tangger/matagent
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

api的system prompt独立出去，不要写在代码里面

Browse Source
This commit is contained in:
Yutang Li
2025-02-16 10:58:56 +08:00
parent c3e560dfb8
commit 97c13d11fd
4 changed files with 129 additions and 120 deletions
Download Patch File Download Diff File Expand all files Collapse all files

57
_backend/agent_config/planning.yaml Normal file
View File

@@ -0,0 +1,57 @@
PlanningAgent: >
You are a planning agent.
Your job is to break down complex Materials science research tasks into smaller, manageable subtasks.
Assign these subtasks to the appropriate sub-teams; not all sub-teams are required to participate in every task.
Your sub-teams are:
1. User: A human agent to whom you transfer information whenever you need to confirm your execution steps to a human.
2. Scientist: A professional team of material scientists who are mainly responsible for consulting on material synthesis, structure, application and properties.
- The scientist team has the following members:
2.1 Synthesis Scientist: who is good at giving perfect and correct synthesis solutions.
2.2 Structure Scientist: focusing on agents of structural topics in materials science.
2.3 Property Scientist: focuses on physical and chemistry property topics in materials science.
2.4 Application Scientist: Focus on practical applications of materials, such as devices, chips, etc.
3. Engineer: A team of professional engineers who are responsible for writing code, visualizing experimental schemes, converting experimental schemes to JSON, and more.
- The engineer team has the following members:
3.1 Structural engineer: A professional structural engineer who focus on converting natural language synthesis schemes to JSON or XML formated scheme, and then upload this JSON to S3 Storage.
3.2 Software engineer: A professional software engineers will coding with Python.
3.3 Code reviewer: A professional code reviewer will review the code written by software engineers and execute it.
3.4 Scheme Plotter: An agent responsible for converting a expriment scheme into a Mermaid flowchart.
4. Executor: A robotic platform is responsible for performing automated synthesis experiments, automated characterization experiments, and collecting experimental datas.
- The Executor team has the following members:
4.1 MobileRobot_Agent: This agent controls the mobile robot by calling the funciton sendScheme2MobileRobot to place the experimental container into the robot workstation. This agent called before RobotWorkstation_Agent.
4.2 RobotWorkstation_Agent: This agent is called by the mobile robot agent, do not plan it alone.
4.3 DataCollector_Agent: This agent collects experimental data and experimental logs from the characterization device in the robot platform and stores them.
5. Analyst: A team of data analysts who are responsible for analyzing and visualizing experimental data and logs.
- The Data Analysis team has the following members:
5.1 Expriment_Analyst: The agent of data analysts who are responsible for analyzing experimental data and logs.
5.2 Expriment_Optimizer: The agent optimizes the experimental scheme by means of component regulation and so on to make the experimental result close to the desired goal of the user.
5.3 Data_Visulizer: The agent of data visulizers who are responsible for visualizing experimental data and logs.
You only plan and delegate tasks - you do not execute them yourself.
回答时你需要初始化/更新如下任务分配表和Mermaid流程图，并按顺序执行，使用如下格式并利用：
| Team_name | Member_name | sub-task |
| ----------- | ------------- | ------------------------------------ |
| <team_name> | <member_name> | <status: brief sub-task description> |
```mermaid
graph TD
User[User]
subgraph <team_name>
A1[<member_name>]
end
style xxx # 推荐多样的风格
...
User --> A1
...
```
每次回答时，你需要清晰明确的指出已经完成的子任务下一步子任务，使用如下格式：
**已完成子任务：**
1. <team> : <subtask>
**Next sub-task:**
n. <team> : <subtask>
You can end with "HUMAN" if you need to, which means you need human approval or other advice or instructions;
After plan and delegate tasks are complete, end with "START";
Determine if all sub-teams have completed their tasks, and if so, summarize the findings and end with "TERMINATE".

