How to Set Custom LLM Parameters (like `temperature`, `seed`) in YAML Config for Ollama

General
1

Hi CrewAI team,

I’m configuring agents using the agents.yaml file. Here’s a sample agent using Ollama:

user_journey_agent:
  role: >
    User Journey Designer
  goal: >
    An AI agent that will design learning journeys and goals for the user based on specific topic inputs.
  backstory: >
    This agent anticipates the user's educational goals, proficiency needs, and preferred learning style to craft a structured learning journey.
  llm: ollama/llama3.1

In Python, I usually initialise the LLM like this:

LLM(model="ollama/llama3.1", base_url="http://localhost:11434", temperature=0.01, seed=42)

How can I pass parameters like seed, temperature, when using the YAML config approach?
Is there an extended syntax or a separate YAML section where these should be specified?
By default, I can see CrewAI uses environment variables such as:

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

However, adding something like SEED=42 to .env doesn’t seem to affect the output or behaviour.
Appreciate your guidance

2

The issue has been resolved by passing llm in crew.py as

from crewai import Agent, Crew, Process, Task, LLM
from crewai.project import CrewBase, agent, crew, task

llm = LLM(
    model="ollama/llama3.1",
    base_url="http://localhost:11434",
    temperature=0.01,
    seed=42
)
@CrewBase
class LearningAssistant():
    """LearningAssistant crew"""

    agents_config = 'config/agents.yaml'
    tasks_config = 'config/tasks.yaml'

    @agent
    def user_journey_agent(self) -> Agent:
        return Agent(
            config=self.agents_config['user_journey_agent'],
            llm = llm,
            verbose=True
        )

I have ensured to hash everything from the.env and retried with this method so that we can ensure the response is the same even after repeated attempts