Reset Responses stub output for each new request

[sylvesterkaczmarek]

Summary

Make the Responses API stub backend restart its deterministic fake-token sequence for every new API request.

The stub accepts the standard new_request flag but currently ignores it, while its token position lives in the module-global token_queue. A second request therefore starts wherever the previous request stopped rather than from the beginning of the stub response.

Fixes #281.

Fix

Reset token_queue to fake_tokens.copy() when new_request=True before returning the first token.

Regression coverage

Adds a deterministic test that:

• consumes the first two fake tokens;
• starts a new request;
• verifies the next token is again fake_tokens[0].

The fake response content, automatic wraparound behavior, setup interface, and simulated latency are otherwise unchanged.

[chatgpt-codex-connector]

Note

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.