feat: add retry mechanism for ModelBehaviorError

[Oxygen56]

Summary

Adds automatic retry for ModelBehaviorError with configurable max_model_retries on RunConfig.

When a model produces a malformed response (invalid tool call JSON, nonexistent tool, etc.), the SDK raises ModelBehaviorError. Previously there was no built-in retry — the agent just failed. This PR adds an automatic retry mechanism that feeds the error message back to the model so it can self-correct.

Changes

• RunConfig.max_model_retries: int = 0 — new parameter controlling how many automatic retries are attempted when ModelBehaviorError occurs (defaults to 0, preserving existing behavior)
• run_single_turn (non-streaming) — wraps model call + response processing in a retry loop; on error, appends a synthetic user message with the error details and retries the model call
• run_single_turn_streamed (streaming) — same retry logic for the streaming path; events from failed attempts are emitted (consistent with existing network-level retry behavior)

How it works

1. The model call completes successfully but produces a malformed response
2. get_single_step_result_from_response raises ModelBehaviorError
3. If retries remain, the error message is appended as a synthetic user message: "Your previous response was invalid: <error message>"
4. The model is called again with the error feedback
5. This repeats until success or retries are exhausted

Usage

from agents import Agent, Runner, RunConfig

config = RunConfig(max_model_retries=3)
result = await Runner.run(agent, "...", run_config=config)

Closes #325

🤖 Generated with Claude Code

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 62fe089a74

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[chatgpt-codex-connector]

Buffer streamed events before retrying malformed turns

When max_model_retries is enabled for streamed runs, a malformed completed response (for example, a nonexistent tool) is only detected later in get_single_step_result_from_response, but this line has already delivered all raw events from the failed attempt to the caller. The retry then emits a second attempt into the same stream, so stream consumers can observe invalid response.output_item.done/tool_called events that are not part of the final accepted turn and may act on a tool call the SDK is about to discard. Consider buffering per-attempt events until the response is accepted, or disabling this retry path for streaming after any events are emitted.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Avoid retrying after tool side effects

This catches ModelBehaviorError around the entire step resolution, not just response parsing. When the error is raised from tool execution (for example, one malformed shell/apply_patch call while another function/computer/custom tool in the same turn has already run, since _execute_tool_plan runs tool groups concurrently), the retry asks the model again and can execute those side-effecting tools a second time. Restrict the retry to validation before any tool side effects, or only retry when no tool execution has started.

Useful? React with 👍 / 👎.