runcontext has no token count in history processor

[jcaplan]

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

trying to generate a summary based on token count instead of number of messages. I can see the token count for each message but the ctx.usage.total_tokens is always 0.

> history.py(57)summary_processor()
(Pdb) p ctx.usage.total_tokens
0
(Pdb) p messages[1].usage
RequestUsage(input_tokens=24, output_tokens=11, details={'accepted_prediction_tokens': 0, 'audio_tokens': 0, 'reasoning_tokens': 0, 'rejected_prediction_tokens': 0})

Example Code

Python, Pydantic AI & LLM client version

pydantic-ai==1.0.10

[jcaplan]

maybe as intended but I think what's happening here is that ctx.usage only reflects the new messages on current round and does not reflect the tokens in the message history. Since it's all getting sent to the LLM, I'd expect to see the global cost tally reflect the message history as well.

[DouweM]

@jcaplan This is expected -- the RunUsage for a given run only contains usage of that run. If you want to include the usage of the history as well, you'll need to store the original run's result.usage() and pass it into the new run alongside message_history: agent.run(..., message_history=result.all_messages(), usage=result.usage()`.