing task management and monitoring platform*
+
+</div>
+
+### Visual Interactive Interface
+<div align="center">
+
+![UI](image/UI.png)
+
+*Modern web application interface providing intuitive user experience*
+
+</div>
+
+## 🚀 Quick Start
+
+### Installation
+
+```bash
+pip install bohr-agent-sdk -i https://pypi.org/simple --upgrade
+```
+
+### Build Your Research Agent in 3 Steps
+
+#### Step 1: Get Project Templates
+
+```bash
+# Get calculation project template
+dp-agent fetch scaffolding --type=calculation
+
+# Get device control project template
+dp-agent fetch scaffolding --type=device
+
+# Get configuration file
+dp-agent fetch config
+```
+
+#### Step 2: Develop Your Agent
+
+**Lab Mode Development Example**
+
+```python
+from typing import Dict, TypedDict
+from dp.agent.device.device import Device, action, BaseParams, SuccessResult
+
+class TakePictureParams(BaseParams):
+    """Picture taking parameters"""
+    horizontal_width: str  # Image horizontal width
+
+class PictureData(TypedDict):
+    """Picture data structure"""
+    image_id: str
+
+class PictureResult(SuccessResult):
+    """Picture taking result"""
+    data: PictureData
+
+class MyDevice(Device):
+    """Custom device class"""
+    device_name = "my_device"
+
+    @action("take_picture")
+    def take_picture(self, params: TakePictureParams) -> PictureResult:
+        """
+        Execute picture taking action
+
+        Through the @action decorator, automatically register this method as an MCP standard service
+        """
+        hw = params.get("horizontal_width", "default")
+        # Execute actual device control logic
+        return PictureResult(
+            message=f"Picture taken with {self.device_name}",
+            data={"image_id": "image_123"}
+        )
+```
+
+**Cloud Mode Development Example**
+
+```python
+"""
+MCP protocol-based cloud device control example
+"""
+import signal
+import sys
+from dp.agent.cloud import mcp, get_mqtt_cloud_instance
+from dp.agent.device.device import TescanDevice, register_mcp_tools
+
+def signal_handler(sig, frame):
+    """Graceful shutdown handling"""
+    print("Shutting down...")
+    get_mqtt_cloud_instance().stop()
+    sys.exit(0)
+
+def main():
+    """Start cloud services"""
+    print("Starting Tescan Device Twin Cloud Services...")
+
+    # Register signal handler
+    signal.signal(signal.SIGINT, signal_handler)
+
+    # Create device instance
+    device = TescanDevice(mcp, device)
+
+    # Automatically register device tools to MCP server
+    # register_mcp_tools implements automatic registration through Python introspection
+    register_mcp_tools(device)
+
+    # Start MCP server
+    print("Starting MCP server...")
+    mcp.run(transport="sse")
+
+if __name__ == "__main__":
+    main()
+```
+
+#### Step 3: Run and Deploy
+
+```bash
+# Local lab environment
+dp-agent run tool device
+
+# Cloud computing environment
+dp-agent run tool cloud
+
+# Scientific calculation mode
+dp-agent run tool calculation
+
+# Start agent (with Web UI)
+dp-agent run agent --config
+
+# Debug mode
+dp-agent run debug
+```
+
+## 🏗️ Project Structure
+
+After running `dp-agent fetch scaffolding`, you'll get a standardized project structure:
+
+```
+your-project/
+├── lab/                    # Lab mode
+│   ├── __init__.py
+│   └── tescan_device.py    # Device control implementation
+├── cloud/                  # Cloud mode
+│   ├── __init__.py
+│   └── mcp_server.py       # MCP service implementation
+├── calculation/            # Calculation mode
+│   └── __init__.py
+├── .env                    # Environment configuration
+└── main.py                 # Main program entry
+```
+
+## ⚙️ Configuration
+
+Configure necessary environment variables in the `.env` file:
+
+```bash
+# MQTT connection configuration
+MQTT_INSTANCE_ID=your_instance_id
+MQTT_ENDPOINT=your_endpoint
+MQTT_DEVICE_ID=your_device_id
+MQTT_GROUP_ID=your_group_id
+MQTT_AK=your_access_key
+MQTT_SK=your_secret_key
+
+# Computing resource configuration
+BOHRIUM_USERNAME=your_username
+BOHRIUM_PASSWORD=your_password
+```
+
+Note: The `dp-agent fetch config` command automatically downloads configuration files and replaces dynamic variables (such as MQTT_DEVICE_ID). For security reasons, this feature is only available in internal network environments.
+
+## 🎯 Application Scenarios
+
+- **Materials Science Computing**: Molecular dynamics simulation, first-principles calculations
+- **Bioinformatics Analysis**: Gene sequence analysis, protein structure prediction
+- **Laboratory Equipment Control**: Intelligent control of research equipment such as electron microscopes and X-ray diffractometers
+- **Data Processing Workflows**: Automated data cleaning, analysis, and visualization
+- **Machine Learning Training**: Model training, hyperparameter optimization, result evaluation
+
+## 🔧 Advanced Features
+
+### File Management
+
+```bash
+# Upload files to cloud
+dp-agent artifact upload <path>
+
+# Download cloud files
+dp-agent artifact download <artifact_id>
+```
+
+### Task Monitoring
+
+The SDK provides real-time task status monitoring, supporting:
+- Task queue management
+- Computing resource scheduling
+- Automatic result collection
+- Exception handling and retry mechanisms
+
+## 📚 Documentation & Support
+
+- 📖 [Detailed Documentation](https://dptechnology.feishu.cn/wiki/ZSj9wbLJEiwdNek0Iu7cKsFanuW)
+