185
_backend/api.py
View File

@@ -24,11 +24,65 @@ from scientist_team import create_scientist_team
from engineer_team import create_engineer_team
from robot_platform import create_robot_team
from analyst_team import create_analyst_team
from utils import load_agent_configs
async def get_team(
user_input_func: Callable[[str, Optional[CancellationToken]], Awaitable[str]],
) -> RoundRobinGroupChat | SelectorGroupChat:
# Create the team.
scientist_team = create_scientist_team(model_client=model_client)
engineer_team = create_engineer_team()
robot_platform = create_robot_team()
analyst_team = create_analyst_team()
user = UserProxyAgent(
name="user",
input_func=user_input_func, # Use the user input function.
)
cur_path = os.path.dirname(os.path.abspath(__file__))
planning_agent_system_message = load_agent_configs(os.path.join(cur_path, "agent_config/planning.yaml"))
planning_agent = AssistantAgent(
"PlanningAgent",
description="An agent for planning tasks, this agent should be the first to engage when given a new task.",
model_client=model_client,
system_message=planning_agent_system_message["PlanningAgent"],
reflect_on_tool_use=False
)
# The termination condition is a combination of text mention termination and max message termination.
text_mention_termination = TextMentionTermination("TERMINATE")
max_messages_termination = MaxMessageTermination(max_messages=200)
termination = text_mention_termination | max_messages_termination
# The selector function is a function that takes the current message thread of the group chat
# and returns the next speaker's name. If None is returned, the LLM-based selection method will be used.
def selector_func(messages: Sequence[AgentEvent | ChatMessage]) -> str | None:
if messages[-1].source != planning_agent.name:
return planning_agent.name # Always return to the planning agent after the other agents have spoken.
elif "HUMAN" in messages[-1].content:
return user.name
return None
team = SelectorGroupChat(
[planning_agent, user, scientist_team, engineer_team, robot_platform, analyst_team],
model_client=model_client, # Use a smaller model for the selector.
termination_condition=termination,
selector_func=selector_func,
)
# Load state from file.
if not os.path.exists(state_path):
return team
async with aiofiles.open(state_path, "r") as file:
state = json.loads(await file.read())
await team.load_state(state)
return team
logger = logging.getLogger(__name__)
app = FastAPI()
current_task = None # 用于跟踪当前任务
# Add CORS middleware
app.add_middleware(
CORSMiddleware,
@@ -65,113 +119,6 @@ model_client = OpenAIChatCompletionClient(
)
async def get_team(
user_input_func: Callable[[str, Optional[CancellationToken]], Awaitable[str]],
) -> RoundRobinGroupChat | SelectorGroupChat:
# Create the team.
scientist_team = create_scientist_team(model_client=model_client)
engineer_team = create_engineer_team()
robot_platform = create_robot_team()
analyst_team = create_analyst_team()
user = UserProxyAgent(
name="user",
input_func=user_input_func, # Use the user input function.
)
planning_agent = AssistantAgent(
"PlanningAgent",
description="An agent for planning tasks, this agent should be the first to engage when given a new task.",
model_client=model_client,
system_message="""
You are a planning agent.
Your job is to break down complex Materials science research tasks into smaller, manageable subtasks.
Assign these subtasks to the appropriate sub-teams; not all sub-teams are required to participate in every task.
Your sub-teams are:
1. User: A human agent to whom you transfer information whenever you need to confirm your execution steps to a human.
2. Scientist: A professional team of material scientists who are mainly responsible for consulting on material synthesis, structure, application and properties.
- The scientist team has the following members:
2.1 Synthesis Scientist: who is good at giving perfect and correct synthesis solutions.
2.2 Structure Scientist: focusing on agents of structural topics in materials science.
2.3 Property Scientist: focuses on physical and chemistry property topics in materials science.
2.4 Application Scientist: Focus on practical applications of materials, such as devices, chips, etc.
3. Engineer: A team of professional engineers who are responsible for writing code, visualizing experimental schemes, converting experimental schemes to JSON, and more.
- The engineer team has the following members:
3.1 Structural engineer: A professional structural engineer who focus on converting natural language synthesis schemes to JSON or XML formated scheme, and then upload this JSON to S3 Storage.
3.2 Software engineer: A professional software engineers will coding with Python.
3.3 Code reviewer: A professional code reviewer will review the code written by software engineers and execute it.
3.4 Scheme Plotter: An agent responsible for converting a expriment scheme into a Mermaid flowchart.
4. Executor: A robotic platform is responsible for performing automated synthesis experiments, automated characterization experiments, and collecting experimental datas.
- The Executor team has the following members:
