Anthropic tool response nested content blocks are stringified (no image response possible?)

[indigoviolet]

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

Anthropic supports image responses https://docs.anthropic.com/en/docs/build-with-claude/tool-use/overview#handling-tool-use-and-tool-result-content-blocks

But we're currently forcing the response to be a str (

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

Line 295 in 91265fb

content=request_part.model_response_str(),

), which means IIUC there's no way to return an image block?

Example Code

Python, Pydantic AI & LLM client version

3.12
0.0.46
Anthropic

[Wh1isper]

I found this too. I think we might be able to use BinaryContent: #1339

[TheFirstMe]

I am having the same issue when using Claude from the Bedrock model.

[TheFirstMe]

I found this too. I think we might be able to use BinaryContent: #1339

#1517 adds support for multi-modal response from tools.

[Wh1isper]

#1517 adds support for multi-modal response from tools.

Seems better, thank you!