☕️ CoffiFilter is a tool to help manage your LLM agents' outbound tool usage.
- A tool to help manage your LLM agents' outbound tool usage.
- Track and manage the tools that your agents are using.
- Switch tools on and off for your agents without having to redeploy them.
- Connect to your own redis server.
- Drop in decorator for langchain functions.
- Python 3.6 or later
- Redis server
You can install CoffiFilter via pip from PyPI:
$ pip install coffifilterCoffiFilter can wrap your Langchain tools to help manage their usage.
import coffifilter
from langchain_community.tools import YouTubeSearchTool
coffifilter.init(
redis_host="your-redishost.redis-cloud.com",
redis_port=11552,
redis_db=0,
redis_password="your-redispassword",
)
youtube_tool = coffifilter.wrap_langchain_tool(YouTubeSearchTool())
...
tools = [..., youtube_tool]
# Create an agent executor by passing in the agent and tools
agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True)
agent_executor.invoke({"input": "can you find videos on langchain"})CoffiFilter can also be used as a decorator for your langchain functions.
import coffifilter
coffifilter.init(
redis_host="your-redishost.redis-cloud.com",
redis_port=11552,
redis_db=0,
redis_password="your-redispassword",
)
@tool
@coffifilter.coffi_filter("summarize_tool")
def summarize_tool(url: str, callbacks: Callbacks = None):
"""Summarize a website."""
text = requests.get(url).text
summary_chain = (
ChatPromptTemplate.from_template(
"Summarize the following text:\n<TEXT {uid}>\n" "{text}" "\n</TEXT {uid}>"
).partial(uid=lambda: uuid.uuid4())
| ChatOpenAI(model="gpt-4o")
| StrOutputParser()
).with_config(run_name="Summarize Text")
return summary_chain.invoke(
{"text": text},
{"callbacks": callbacks},
)The current design is to check the redis server for the tool status before executing the function. It literally checks if the tool is on or off by checking the value of the key in the redis server. If the key is not found, it will default to off. If the key is found, it will check if the value is "true" or "false".
The decorator will raise a ValueError if the tool is off.
Eg. If the tool is off, the following error will be raised:
ValueError(f"Tool '{filter_string}' is not enabled")- Better documentation.
- Better langchain integration.
- Better error handling.
- Local first approach; avoiding redis server if not needed.
- User tracking and IFTTT tool usage.
Contributions are very welcome. To learn more, see the Contributor Guide.
Distributed under the terms of the GPL 3.0 license, CoffiFilter is free and open source software.
If you encounter any problems, please file an issue along with a detailed description.
This project was generated from @cjolowicz's Hypermodern Python Cookiecutter template.