{bohr_agent_sdk-0.1.102.dist-info → bohr_agent_sdk-0.1.104.dist-info}/RECORD

@@ -9,7 +9,7 @@ dp/agent/adapter/camel/__init__.py,sha256=RN1NhdmsJyN43fTxTXFld4UKZksjpSV0b2QvFn
 dp/agent/adapter/camel/client/__init__.py,sha256=ld-r0_WsZLFv6yyrmxjWmR8JgnrQzOw4fX0hwVHzciY,93
 dp/agent/adapter/camel/client/calculation_mcp_client.py,sha256=JZZUYYfMgXvHzK4f6IJp-ia33wn3aYZqDSDVa5yKtdc,1860
 dp/agent/cli/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
-dp/agent/cli/cli.py,sha256=3CPdZW3nn_n5-1Ohak_LOOHqNQJbnYZHoTXIVjI19VU,12367
+dp/agent/cli/cli.py,sha256=Xp3Q7gp50ttsidPG1i3bOQ0-_pMN7yILePJ-ZyOogvA,12363
 dp/agent/cli/templates/__init__.py,sha256=h5__iNn8QzUyYpCUORJO9GTd42tga_BFQ3ZjA3QtSCI,15
 dp/agent/cli/templates/main.py.template,sha256=gEv_naKkBKUmVY1aGM_RZPWqXhkYDIrk4eu7uQFpWrA,1986
 dp/agent/cli/templates/calculation/simple.py.template,sha256=AkOrMWZf9YKQuLnl5yRgl2V-fepFuZJ1qDcTc9S5Gj0,395
@@ -60,21 +60,21 @@ dp/agent/device/device/__init__.py,sha256=w7_1S16S1vWUq0RGl0GFgjq2vFkc5oNvy8cQTn
 dp/agent/device/device/device.py,sha256=9ZRIJth-4qMO-i-u_b_cO3d6a4eTbTQjPaxFsV_zEkc,9643
 dp/agent/device/device/types.py,sha256=JuxB-hjf1CjjvfBxCLwRAXVFlYS-nPEdiJpBWLFVCzo,1924
 dp/agent/server/__init__.py,sha256=rckaYd8pbYyB4ENEhgjXKeGMXjdnrgcJpdM1gu5u1Wc,508
-dp/agent/server/calculation_mcp_server.py,sha256=eClRP7A-t5hMGyTh81KC3GAKjSPNJIylOrOKyzqwo8o,11459
+dp/agent/server/calculation_mcp_server.py,sha256=iRFOdgTxySMGk7ZaSseNssEp-A7zT5cW1Ym2_MIKnG4,12602
 dp/agent/server/preprocessor.py,sha256=XUWu7QOwo_sIDMYS2b1OTrM33EXEVH_73vk-ju1Ok8A,1264
-dp/agent/server/utils.py,sha256=8jgYZEW4XBp86AF2Km6QkwHltBmrnS-soTpHov7ZEJw,4501
+dp/agent/server/utils.py,sha256=ui3lca9EagcGqmYf8BKLsPARIzXxJ3jgN98yuEO3OSQ,1668
 dp/agent/server/executor/__init__.py,sha256=s95M5qKQk39Yi9qaVJZhk_nfj54quSf7EDghR3OCFUA,248
 dp/agent/server/executor/base_executor.py,sha256=EFJBsYVYAvuRbiLAbLOwLTw3h7ScjN025xnSP4uJHrQ,2052
-dp/agent/server/executor/dispatcher_executor.py,sha256=urpzmKH_tBOgblBdJEa3y8eEhXqUDrdcdWCnUdJpfZk,9420
+dp/agent/server/executor/dispatcher_executor.py,sha256=p2ISxvLUcR1QOPF5BxDLD7AFPFibmw5uKJJ_fL8zecY,10836
 dp/agent/server/executor/local_executor.py,sha256=wYCclNZFkLb3v7KpW1nCnupO8piBES-esYlDAuz86zk,6120
 dp/agent/server/storage/__init__.py,sha256=Sgsyp5hb0_hhIGugAPfQFzBHt_854rS_MuMuE3sn8Gs,389
 dp/agent/server/storage/base_storage.py,sha256=728-oNG6N8isV95gZVnyi4vTznJPJhSjxw9Gl5Y_y5o,2356
 dp/agent/server/storage/bohrium_storage.py,sha256=EsKX4dWWvZTn2TEhZv4zsvihfDK0mmPFecrln-Ytk40,10488
-dp/agent/server/storage/http_storage.py,sha256=w0lY95wQqKmjXTGFRhEG2hLu8GBFwgqG8ocm5lJ_fYc,1470
+dp/agent/server/storage/http_storage.py,sha256=KiySq7g9-iJr12XQCKKyJLn8wJoDnSRpQAR5_qPJ1ZU,1471
 dp/agent/server/storage/local_storage.py,sha256=t1wfjByjXew9ws3PuUxWxmZQ0-Wt1a6t4wmj3fW62GI,1352
 dp/agent/server/storage/oss_storage.py,sha256=pgjmi7Gir3Y5wkMDCvU4fvSls15fXT7Ax-h9MYHFPK0,3359
-bohr_agent_sdk-0.1.102.dist-info/METADATA,sha256=7FQMW--VfxyMkPlKfbj9_j5oMWFExoM_0cpxTG1a4-Y,6495
-bohr_agent_sdk-0.1.102.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-bohr_agent_sdk-0.1.102.dist-info/entry_points.txt,sha256=5n5kneF5IbDQtoQ2WfF-QuBjDtsimJte9Rv9baSGgc0,86
-bohr_agent_sdk-0.1.102.dist-info/top_level.txt,sha256=87xLUDhu_1nQHoGLwlhJ6XlO7OsjILh6i1nX6ljFzDo,3
-bohr_agent_sdk-0.1.102.dist-info/RECORD,,
+bohr_agent_sdk-0.1.104.dist-info/METADATA,sha256=h9BBCLrH_qtMCtfHzoCpk0WIgppExXMvQ8jwnXhWFHg,10226
+bohr_agent_sdk-0.1.104.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+bohr_agent_sdk-0.1.104.dist-info/entry_points.txt,sha256=5n5kneF5IbDQtoQ2WfF-QuBjDtsimJte9Rv9baSGgc0,86
+bohr_agent_sdk-0.1.104.dist-info/top_level.txt,sha256=87xLUDhu_1nQHoGLwlhJ6XlO7OsjILh6i1nX6ljFzDo,3
+bohr_agent_sdk-0.1.104.dist-info/RECORD,,

