mirror of
https://github.com/Mintplex-Labs/langchain-python.git
synced 2026-07-22 09:15:31 -04:00
Jason Liu
54f9e4287f
# Problem
I noticed that in order to change the prefix of the prompt in the
`zero-shot-react-description` agent
we had to dig around to subset strings deep into the agent's attributes.
It requires the user to inspect a long chain of attributes and classes.
`initialize_agent -> AgentExecutor -> Agent -> LLMChain -> Prompt from
Agent.create_prompt`
``` python
agent = initialize_agent(
tools=tools,
llm=fake_llm,
agent="zero-shot-react-description"
)
prompt_str = agent.agent.llm_chain.prompt.template
new_prompt_str = change_prefix(prompt_str)
agent.agent.llm_chain.prompt.template = new_prompt_str
```
# Implemented Solution
`initialize_agent` accepts `**kwargs` but passes it to `AgentExecutor`
but not `ZeroShotAgent`, by simply giving the kwargs to the agent class
methods we can support changing the prefix and suffix for one agent
while allowing future agents to take advantage of `initialize_agent`.
```
agent = initialize_agent(
tools=tools,
llm=fake_llm,
agent="zero-shot-react-description",
agent_kwargs={"prefix": prefix, "suffix": suffix}
)
```
To be fair, this was before finding docs around custom agents here:
https://langchain.readthedocs.io/en/latest/modules/agents/examples/custom_agent.html?highlight=custom%20#custom-llmchain
but i find that my use case just needed to change the prefix a little.
# Changes
* Pass kwargs to Agent class method
* Added a test to check suffix and prefix
---------
Co-authored-by: Jason Liu <jason@jxnl.coA>
73 lines
2.5 KiB
Python
73 lines
2.5 KiB
Python
"""Load agent."""
|
|
from typing import Any, List, Optional
|
|
|
|
from langchain.agents.agent import AgentExecutor
|
|
from langchain.agents.loading import AGENT_TO_CLASS, load_agent
|
|
from langchain.agents.tools import Tool
|
|
from langchain.callbacks.base import BaseCallbackManager
|
|
from langchain.llms.base import BaseLLM
|
|
|
|
|
|
def initialize_agent(
|
|
tools: List[Tool],
|
|
llm: BaseLLM,
|
|
agent: Optional[str] = None,
|
|
callback_manager: Optional[BaseCallbackManager] = None,
|
|
agent_path: Optional[str] = None,
|
|
agent_kwargs: Optional[dict] = None,
|
|
**kwargs: Any,
|
|
) -> AgentExecutor:
|
|
"""Load agent given tools and LLM.
|
|
|
|
Args:
|
|
tools: List of tools this agent has access to.
|
|
llm: Language model to use as the agent.
|
|
agent: The agent to use. Valid options are:
|
|
`zero-shot-react-description`
|
|
`react-docstore`
|
|
`self-ask-with-search`
|
|
`conversational-react-description`
|
|
If None and agent_path is also None, will default to
|
|
`zero-shot-react-description`.
|
|
callback_manager: CallbackManager to use. Global callback manager is used if
|
|
not provided. Defaults to None.
|
|
agent_path: Path to serialized agent to use.
|
|
**kwargs: Additional key word arguments to pass to the agent.
|
|
|
|
Returns:
|
|
An agent.
|
|
"""
|
|
if agent is None and agent_path is None:
|
|
agent = "zero-shot-react-description"
|
|
if agent is not None and agent_path is not None:
|
|
raise ValueError(
|
|
"Both `agent` and `agent_path` are specified, "
|
|
"but at most only one should be."
|
|
)
|
|
if agent is not None:
|
|
if agent not in AGENT_TO_CLASS:
|
|
raise ValueError(
|
|
f"Got unknown agent type: {agent}. "
|
|
f"Valid types are: {AGENT_TO_CLASS.keys()}."
|
|
)
|
|
agent_cls = AGENT_TO_CLASS[agent]
|
|
agent_kwargs = agent_kwargs or {}
|
|
agent_obj = agent_cls.from_llm_and_tools(
|
|
llm, tools, callback_manager=callback_manager, **agent_kwargs
|
|
)
|
|
elif agent_path is not None:
|
|
agent_obj = load_agent(
|
|
agent_path, llm=llm, tools=tools, callback_manager=callback_manager
|
|
)
|
|
else:
|
|
raise ValueError(
|
|
"Somehow both `agent` and `agent_path` are None, "
|
|
"this should never happen."
|
|
)
|
|
return AgentExecutor.from_agent_and_tools(
|
|
agent=agent_obj,
|
|
tools=tools,
|
|
callback_manager=callback_manager,
|
|
**kwargs,
|
|
)
|