Add output_json_schema method to Agent class

[g-eoj]

This PR adds a method to get the json schema for agent output.

It aims to support all output types, including special ones:

• text only
• BinaryImage
• DeferredToolRequests

Simple example:

agent = Agent('test')
agent.output_json_schema()

{'type': 'string'}

Override agent output types:

agent = Agent('test')
agent.output_json_schema(output_type=bool)

{'properties': {'response': {'type': 'boolean'}},
 'required': ['response'],
 'type': 'object',
 'title': 'bool'}

Discriminated JSON whenever there is more than one output type:

agent = Agent('test', output_type=[str,bool])
agent.output_json_schema()

{'type': 'object',
 'properties': {'result': {'anyOf': [{'type': 'object',
     'properties': {'kind': {'type': 'string', 'const': 'final_result'},
      'data': {'properties': {'response': {'type': 'boolean'}},
       'required': ['response'],
       'type': 'object'}},
     'required': ['kind', 'data'],
     'additionalProperties': False,
     'title': 'final_result'},
    {'type': 'object',
     'properties': {'kind': {'type': 'string', 'const': 'str'},
      'data': {'type': 'string'}},
     'required': ['kind', 'data'],
     'additionalProperties': False}]}},
 'required': ['result'],
 'additionalProperties': False}

Closes #3225

[g-eoj]

@DouweM is there a source of truth I can reference for what the schema output should be for tests? I think I will get stuck figuring this out so any guidance will be appreciated.

Edit: I added snapshots of expected output to the tests. Please review them carefully as I'm guessing what they are supposed to be.

[g-eoj]

@DouweM thanks for the reviews so far.

Fixed: The PR has an issue where it always adds str to the output schema if the AutoOutputSchema is selected. However I think this is due to a bug not caused by the PR. I think it is because

pydantic-ai/pydantic_ai_slim/pydantic_ai/_output.py

Line 398 in bcd3d83

text_processor=processor,

will always set allows_text to True. This doesn't seem right but I'll wait for confirmation.

[g-eoj]

I don't know how to properly fix the coverage issue 😬

[DouweM]

@g-eoj Let me know when this is ready for review again!

[g-eoj]

@DouweM it is ready! Apologies.

[DouweM]

FYI The coverage CI step complains about this line and the next one being not being covered by tests

[g-eoj]

I thought my tests triggered the loop, in fact I purposely added one with non-unique keys. However, the offending code is gone now.

[g-eoj]

@DouweM how should duplicate tool output types be handled? For example:

output_type=[ToolOutput(bool, name='one'), ToolOutput(bool, name='two'), ToolOutput(bool, name='three')]

Can it be:

{'anyOf': [{'type': 'boolean'}, {'type': 'boolean'}, {'type': 'boolean'}]}  

Or should duplicates be removed?

[g-eoj]

@DouweM ready for review. The test failure doesn't appear to be from this PR. I have one question about whether we should remove duplicate output types: #3454 (comment)

[DouweM]

@g-eoj Yeah that's fine to be deduplicated. The code using agent.run(...).output wouldn't be able to tell the different bools apart either (except by inspecting the message history for tool call parts).

I'll do another review later or early next week, I have some some more PRs to get through :)