Source code for langchain_community.tools.memorize.tool
from abc import abstractmethod
from typing import Any, Optional, Protocol, Sequence, runtime_checkable
from langchain_core.callbacks import (
AsyncCallbackManagerForToolRun,
CallbackManagerForToolRun,
)
from langchain_core.pydantic_v1 import Field
from langchain_core.tools import BaseTool
from langchain_community.llms.gradient_ai import TrainResult
[docs]@runtime_checkable
class TrainableLLM(Protocol):
"""Protocol for trainable language models."""
[docs] @abstractmethod
def train_unsupervised(
self,
inputs: Sequence[str],
**kwargs: Any,
) -> TrainResult:
...
[docs] @abstractmethod
async def atrain_unsupervised(
self,
inputs: Sequence[str],
**kwargs: Any,
) -> TrainResult:
...
[docs]class Memorize(BaseTool):
"""Tool that trains a language model."""
name: str = "memorize"
description: str = (
"Useful whenever you observed novel information "
"from previous conversation history, "
"i.e., another tool's action outputs or human comments. "
"The action input should include observed information in detail, "
"then the tool will fine-tune yourself to remember it."
)
llm: TrainableLLM = Field()
def _run(
self,
information_to_learn: str,
run_manager: Optional[CallbackManagerForToolRun] = None,
) -> str:
train_result = self.llm.train_unsupervised((information_to_learn,))
return f"Train complete. Loss: {train_result['loss']}"
async def _arun(
self,
information_to_learn: str,
run_manager: Optional[AsyncCallbackManagerForToolRun] = None,
) -> str:
train_result = await self.llm.atrain_unsupervised((information_to_learn,))
return f"Train complete. Loss: {train_result['loss']}"