Agent llama tools wont work properly #2973
Description
Initial Checks
- I confirm that I'm using the latest version of Pydantic AI
- I confirm that I searched for my issue in https://github.com/pydantic/pydantic-ai/issues before opening this issue
Description
Im trying to make local llama3.1:8b model use weather tool to suggest clothes to wear.
In response Im getting something like:
I can only respond with formatted JSON as per AI model restrictions.
Or:
I'll respond with a JSON object for a function call that best answers your prompt
(though initial params wont change)
And in (very) rare cases something relevant like waterproof shoes
My setup:
Windows 10, nvidia 4070, virtual ubuntu 24.04 (WSL v2)
Installed llama LLM with $ ollama pull llama3.1:8b
And commands like ollama run llama3.1:8b works fine. But not the tool, let me show you example:
Example Code
import asyncio
from dataclasses import dataclass
from typing import Any
from pydantic import BaseModel, Field
from pydantic_ai import Agent, RunContext, NativeOutput
from pydantic_ai.models.openai import OpenAIChatModel
from pydantic_ai.providers.ollama import OllamaProvider
class AgentOutput(BaseModel):
response_text: str = Field(description="Message to the user")
# Initialize the Ollama provider
ollama_provider = OllamaProvider(
base_url="http://localhost:11434/v1",
)
model = OpenAIChatModel(
model_name="llama3.1:8b",
provider=ollama_provider,
)
main_agent = Agent(
model=model,
output_type=AgentOutput,
system_prompt=(
"You are a smart home assistant. ",
"Use the available tools when needed. ",
"When asked about weather, always use the 'get_location_weather' tool."
),
retries=5,
)
@main_agent.tool_plain
async def get_location_weather(
location: str = Field(description="The location to get weather for"),
# *args
) -> dict[str, Any]:
"""
Get the current weather for a specific location.
Args:
location: Location to check weather
"""
print(f"[DEBUG] Tool used.")
print(f"[DEBUG] Location type: '{type(location)}'")
print(f"[DEBUG] Location val: '{location}'")
# In a real implementation, you would call a weather API here
return {location: f"The weather in {location} is currently rainy with a temperature of 15°C."}
async def main():
result = await main_agent.run(
"Check the weather in Paris and suggest what am I need to wear?",
)
print(result.output)
if __name__ == "__main__":
asyncio.run(main())Python, Pydantic AI & LLM client version
python==3.12.3
pydantic==2.11.9
pydantic-ai==1.0.9
pydantic-ai-slim==1.0.9
pydantic-evals==1.0.9
pydantic-graph==1.0.9
pydantic-settings==2.10.1
pydantic_core==2.33.2