dp/agent/cli/cli.py

@@ -188,7 +188,7 @@ def calculation():
 def agent(ui, config, port, module, agent_name, dev):
     """Run the science agent with optional UI interface."""
     if not ui:
-            click.echo("Starting agent in console mode...")
+        click.echo("Starting agent in console mode...")
         click.echo("Console mode not yet implemented.")
         return
     

dp/agent/server/calculation_mcp_server.py

@@ -6,17 +6,23 @@ from copy import deepcopy
 from datetime import datetime
 from pathlib import Path
 from urllib.parse import urlparse
-from typing import Literal, Optional, get_origin
+from typing import Any, Literal, Optional, TypedDict
 
+import mcp
 from mcp.server.fastmcp import FastMCP
-from mcp.server.fastmcp.server import Context
-from mcp.server.fastmcp.utilities.func_metadata import _get_typed_signature
+from mcp.server.fastmcp.utilities.context_injection import (
+    find_context_parameter,
+)
+from mcp.server.fastmcp.utilities.func_metadata import (
+    _get_typed_signature,
+    func_metadata,
+)
 from starlette.responses import JSONResponse
 from starlette.routing import Route
 
 from .executor import executor_dict
 from .storage import storage_dict
-from .utils import get_logger, get_metadata, convert_to_content, Tool
+from .utils import get_logger, JobResult, Tool
 logger = get_logger(__name__)
 
 
@@ -133,7 +139,7 @@ def handle_output_artifacts(results, exec_id, storage):
 
 
 def get_job_results(job_id: str, executor: Optional[dict] = None,
-                    storage: Optional[dict] = None) -> dict:
+                    storage: Optional[dict] = None) -> Any:
     """
     Get results of a calculation job
     Args:
@@ -148,63 +154,80 @@ def get_job_results(job_id: str, executor: Optional[dict] = None,
         results, output_artifacts = handle_output_artifacts(
             results, exec_id, storage)
         logger.info("Job %s result is %s" % (job_id, results))
-    return convert_to_content(results, job_info={
+    return JobResult(result=results, job_info={
         "output_artifacts": output_artifacts,
     })
 
 
 class CalculationMCPServer:
-    def __init__(self, *args, preprocess_func=None, **kwargs):
+    def __init__(self, *args, preprocess_func=None, fastmcp_mode=False,
+                 **kwargs):
+        """
+        Args:
+            preprocess_func: The preprocess function for all tools
+            fastmcp_mode: compatible for fastmcp.FastMCP
+        """
         self.preprocess_func = preprocess_func
+        self.fastmcp_mode = fastmcp_mode
         self.mcp = FastMCP(*args, **kwargs)
 
