Ignore leading whitespace when streaming text, fixing run_stream + Ollama + Qwen3

[DouweM]

Ollama + Qwen3 will emit <think>\n</think>\n\n ahead of tool calls, which we don't want to end up treating as a final result when using run_stream (as that would stop the run before even handling the tool calls).

(See example code below for context on this output)

Before, the TextPart(content='\n\n') is treated as the final result:

PartStartEvent(index=0, part=ThinkingPart(content='')),
PartDeltaEvent(index=0, delta=ThinkingPartDelta(content_delta='\n')),
PartStartEvent(index=1, part=TextPart(content='\n\n')),
FinalResultEvent(tool_name=None, tool_call_id=None),
PartStartEvent(index=2, part=ToolCallPart(tool_name='roll_dice', args='{}', tool_call_id='call_q2vb1z22')),    
...

After, only the actual text after the tool call is treated as the final result: (If it was a final_result call, that call would be treated as the final result)

PartStartEvent(index=0, part=ThinkingPart(content=''))
PartDeltaEvent(index=0, delta=ThinkingPartDelta(content_delta='\n'))
PartStartEvent(index=1, part=ToolCallPart(tool_name='roll_dice', args='{}', tool_call_id='call_frxbo7tq'))
...
PartStartEvent(index=1, part=TextPart(content='You'))
FinalResultEvent(tool_name=None, tool_call_id=None)
PartDeltaEvent(index=1, delta=TextPartDelta(content_delta=' rolled'))
PartDeltaEvent(index=1, delta=TextPartDelta(content_delta=' a'))
PartDeltaEvent(index=1, delta=TextPartDelta(content_delta=' '))
PartDeltaEvent(index=1, delta=TextPartDelta(content_delta='5'))
PartDeltaEvent(index=1, delta=TextPartDelta(content_delta='.'))

Code

import asyncio
import random

from pydantic_ai import Agent
from pydantic_ai.models.openai import OpenAIModel
from pydantic_ai.providers.openai import OpenAIProvider


def roll_dice() -> str:
    """Roll a six-sided die and return the result."""
    return str(random.randint(1, 6))


# --- Agent and Model Configuration ---


ollama_model = OpenAIModel(
    model_name='qwen3:1.7b',
    provider=OpenAIProvider(base_url='http://localhost:11434/v1'),
    settings={'temperature': 0.2},
)

default_system_prompt = """
You are a helpful assistant.
Do not ask any follow-up questions.
"""

model_settings = {
    'temperature': 0.2,
}

tools = [roll_dice]

agent = Agent(model=ollama_model, tools=tools, system_prompt=default_system_prompt, output_type=str)


async def main():
    prompt = 'roll the dice'

    async with agent.iter(prompt) as agent_run:
        async for node in agent_run:
            if Agent.is_model_request_node(node) or Agent.is_call_tools_node(node):
                async with node.stream(agent_run.ctx) as request_stream:
                    async for event in request_stream:
                        print(event, flush=True)


# ---

if __name__ == '__main__':
    asyncio.run(main())

[github-actions]

Docs Preview

commit:	d145396
Preview URL:	https://91e4334f-pydantic-ai-previews.pydantic.workers.dev