mirror of
https://github.com/eigent-ai/eigent.git
synced 2026-05-24 05:26:42 +00:00
273 lines
8.3 KiB
Python
273 lines
8.3 KiB
Python
# ========= Copyright 2023-2026 @ CAMEL-AI.org. All Rights Reserved. =========
|
|
# Licensed under the Apache License, Version 2.0 (the "License");
|
|
# you may not use this file except in compliance with the License.
|
|
# You may obtain a copy of the License at
|
|
#
|
|
# http://www.apache.org/licenses/LICENSE-2.0
|
|
#
|
|
# Unless required by applicable law or agreed to in writing, software
|
|
# distributed under the License is distributed on an "AS IS" BASIS,
|
|
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
# See the License for the specific language governing permissions and
|
|
# limitations under the License.
|
|
# ========= Copyright 2023-2026 @ CAMEL-AI.org. All Rights Reserved. =========
|
|
|
|
from abc import ABC, abstractmethod
|
|
from typing import Any, Dict, List, Optional, Tuple
|
|
|
|
from camel.environments.models import Action, Observation, StepResult
|
|
from camel.extractors.base import BaseExtractor
|
|
from camel.logger import get_logger
|
|
|
|
logger = get_logger(__name__)
|
|
|
|
|
|
class MultiStepEnv(ABC):
|
|
r"""A multi-step environment for reinforcement learning with LLMs."""
|
|
|
|
def __init__(
|
|
self,
|
|
extractor: BaseExtractor,
|
|
max_steps: Optional[int] = None,
|
|
**kwargs,
|
|
) -> None:
|
|
r"""Initialize the environment.
|
|
|
|
Args:
|
|
extractor: Extractor to process LLM responses.
|
|
max_steps: Maximum steps per episode.
|
|
**kwargs: Additional environment parameters.
|
|
"""
|
|
self.extractor = extractor
|
|
self.max_steps = max_steps
|
|
self._metadata = kwargs
|
|
|
|
# State tracking
|
|
self._is_setup: bool = False
|
|
self._current_step: int = 0
|
|
self._episode_ended: bool = False
|
|
self._state: Dict[str, Any] = self._get_initial_state()
|
|
self._last_observation: Optional[Observation] = None
|
|
self._episode_history: List[Tuple[Observation, Action]] = []
|
|
|
|
async def setup(self) -> None:
|
|
r"""Set up the environment by initializing the verifier and extractor.
|
|
|
|
This method ensures that the environment is ready for interaction.
|
|
It sets up necessary components, including the verifier and extractor.
|
|
|
|
Raises:
|
|
Exception: If setup fails due to an internal error.
|
|
"""
|
|
|
|
if self._is_setup:
|
|
return
|
|
|
|
try:
|
|
await self.extractor.setup()
|
|
await self._setup()
|
|
self._is_setup = True
|
|
logger.info('Environment setup completed successfully')
|
|
except Exception as e:
|
|
logger.error(f'Failed to setup environment: {e}')
|
|
raise
|
|
|
|
async def _setup(self) -> None:
|
|
return
|
|
|
|
async def close(self) -> None:
|
|
r"""Clean up and close all resources used by the environment.
|
|
This method shuts down the verifier, calls the internal
|
|
close function that is implemented in any MultiStepEnv,
|
|
and ensures that the environment is properly closed.
|
|
|
|
Raises:
|
|
Exception: If an error occurs while closing the environment.
|
|
"""
|
|
if not self._is_setup:
|
|
return
|
|
|
|
try:
|
|
await self.extractor.cleanup()
|
|
|
|
await self._close()
|
|
|
|
self._is_setup = False
|
|
logger.info('Environment teardown completed successfully')
|
|
except Exception as e:
|
|
logger.error(f'Failed to teardown environment: {e}')
|
|
raise
|
|
|
|
async def _close(self) -> None:
|
|
return
|
|
|
|
async def reset(self) -> Observation:
|
|
r"""Reset the environment to an initial state.
|
|
|
|
Returns:
|
|
Observation: The initial observation for the episode.
|
|
|
|
Raises:
|
|
RuntimeError: If we fail to get the initial observation.
|
|
"""
|
|
|
|
if not self._is_setup:
|
|
logger.warning(
|
|
"reset() called on un-setup environment. Setting up..."
