apunkt/MetaGPT
1
0
Fork 0
mirror of https://github.com/FoundationAgents/MetaGPT.git synced 2026-06-08 15:05:17 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

Add a function to analyze human responses.

Browse source
This commit is contained in:
黄伟韬 2024-09-26 13:42:43 +08:00
parent 128312be59
commit 955ec263c6
2 changed files with 31 additions and 15 deletions
Download patch file Download diff file Expand all files Collapse all files

24
metagpt/prompts/di/role_zero.py
View file

@ -119,18 +119,20 @@ END_COMMAND = """
```
"""
ASK_HUMAN_COMMAND = """
```json
[
{
"command_name": "RoleZero.ask_human",
"args": {
"question": "I'm a little uncertain about the next step, could you provide me with some guidance? If you want to stop, please include <STOP> in your response."
}
}
]
```
SUMMARY_TROUBLE = """
You are repeating the same action.
Please summarize what is confusing or troubling you, and tell me directly.
Do not output any commands.
Your response must be in {language} and should be within 30 words.
"""
DUPULICATE_QUESTTION_FORMAT = """
I meet the following trouble:
{trouble}
Could you provide me with some guidance?
If want to stop, please contain \"<STOP>\" in your guidance.
"""
ASK_HUMAN_COMMAND = [{"command_name": "RoleZero.ask_human", "args": {"question": ""}}]
JSON_REPAIR_PROMPT = """
## json data
{json_data}

22
metagpt/roles/di/role_zero.py
View file

@ -21,6 +21,7 @@ from metagpt.prompts.di.role_zero import (
ASK_HUMAN_COMMAND,
CMD_PROMPT,
DETECT_LANGUAGE_PROMPT,
DUPULICATE_QUESTTION_FORMAT,
END_COMMAND,
JSON_REPAIR_PROMPT,
QUICK_RESPONSE_SYSTEM_PROMPT,
@ -31,6 +32,7 @@ from metagpt.prompts.di.role_zero import (
REPORT_TO_HUMAN_PROMPT,
ROLE_INSTRUCTION,
SUMMARY_PROMPT,
SUMMARY_TROUBLE,
SYSTEM_PROMPT,
)
from metagpt.roles import Role
@ -120,7 +122,7 @@ class RoleZero(Role):
"Plan.append_task": self.planner.plan.append_task,
"Plan.reset_task": self.planner.plan.reset_task,
"Plan.replace_task": self.planner.plan.replace_task,
"RoleZero.ask_human": self.ask_human,
"RoleZero.ask_human": self.request_and_analyze_human_response,
"RoleZero.reply_to_human": self.reply_to_human,
"SearchEnhancedQA.run": SearchEnhancedQA().run,
}
@ -386,7 +388,7 @@ class RoleZero(Role):
async def _check_duplicates(self, req: list[dict], command_rsp: str, check_window: int = 10):
past_rsp = [mem.content for mem in self.rc.memory.get(check_window)]
if command_rsp in past_rsp:
if command_rsp in past_rsp and '"command_name": "end"' not in command_rsp:
# Normal response with thought contents are highly unlikely to reproduce
# If an identical response is detected, it is a bad response, mostly due to LLM repeating generated content
# In this case, ask human for help and regenerate
@ -395,10 +397,15 @@ class RoleZero(Role):
# Hard rule to ask human for help
if past_rsp.count(command_rsp) >= 3:
if '"command_name": "Plan.finish_current_task",' in command_rsp:
# Detect the deplicate of "Plan.finish_current_task" command, use command "end" to finish the task
# Detect the duplicate of the 'Plan.finish_current_task' command, and use the 'end' command to finish the task.
logger.warning(f"Duplicate response detected: {command_rsp}")
return END_COMMAND
return ASK_HUMAN_COMMAND
trouble_content = await self.llm.aask(
req + [UserMessage(content=SUMMARY_TROUBLE.format(language=self.respond_language))]
)
ASK_HUMAN_COMMAND[0]["args"]["question"] = DUPULICATE_QUESTTION_FORMAT.format(trouble=trouble_content)
ask_human_command = "```json\n" + json.dumps(ASK_HUMAN_COMMAND, indent=4, ensure_ascii=False) + "\n```"
return ask_human_command
# Try correction by self
logger.warning(f"Duplicate response detected: {command_rsp}")
regenerate_req = req + [UserMessage(content=REGENERATE_PROMPT)]
@ -557,6 +564,13 @@ class RoleZero(Role):
return "Not in MGXEnv, command will not be executed."
return await self.rc.env.ask_human(question, sent_from=self)
async def request_and_analyze_human_response(self, question: str) -> str:
"""This fucntion will call ask_human and then analyze and decorate the human response."""
human_response = await self.ask_human(question)
if "<STOP>" in human_response:
human_response += f"\nUser ask you to stop the current task immediately.\nIn your next response, use end command to stop the current task.\nLike:\n{END_COMMAND}"
return human_response
async def reply_to_human(self, content: str) -> str:
"""Reply to human user with the content provided. Use this when you have a clear answer or solution to the user's question."""
# NOTE: Can be overwritten in remote setting