-    def add_patched_tool(self, fn, new_fn, name, is_async=False, doc=None):
-        # patch the metadata of the tool
-        context_kwarg = None
-        sig = inspect.signature(fn)
-        for param_name, param in sig.parameters.items():
-            if get_origin(param.annotation) is not None:
-                continue
-            if issubclass(param.annotation, Context):
-                context_kwarg = param_name
-                break
-        # combine parameters
-        parameters = []
-        for param in _get_typed_signature(fn).parameters.values():
-            if param.annotation is Path:
-                parameters.append(inspect.Parameter(
-                    name=param.name, default=param.default,
-                    annotation=str, kind=param.kind))
-            elif param.annotation is Optional[Path]:
-                parameters.append(inspect.Parameter(
-                    name=param.name, default=param.default,
-                    annotation=Optional[str], kind=param.kind))
-            else:
-                parameters.append(param)
-        for param in _get_typed_signature(new_fn).parameters.values():
-            if param.name != "kwargs":
-                parameters.append(param)
-        func_arg_metadata = get_metadata(
-            name,
-            parameters=parameters,
+    def add_patched_tool(self, fn, new_fn, name, is_async=False, doc=None,
+                         override_return_annotation=False):
+        """patch the metadata of the tool"""
+        context_kwarg = find_context_parameter(fn)
+
+        def _get_typed_signature_patched(call):
+            """patch parameters"""
+            typed_signature = _get_typed_signature(call)
+            new_typed_signature = _get_typed_signature(new_fn)
+            parameters = []
+            for param in typed_signature.parameters.values():
+                if param.annotation is Path:
+                    parameters.append(inspect.Parameter(
+                        name=param.name, default=param.default,
+                        annotation=str, kind=param.kind))
+                elif param.annotation is Optional[Path]:
+                    parameters.append(inspect.Parameter(
+                        name=param.name, default=param.default,
+                        annotation=Optional[str], kind=param.kind))
+                else:
+                    parameters.append(param)
+            for param in new_typed_signature.parameters.values():
+                if param.name != "kwargs":
+                    parameters.append(param)
+            return inspect.Signature(
+                parameters,
+                return_annotation=(new_typed_signature.return_annotation
+                                   if override_return_annotation
+                                   else typed_signature.return_annotation))
+
+        # Due to the frequent changes of MCP, we use a patching style here
+        mcp.server.fastmcp.utilities.func_metadata._get_typed_signature = \
+            _get_typed_signature_patched
+        func_arg_metadata = func_metadata(
+            fn,
             skip_names=[context_kwarg] if context_kwarg is not None else [],
-            globalns=getattr(fn, "__globals__", {})
+            structured_output=None,
         )
-        json_schema = func_arg_metadata.arg_model.model_json_schema()
+        mcp.server.fastmcp.utilities.func_metadata._get_typed_signature = \
+            _get_typed_signature
+        if self.fastmcp_mode and func_arg_metadata.wrap_output:
+            # Only simulate behavior of fastmcp for output_schema
+            func_arg_metadata.output_schema["x-fastmcp-wrap-result"] = True
+        parameters = func_arg_metadata.arg_model.model_json_schema(
+            by_alias=True)
         tool = Tool(
             fn=new_fn,
             name=name,
             description=doc or fn.__doc__,
-            parameters=json_schema,
+            parameters=parameters,
             fn_metadata=func_arg_metadata,
             is_async=is_async,
             context_kwarg=context_kwarg,
-            annotations=None,
         )
         self.mcp._tool_manager._tools[name] = tool
 
     def add_tool(self, fn, *args, **kwargs):
-        self.mcp.add_tool(fn, *args, **kwargs)
         tool = Tool.from_function(fn, *args, **kwargs)
         self.mcp._tool_manager._tools[tool.name] = tool
         return tool
@@ -215,7 +238,9 @@ class CalculationMCPServer:
 
         def decorator(fn: Callable) -> Callable:
             def submit_job(executor: Optional[dict] = None,
-                           storage: Optional[dict] = None, **kwargs):
+                           storage: Optional[dict] = None,
+                           **kwargs) -> TypedDict("results", {
+                               "job_id": str, "extra_info": Optional[dict]}):
                 trace_id = datetime.today().strftime('%Y-%m-%d-%H:%M:%S.%f')
                 logger.info("Job processing (Trace ID: %s)" % trace_id)
                 with set_directory(trace_id):
@@ -233,7 +258,7 @@ class CalculationMCPServer:
                     "job_id": job_id,
                     "extra_info": res.get("extra_info"),
                 }
-                return convert_to_content(result, job_info={
+                return JobResult(result=result, job_info={
                     "trace_id": trace_id,
                     "executor_type": executor_type,
                     "job_id": job_id,
@@ -263,7 +288,7 @@ class CalculationMCPServer:
                     logger.info("Job %s result is %s" % (job_id, results))
                     await context.log(level="info", message="Job %s result is"
                                       " %s" % (job_id, results))
-                return convert_to_content(results, job_info={
+                return JobResult(result=results, job_info={
                     "trace_id": trace_id,
                     "executor_type": executor_type,
                     "job_id": job_id,
@@ -273,8 +298,9 @@ class CalculationMCPServer:
                 })
 