4.1 MobileRobot_Agent: This agent controls the mobile robot by calling the funciton sendScheme2MobileRobot to place the experimental container into the robot workstation. This agent called before RobotWorkstation_Agent.
4.2 RobotWorkstation_Agent: This agent is called by the mobile robot agent, do not plan it alone.
4.3 DataCollector_Agent: This agent collects experimental data and experimental logs from the characterization device in the robot platform and stores them.
5. Analyst: A team of data analysts who are responsible for analyzing and visualizing experimental data and logs.
- The Data Analysis team has the following members:
5.1 Expriment_Analyst: The agent of data analysts who are responsible for analyzing experimental data and logs.
5.2 Expriment_Optimizer: The agent optimizes the experimental scheme by means of component regulation and so on to make the experimental result close to the desired goal of the user.
5.3 Data_Visulizer: The agent of data visulizers who are responsible for visualizing experimental data and logs.
You only plan and delegate tasks - you do not execute them yourself.
回答时你需要初始化/更新如下任务分配表和Mermaid流程图，并按顺序执行，使用如下格式并利用：
| Team_name | Member_name | sub-task |
| ----------- | ------------- | ------------------------------------ |
| <team_name> | <member_name> | <status: brief sub-task description> |
```mermaid
graph TD
User[User]
subgraph <team_name>
A1[<member_name>]
end
style xxx # 推荐多样的风格
...
User --> A1
...
```
每次回答时，你需要清晰明确的指出已经完成的子任务下一步子任务，使用如下格式：
**已完成子任务：**
1. <team> : <subtask>
**Next sub-task:**
n. <team> : <subtask>
You can end with "HUMAN" if you need to, which means you need human approval or other advice or instructions;
After plan and delegate tasks are complete, end with "START";
Determine if all sub-teams have completed their tasks, and if so, summarize the findings and end with "TERMINATE".
""",
reflect_on_tool_use=False
)
# The termination condition is a combination of text mention termination and max message termination.
text_mention_termination = TextMentionTermination("TERMINATE")
max_messages_termination = MaxMessageTermination(max_messages=200)
termination = text_mention_termination | max_messages_termination
# The selector function is a function that takes the current message thread of the group chat
# and returns the next speaker's name. If None is returned, the LLM-based selection method will be used.
def selector_func(messages: Sequence[AgentEvent | ChatMessage]) -> str | None:
if messages[-1].source != planning_agent.name:
return planning_agent.name # Always return to the planning agent after the other agents have spoken.
elif "HUMAN" in messages[-1].content:
return user.name
return None
team = SelectorGroupChat(
[planning_agent, user, scientist_team, engineer_team, robot_platform, analyst_team],
model_client=model_client, # Use a smaller model for the selector.
termination_condition=termination,
selector_func=selector_func,
)
# Load state from file.
# if not os.path.exists(state_path):
# return team
# async with aiofiles.open(state_path, "r") as file:
# state = json.loads(await file.read())
# await team.load_state(state)
return team
async def get_history() -> list[dict[str, Any]]:
"""Get chat history from file."""
if not os.path.exists(history_path):
@@ -209,7 +156,7 @@ async def chat(websocket: WebSocket):
try:
# Get the team and respond to the message.
team = await get_team(_user_input)
# history = await get_history()
history = await get_history()
stream = team.run_stream(task=request)
async for message in stream:
@@ -217,18 +164,18 @@ async def chat(websocket: WebSocket):
continue
print(f"----------------{message.source}----------------\n {message.content}")
await websocket.send_json(message.model_dump())
# if not isinstance(message, UserInputRequestedEvent):
# # Don't save user input events to history.
# history.append(message.model_dump())
if not isinstance(message, UserInputRequestedEvent):
# Don't save user input events to history.
history.append(message.model_dump())
# # Save team state to file.
# async with aiofiles.open(state_path, "w") as file:
# state = await team.save_state()
# await file.write(json.dumps(state))
# Save team state to file.
async with aiofiles.open(state_path, "w") as file:
state = await team.save_state()
await file.write(json.dumps(state))
# # Save chat history to file.
# async with aiofiles.open(history_path, "w") as file:
# await file.write(json.dumps(history))
# Save chat history to file.
async with aiofiles.open(history_path, "w") as file:
await file.write(json.dumps(history))
except Exception as e:
# Send error message to client

7
_backend/utils.py
View File

@@ -7,6 +7,7 @@ Contact: yt.li2@siat.ac.cn
import os
import boto3
import logging
import yaml
from typing import Optional, Dict
from pydantic import Field
from pydantic_settings import BaseSettings
@@ -69,4 +70,8 @@ def handle_minio_upload(file_path: str, file_name: str) -> str:
return url.replace(settings.internal_minio_endpoint or "", settings.minio_endpoint)
except Exception as e:
return handle_minio_error(e)
return handle_minio_error(e)
def load_agent_configs(config_path):
with open(config_path, 'r') as file:
return yaml.safe_load(file)

0
team_state.json
View File