langchain/templates/solo-performance-prompting-agent/solo_performance_prompting_agent/agent.py
Sayandip 8dbbcf0b6c
Adding a template for Solo Performance Prompting Agent (#12627)
**Description:** This template creates an agent that transforms a single
LLM into a cognitive synergist by engaging in multi-turn
self-collaboration with multiple personas.
**Tag maintainer:** @hwchase17

---------

Co-authored-by: Sayandip Sarkar <sayandip.sarkar@skypointcloud.com>
Co-authored-by: Erick Friis <erick@langchain.dev>
2023-11-01 08:10:07 -07:00

39 lines
1.1 KiB
Python

from langchain.agents import AgentExecutor
from langchain.agents.format_scratchpad import format_xml
from langchain.llms import OpenAI
from langchain.pydantic_v1 import BaseModel
from langchain.tools import DuckDuckGoSearchRun
from langchain.tools.render import render_text_description
from solo_performance_prompting_agent.parser import parse_output
from solo_performance_prompting_agent.prompts import conversational_prompt
_model = OpenAI()
_tools = [DuckDuckGoSearchRun()]
_prompt = conversational_prompt.partial(
tools=render_text_description(_tools),
tool_names=", ".join([t.name for t in _tools]),
)
_llm_with_stop = _model.bind(stop=["</tool_input>", "</final_answer>"])
agent = (
{
"question": lambda x: x["question"],
"agent_scratchpad": lambda x: format_xml(x["intermediate_steps"]),
}
| _prompt
| _llm_with_stop
| parse_output
)
class AgentInput(BaseModel):
question: str
agent_executor = AgentExecutor(
agent=agent, tools=_tools, verbose=True, handle_parsing_errors=True
).with_types(input_type=AgentInput)
agent_executor = agent_executor | (lambda x: x["output"])