             self.add_patched_tool(fn, run_job, fn.__name__, is_async=True)
-            self.add_patched_tool(fn, submit_job, "submit_" + fn.__name__,
-                                  doc="Submit a job")
+            self.add_patched_tool(
+                fn, submit_job, "submit_" + fn.__name__, doc="Submit a job",
+                override_return_annotation=True)
             self.add_tool(query_job_status)
             self.add_tool(terminate_job)
             self.add_tool(get_job_results)
@@ -284,9 +310,10 @@ class CalculationMCPServer:
     def run(self, **kwargs):
         if os.environ.get("DP_AGENT_RUNNING_MODE") in ["1", "true"]:
             return
-        async def health_check(request) :
+
+        async def health_check(request):
             return JSONResponse({"status": "ok"})
-        
+
         self.mcp._custom_starlette_routes.append(
             Route(
                 "/health",

dp/agent/server/executor/dispatcher_executor.py

@@ -9,6 +9,7 @@ from pathlib import Path
 
 import jsonpickle
 from dpdispatcher import Machine, Resources, Task, Submission
+from dpdispatcher.utils.job_status import JobStatus
 
 from .base_executor import BaseExecutor
 from .... import __path__
@@ -33,6 +34,50 @@ def get_source_code(fn):
     return "".join(pre_lines + source_lines) + "\n"
 
 
+def get_func_def_script(fn):
+    script = ""
+    packages = []
+    fn_name = fn.__name__
+    module_name = fn.__module__
+    module = sys.modules[module_name]
+    if getattr(module, fn_name, None) is not fn:
+        # cannot import from module, maybe a local function
+        import cloudpickle
+        packages.extend(cloudpickle.__path__)
+        script += "import cloudpickle\n"
+        script += "%s = cloudpickle.loads(%s)\n" % \
+            (fn_name, cloudpickle.dumps(fn))
+    elif module_name in ["__main__", "__mp_main__"]:
+        if hasattr(module, "__file__"):
+            name = os.path.splitext(os.path.basename(module.__file__))[0]
+            if getattr(module, "__package__", None):
+                package = module.__package__
+                package_name = package.split('.')[0]
+                module = importlib.import_module(package_name)
+                packages.extend(module.__path__)
+                script += "from %s.%s import %s\n" % (
+                    package, name, fn_name)
+            else:
+                packages.append(module.__file__)
+                script += "from %s import %s\n" % (name, fn_name)
+        else:
+            # cannot get file of __main__, maybe in the interactive mode
+            import cloudpickle
+            packages.extend(cloudpickle.__path__)
+            script += "import cloudpickle\n"
+            script += "%s = cloudpickle.loads(%s)\n" % \
+                (fn_name, cloudpickle.dumps(fn))
+    else:
+        package_name = module_name.split('.')[0]
+        module = importlib.import_module(package_name)
+        if hasattr(module, "__path__"):
+            packages.extend(module.__path__)
+        elif hasattr(module, "__file__"):
+            packages.append(module.__file__)
+        script += "from %s import %s\n" % (module_name, fn_name)
+    return script, packages
+
+
 class DispatcherExecutor(BaseExecutor):
     def __init__(
             self,
@@ -86,24 +131,8 @@ class DispatcherExecutor(BaseExecutor):
     def submit(self, fn, kwargs):
         script = ""
         fn_name = fn.__name__
-        module_name = fn.__module__
-        import_func_line = None
-        if module_name in ["__main__", "__mp_main__"]:
-            module = sys.modules[module_name]
-            if hasattr(module, "__file__"):
-                self.python_packages.append(module.__file__)
-                name = os.path.splitext(os.path.basename(module.__file__))[0]
-                import_func_line = "from %s import %s\n" % (name, fn_name)
-            else:
-                script += get_source_code(fn)
-        else:
-            package_name = module_name.split('.')[0]
-            module = importlib.import_module(package_name)
-            if hasattr(module, "__path__"):
-                self.python_packages.extend(module.__path__)
-            elif hasattr(module, "__file__"):
-                self.python_packages.append(module.__file__)
-            import_func_line = "from %s import %s\n" % (module_name, fn_name)
+        func_def_script, packages = get_func_def_script(fn)
+        self.python_packages.extend(packages)
 