|
|
)
|
|
await self.setup()
|
|
|
|
# Reset state
|
|
self._current_step = 0
|
|
self._episode_ended = False
|
|
self._episode_history = []
|
|
self._state = self._get_initial_state()
|
|
|
|
# Get initial observation
|
|
observation = self._get_next_observation()
|
|
if observation is None:
|
|
raise RuntimeError("Failed to get initial observation")
|
|
|
|
self._last_observation = observation
|
|
|
|
return observation
|
|
|
|
async def step(
|
|
self, action: Action
|
|
) -> Tuple[Observation, float, bool, Dict[str, Any]]:
|
|
r"""Take a step in the environment using the given action.
|
|
|
|
This method updates the environment state based on the LLM's response,
|
|
computes rewards, checks if the episode is done, and based on that
|
|
gets the next or final observation.
|
|
|
|
Args:
|
|
action (Action): The action containing the LLM response.
|
|
|
|
Returns:
|
|
StepResult containing next observation, total reward, a dictionary
|
|
of rewards, done flag, and info.
|
|
|
|
Raises:
|
|
RuntimeError: If the environment is not set up, the episode has
|
|
ended, or there is no valid current observation.
|
|
"""
|
|
if self.max_steps and self._current_step >= self.max_steps:
|
|
return StepResult(
|
|
observation=self._get_terminal_observation(),
|
|
reward=0,
|
|
rewards_dict={},
|
|
done=True,
|
|
info={"reason": "max_steps_reached"},
|
|
).as_tuple()
|
|
|
|
if not self._is_setup:
|
|
raise RuntimeError("Environment not set up. Call setup() first.")
|
|
if self._episode_ended:
|
|
raise RuntimeError("Episode has ended. Call reset() first.")
|
|
if self._last_observation is None:
|
|
raise RuntimeError("No current observation. Call reset() first.")
|
|
|
|
self._current_step += 1
|
|
|
|
current_obs: Observation = self._last_observation
|
|
self._episode_history.append((current_obs, action))
|
|
|
|
# Update the environment state based on the action
|
|
await self._update_state(action)
|
|
|
|
# Compute rewards
|
|
total_reward, rewards_dict = await self.compute_reward()
|
|
|
|
# Check termination
|
|
done = self.is_done()
|
|
|
|
# Get next observation based on the updated state
|
|
next_obs = (
|
|
self._get_terminal_observation()
|
|
if done
|
|
else self._get_next_observation()
|
|
)
|
|
|
|
self._last_observation = next_obs
|
|
self._episode_ended = done
|
|
|
|
return StepResult(
|
|
observation=next_obs,
|
|
reward=total_reward,
|
|
rewards_dict=rewards_dict,
|
|
done=done,
|
|
info={
|
|
"extraction_result": await self.extractor.extract(
|
|
action.llm_response
|
|
),
|
|
"step": self._current_step,
|
|
"state": self._state, # Updated state
|
|
},
|
|
).as_tuple()
|
|
|
|
@abstractmethod
|
|
def _get_initial_state(self) -> Dict[str, Any]:
|
|
pass
|
|
|
|
@abstractmethod
|
|
async def _update_state(self, action: Action) -> None:
|
|
pass
|
|
|
|
@abstractmethod
|
|
def _get_next_observation(self) -> Observation:
|
|
pass
|
|
|
|
@abstractmethod
|
|
def _get_terminal_observation(self) -> Observation:
|
|
pass
|
|
|
|
@abstractmethod
|
|
async def compute_reward(
|
|
self,
|
|
) -> Tuple[float, Dict[str, float]]:
|
|
pass
|
|
|
|
def is_done(self) -> bool:
|
|
r"""Check if the episode should terminate.
|
|
|
|
This function terminates the episode if the maximum number of
|
|
steps is reached or if any other terminating criterion is met.
|
|
|
|
Returns:
|
|
bool: A boolean flag.
|
|
"""
|
|
|
|
# After too many steps
|
|
if self.max_steps and self._current_step >= self.max_steps:
|
|
return True
|
|
|
|
# Further termination logic can be implemented in subclass
|
|
if self._is_done():
|
|
return True
|
|
|
|
return False
|
|
|
|
@abstractmethod
|
|
def _is_done(self) -> bool:
|
|
pass
|
|
|
|
@property
|
|
def metadata(self) -> Dict[str, Any]:
|
|
r"""Retrieve the metadata of the environment.
|
|
|
|
This provides additional parameters and configuration details.
|
|
|
|
Returns:
|
|
Dict[str, Any]: A copy of the environment's metadata.
|
|
"""
|
|
return self._metadata.copy()
|
|
|
|
@property
|
|
def current_step(self) -> int:
|
|
r"""Get the current step number.
|
|
|
|
Returns:
|
|
int: The number of the step we are currently in.
|
|
"""
|
|
return self._current_step
|