docs: fix model name in ollama example

[imfing]

This PR essentially reverts #279

When I followed

pydantic-ai/docs/api/models/ollama.md

Line 29 in af79e3d

agent = Agent('llama3.2', result_type=CityLocation)

I got pydantic_ai.exceptions.UserError: Unknown model: llama3.2

Traceback (most recent call last):
  File "/dev/github.com/pydantic/pydantic-ai/test.py", line 11, in <module>
    agent = Agent('llama3.2', result_type=CityLocation)
            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/dev/github.com/pydantic/pydantic-ai/.venv/lib/python3.12/site-packages/pydantic_ai/agent.py", line 160, in __init__
    self.model = models.infer_model(model)
                 ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/dev/github.com/pydantic/pydantic-ai/.venv/lib/python3.12/site-packages/pydantic_ai/models/__init__.py", line 302, in infer_model
    raise UserError(f'Unknown model: {model}')
pydantic_ai.exceptions.UserError: Unknown model: llama3.2

turns out the Agent class initialization checks the prefix:

pydantic-ai/pydantic_ai_slim/pydantic_ai/models/__init__.py

Line 66 in af79e3d

'ollama:llama3.2',

pydantic-ai/pydantic_ai_slim/pydantic_ai/models/__init__.py

Line 293 in af79e3d

elif model.startswith('ollama:'):

[samuelcolvin]

thanks, sorry for the mistake.

[asmith26]

Apologies for the mistake/inconvenience, I've just realised this myself - I was exploring using Agent by itself vs OllamaModel with Agent and must have gotten in a bit of a muddle:

from pydantic_ai import Agent

# prefix needed here (parsed from https://github.com/pydantic/pydantic-ai/blob/339f89c76837c2264322e0e6e570c957516a6f79/pydantic_ai_slim/pydantic_ai/models/__init__.py#L66)
agent = Agent('ollama:llama3.2', result_type=CityLocation)  

vs

from pydantic_ai.models.ollama import OllamaModel

# no prefix needed here (parsed from: https://github.com/pydantic/pydantic-ai/blob/339f89c76837c2264322e0e6e570c957516a6f79/pydantic_ai_slim/pydantic_ai/models/ollama.py#L32)
ollama_model = OllamaModel(model_name='llama3.2')

from pydantic_ai import Agent
agent = Agent(ollama_model, result_type=CityLocation)

Apologies again for the mistake/inconvenience, and many thanks for the fix/amazing lib!

[srishrachamalla7]

I was having some issues with:

agent = Agent('ollama:llama3.2:1b', result_type=CityLocation)

But after seeing this, I tried:

ollama_model = OllamaModel(model_name='llama3.2:1b')

which is now working fine.

Thanks so much for your help and the amazing library!

Here's my updated code:

from pydantic import BaseModel
from pydantic_ai.models.ollama import OllamaModel

class CityLocation(BaseModel):
    city: str
    country: str

# agent = Agent('ollama:llama3.2 ', result_type=CityLocation)  # Original line
ollama_model = OllamaModel(model_name='llama3.2:1b')  # Updated line

from pydantic_ai import Agent
agent = Agent(ollama_model, result_type=CityLocation)

result = agent.run_sync('Where were the olympics held in 2012?')
print(result.data)
#> city='London' country='United Kingdom'
print(result.usage())
"""
Usage(requests=1, request_tokens=57, response_tokens=8, total_tokens=65, details=None)
"""

Thanks again!