         script += "import asyncio, jsonpickle, os\n"
         script += "from pathlib import Path\n\n"
@@ -112,8 +141,9 @@ class DispatcherExecutor(BaseExecutor):
         script += "    kwargs = jsonpickle.loads(r'''%s''')\n" % \
                   jsonpickle.dumps(kwargs)
         script += "    try:\n"
-        if import_func_line is not None:
-            script += "        " + import_func_line
+        for line in func_def_script.splitlines():
+            if line:
+                script += "        " + line + "\n"
         if inspect.iscoroutinefunction(fn):
             script += "        results = asyncio.run(%s(**kwargs))\n" % fn_name
         else:
@@ -198,17 +228,23 @@ class DispatcherExecutor(BaseExecutor):
         submission = Submission.deserialize(
             submission_dict=json.loads(content))
         submission.update_submission_state()
-        if not submission.check_all_finished():
+        if not submission.check_all_finished() and not any(
+            job.job_state in [JobStatus.terminated, JobStatus.unknown,
+                              JobStatus.unsubmitted]
+                for job in submission.belonging_jobs):
             return "Running"
         try:
             submission.run_submission(exit_on_submit=True)
         except Exception as e:
             logger.error(e)
             return "Failed"
-        if os.path.isfile("results.txt"):
-            return "Succeeded"
+        if submission.check_all_finished():
+            if os.path.isfile("results.txt"):
+                return "Succeeded"
+            else:
+                return "Failed"
         else:
-            return "Failed"
+            return "Running"
 
     def terminate(self, job_id):
         machine = Machine.load_from_dict(self.machine)

dp/agent/server/storage/http_storage.py

@@ -51,4 +51,4 @@ class HTTPStorage(BaseStorage):
 
 
 class HTTPSStorage(HTTPStorage):
-    scheme = "https"
+    scheme = "https"

dp/agent/server/utils.py

@@ -1,26 +1,10 @@
-import inspect
 import logging
 import traceback
-from collections.abc import Sequence
-from typing import Annotated, Any, List, Optional
+from typing import Any
 
-import jsonpickle
 import mcp
-from mcp.server.fastmcp.exceptions import InvalidSignature
-from mcp.server.fastmcp.utilities.func_metadata import (
-    ArgModelBase,
-    _get_typed_annotation,
-    FuncMetadata,
-)
-from mcp.server.fastmcp.utilities.types import Image
-from mcp.types import (
-    EmbeddedResource,
-    ImageContent,
-    TextContent,
-)
-from pydantic import Field, WithJsonSchema, create_model
-from pydantic.fields import FieldInfo
-from pydantic_core import PydanticUndefined
+from mcp.types import TextContent
+from pydantic import BaseModel
 
 
 def get_logger(name, level="INFO",
@@ -34,110 +18,34 @@ def get_logger(name, level="INFO",
     return logger
 
 
-def get_metadata(
-    func_name: str,
-    parameters: List[inspect.Parameter],
-    skip_names: Sequence[str] = (),
-    globalns: dict = {},
-) -> FuncMetadata:
-    dynamic_pydantic_model_params: dict[str, Any] = {}
-    for param in parameters:
-        if param.name.startswith("_"):
-            raise InvalidSignature(
-                f"Parameter {param.name} of {func_name} cannot start with '_'"
-            )
-        if param.name in skip_names:
-            continue
-        annotation = param.annotation
-
-        # `x: None` / `x: None = None`
-        if annotation is None:
-            annotation = Annotated[
-                None,
-                Field(
-                    default=param.default
-                    if param.default is not inspect.Parameter.empty
-                    else PydanticUndefined
-                ),
-            ]
-
-        # Untyped field
-        if annotation is inspect.Parameter.empty:
-            annotation = Annotated[
-                Any,
-                Field(),
-                # 🤷
-                WithJsonSchema({"title": param.name, "type": "string"}),
-            ]
-
-        field_info = FieldInfo.from_annotated_attribute(
-            _get_typed_annotation(annotation, globalns),
-            param.default
-            if param.default is not inspect.Parameter.empty
-            else PydanticUndefined,
-        )
-        dynamic_pydantic_model_params[param.name] = (
-            field_info.annotation, field_info)
-        continue
-
-    arguments_model = create_model(
-        f"{func_name}Arguments",
-        **dynamic_pydantic_model_params,
-        __base__=ArgModelBase,
-    )
-    resp = FuncMetadata(arg_model=arguments_model)
-    return resp
-
-
-def convert_to_content(
-    result: Any,
-    job_info: Optional[dict] = None,
-) -> Sequence[TextContent | ImageContent | EmbeddedResource]:
-    """Convert a result to a sequence of content objects."""
-    other_contents = []
-    if isinstance(result, Image):
-        other_contents.append(result.to_image_content())
-        result = None
-
-    if isinstance(result, TextContent | ImageContent | EmbeddedResource):
-        other_contents.append(result)
-        result = None
-
-    if isinstance(result, list | tuple):
-        for item in result.copy():
-            if isinstance(item, Image):
-                other_contents.append(item.to_image_content())
-                result.remove(item)
-            elif isinstance(
-                    result, TextContent | ImageContent | EmbeddedResource):
-                other_contents.append(item)
-                result.remove(item)
-
-    if isinstance(result, dict):
-        for key, value in list(result.items()):
-            if isinstance(value, Image):
-                other_contents.append(value.to_image_content())
-                del result[key]
-            elif isinstance(
-                    value, TextContent | ImageContent | EmbeddedResource):
-                other_contents.append(value)
-                del result[key]
-
-    if not isinstance(result, str):
-        result = jsonpickle.dumps(result)
-
-    return [TextContent(type="text", text=result, job_info=job_info)] \
-        + other_contents
+class JobResult(BaseModel):
+    result: Any
+    job_info: dict
 
