π The First Self-Improving agents with RL / Prompting Optimization
Python agentaigptmulti-agentpythonreinforcement-learningself-improvementsociety-of-mindswarm-intelligence
from abc import ABC, abstractmethod
from typing import List, Union, Optional
from swarm.llm.format import Message
class LLM(ABC):
DEFAULT_MAX_TOKENS = 1000
DEFAULT_TEMPERATURE = 0.2
DEFUALT_NUM_COMPLETIONS = 1
@abstractmethod
async def agen(
self,
messages: List[Message],
max_tokens: Optional[int] = None,
temperature: Optional[float] = None,
num_comps: Optional[int] = None,
) -> Union[List[str], str]:
pass
@abstractmethod
def gen(
self,
messages: List[Message],
max_tokens: Optional[int] = None,
temperature: Optional[float] = None,
num_comps: Optional[int] = None,
) -> Union[List[str], str]:
pass