Use local enterprise LLMs

LLMs
1

Does CrewAI allow using local enterprise LLMs (e.g Azure Open AI or GGUF files downloaded locally ) ?

2

Hi and welcome to the community

This might help your search

CrewAI supports OLLAMA some examples here LLMs - CrewAI

llm = LLM(
    model="ollama/llama3:70b",
    base_url="http://localhost:11434"
)
3

Thanks for the response. Where or which files does this need to configured ?

4

you can set either in .env as example like

MODEL=ollama/llama3.1
API_BASE=http://localhost:11434

or by import in crew.py as below

from crewai import LLM
llm = LLM(
    model="ollama/llama3.1",
    base_url="http://localhost:11434"
)

you can also pass the llm within the class as

self.llm = LLM(
    model="ollama/llama3.1",
    base_url="http://localhost:11434"
)

if you chose the second method don’t forget to specify the llm either in agent or task like

    def researcher(self) -> Agent:
        return Agent(
            config=self.agents_config['researcher'],
            llm=self.llm #llm <<--------------
            verbose=True
        )
5

Please refer the litellm docs for this. vllm and other providers are supported there.