 
 class Tool(mcp.server.fastmcp.tools.Tool):
     """
     Workaround MCP server cannot print traceback
-    Remove this if MCP has proper support
+    Add job info to first unstructured content
     """
     async def run(self, *args, **kwargs):
         try:
-            return await super().run(*args, **kwargs)
+            kwargs["convert_result"] = False
+            result = await super().run(*args, **kwargs)
+            if isinstance(result, JobResult):
+                job_info = result.job_info
+                result = self.fn_metadata.convert_result(result.result)
+                if isinstance(result, tuple) and len(result) == 2:
+                    unstructured_content, _ = result
+                else:
+                    unstructured_content = result
+                if len(unstructured_content) == 0:
+                    unstructured_content.append(
+                        TextContent(type="text", text="null"))
+                unstructured_content[0].job_info = job_info
+            else:
+                result = self.fn_metadata.convert_result(result)
+            return result
         except Exception as e:
             traceback.print_exc()
             raise e

bohr_agent_sdk-0.1.102.dist-info/METADATA

@@ -1,228 +0,0 @@
-Metadata-Version: 2.4
-Name: bohr-agent-sdk
-Version: 0.1.102
-Summary: SDK for scientific agents
-Home-page: https://github.com/dptech-corp/bohr-agent-sdk/
-Author: DP Technology
-Maintainer-email: liupeng <liupeng@dp.tech>, zjgemi <liuxzj@dp.tech>
-License: MIT
-Project-URL: Homepage, https://github.com/dptech-corp/bohr-agent-sdk
-Project-URL: repository, https://github.com/dptech-corp/bohr-agent-sdk
-Project-URL: Bug Reports, https://github.com/dptech-corp/bohr-agent-sdk/issues
-Keywords: agent SDK,AI for science
-Classifier: License :: OSI Approved :: MIT License
-Classifier: Programming Language :: Python :: 3
-Classifier: Programming Language :: Python :: 3.10
-Classifier: Programming Language :: Python :: 3.11
-Classifier: Programming Language :: Python :: 3.12
-Classifier: Programming Language :: Python :: 3.13
-Requires-Python: >=3.10
-Description-Content-Type: text/markdown
-Requires-Dist: click>=8.0.0
-Requires-Dist: mcp
-Requires-Dist: python-dotenv>=1.0.0
-Requires-Dist: typing-extensions>=4.8.0
-Requires-Dist: dpdispatcher>=0.6.8
-Requires-Dist: lbg>=1.2.29
-Requires-Dist: jsonpickle>=3.0.3
-Requires-Dist: psutil>=5.9.6
-Requires-Dist: paho-mqtt>=2.1.0
-Requires-Dist: redis>=6.2.0
-Requires-Dist: twine>=6.1.0
-Requires-Dist: build>=1.2.2.post1
-Requires-Dist: watchdog>=6.0.0
-Requires-Dist: fastapi>=0.116.0
-Requires-Dist: bohrium-open-sdk==0.1.5
-Provides-Extra: device
-Requires-Dist: pywinauto-recorder>=0.1.0; extra == "device"
-Provides-Extra: cloud
-Requires-Dist: paho-mqtt>=1.6.1; extra == "cloud"
-Requires-Dist: redis>=5.0.1; extra == "cloud"
-Requires-Dist: aiohttp>=3.9.1; extra == "cloud"
-Provides-Extra: dev
-Requires-Dist: pytest>=7.4.0; extra == "dev"
-Requires-Dist: pytest-asyncio>=0.23.0; extra == "dev"
-Requires-Dist: pytest-cov>=4.1.0; extra == "dev"
-Requires-Dist: black>=23.11.0; extra == "dev"
-Requires-Dist: isort>=5.12.0; extra == "dev"
-Requires-Dist: mypy>=1.7.0; extra == "dev"
-Requires-Dist: pylint>=3.0.0; extra == "dev"
-Requires-Dist: google-adk; extra == "dev"
-Provides-Extra: docs
-Requires-Dist: sphinx>=7.2.0; extra == "docs"
-Requires-Dist: sphinx-rtd-theme>=1.3.0; extra == "docs"
-Provides-Extra: all
-Requires-Dist: bohr-agent-sdk[bohrium,cloud,dev,device,dispatcher,docs]; extra == "all"
-Dynamic: home-page
-Dynamic: requires-python
-
-# Bohrium Science Agent SDK
-
-这是DP Tech的Bohrium Science Agent SDK，提供了一个命令行工具dp-agent，用于管理科学计算任务。同时提供了Python SDK用于开发自定义的科学计算应用。
-
-## 安装
-
-```bash
-pip install bohr-agent-sdk -i https://pypi.org/simple --upgrade
-```
-
-## CLI 使用方法
-
-安装后，您可以使用以下命令：
-
-### 获取资源
-
-```bash
-# 获取基础代码结构
-dp-agent fetch scaffolding --type=calculation/device
-
-# 获取配置文件
-dp-agent fetch config
-```
-
-`fetch config` 命令会下载 .env 配置文件并替换部分动态变量，如 MQTT_DEVICE_ID。
-注意：出于安全考虑，此功能仅在内网环境可用。其他环境需要手动配置。
-
-### 运行命令
-
-```bash
-# 运行实验环境
-dp-agent run tool device
-
-# 运行云环境
-dp-agent run tool cloud
-
-# 运行计算环境
-dp-agent run tool calculation
-
-# 运行代理
-dp-agent run agent --config config.json
-
-# 调试模式
-dp-agent run debug
-```
-
-## SDK 快速入门
-
-Bohrium Science Agent SDK 提供了两种主要的开发模式：实验室模式（Lab）和云模式（Cloud）。
-
-### 基础结构
-
-安装完成并运行 `dp-agent fetch scaffolding` 后，您将获得以下基础项目结构：
-
-```
-your-project/
-├── lab/                # 实验室模式相关代码
-│   ├── __init__.py
-│   └── tescan_device.py  # 设备控制示例
-├── cloud/              # 云模式相关代码
-│   └── __init__.py
-└── main.py            # 主程序入口
-```
-
-### 实验室模式开发
-
-实验室模式主要用于控制本地实验设备。以下是一个基于 Tescan 设备的示例：
-
-```python
-from typing import Dict, TypedDict
-from dp.agent.device.device import Device, action, BaseParams, SuccessResult
-
-class TakePictureParams(BaseParams):
-    """拍照参数"""
-    horizontal_width: str
-
-class PictureData(TypedDict):
-    """照片数据"""
-    image_id: str
-
-class PictureResult(SuccessResult):
-    """拍照结果"""
-    data: PictureData
-
-class MyDevice(Device):
-    device_name = "my_device"
-    
-    @action("take_picture")
-    def take_picture(self, params: TakePictureParams) -> PictureResult:
-        """拍照动作
-        
-        Args:
-            params: 拍照参数
-                - horizontal_width: 图片水平宽度
-        """
-        hw = params.get("horizontal_width", "default")
-        return PictureResult(
-            message=f"Picture taken with {self.device_name}",
-            data={"image_id": "image_123"}
-        )
-```
-
-### 云端开发
-
-云模式基于 MCP (Message Control Protocol) 实现，用于处理远程设备控制和任务调度。register_mcp_tools 通过 python 的自省和反射机制实现了设备控制的自动注册，无需重复实现操作定义。
-以下展示如何创建设备并注册到 MCP 服务器：
-
-```python
-"""
-Example of using the bohr-agent-sdk cloud functionality.
-"""
-import signal
-import sys
-from dp.agent.cloud import mcp, get_mqtt_cloud_instance
-from dp.agent.device.device import TescanDevice, register_mcp_tools
-
-def signal_handler(sig, frame):
-    """Handle SIGINT signal to gracefully shutdown."""
-    print("Shutting down...")
-    get_mqtt_cloud_instance().stop()
-    sys.exit(0)
-
-def main():
-    """Start the cloud services."""
-    print("Starting Tescan Device Twin Cloud Services...")
-    
-    # Register signal handler
-    signal.signal(signal.SIGINT, signal_handler)
-    
-    # Create device instance
-    device = TescanDevice(mcp, device)
-    
-    # Register device tools
-    register_mcp_tools(device)
-    
-    # Start MCP server
-    print("Starting MCP server...")
-    mcp.run(transport="sse")
-
-if __name__ == "__main__":
-    main()
-```
-
-
-### 配置说明
-
-在 `.env` 文件中配置必要的环境变量：
-
-```
-MQTT_INSTANCE_ID=your_instance_id
-MQTT_ENDPOINT=your_endpoint
-MQTT_DEVICE_ID=your_device_id
-MQTT_GROUP_ID=your_group_id
-MQTT_AK=your_access_key
-MQTT_SK=your_secret_key
-```
-
-### 主要功能
-
-- 设备控制接口（Lab模式）
-  - 设备初始化
-  - 命令执行
-  - 状态监控
-  
-- 云端任务处理（Cloud模式）
-  - 任务队列管理
-  - 计算资源调度
-  - 结果回传
-
-更详细的API文档请参考代码中的注释。