Audio transcription response is typed too generically

[saxon134]

AudioResource.transcription() returns APIResponse for /v1/audio/transcriptions, but the mocked API response contains a top-level text field and the repository's own test accesses .text. Because APIResponse does not declare that attribute, python -m mypy src tests fails with: tests/test_client.py:165: error: "APIResponse" has no attribute "text". Add a dedicated transcription response model with text or change the typed surface so this endpoint exposes the documented field.

File: src/globalrouter/_resources.py

Line: 536

Severity: medium

Summary: generic response type for audio transcription

[saxon134]

Coding Loop Plan

Refined Issue

Audio transcription methods currently return the generic APIResponse type even though transcription responses expose a concrete text field. Add a dedicated typed response model and use it for sync and async transcription methods.

Implementation Plan

Implementation Plan

Keep the change scoped to the audio transcription response surface. Add a dedicated model for transcription results, update only the transcription methods to return that model, export the new public type, and add a regression test that verifies both runtime parsing and the method annotations. Do not change request payload behavior or unrelated audio endpoints.

Development Tasks

1. Add a dedicated audio transcription response type
   Create an AudioTranscription model in src/globalrouter/_models.py with a required text: str field, relying on the existing GlobalRouterModel extra="allow" behavior so additional provider fields continue to round-trip. Import and export this model from src/globalrouter/init.py so users can reference the public type.
   Files: src/globalrouter/_models.py, src/globalrouter/init.py
   Tests: python -m mypy src tests

2. Use the transcription type in audio resource methods
   Update AudioResource.transcription and AudioResource.transcription_async in src/globalrouter/_resources.py to return AudioTranscription instead of APIResponse, and pass AudioTranscription into request_model/request_model_async for /v1/audio/transcriptions. Leave AudioResource.speech unchanged because this issue is only about transcription.
   Files: src/globalrouter/_resources.py
   Tests: python -m mypy src tests, python -m pytest tests/test_client.py::test_native_surface_and_sse_streaming

3. Add regression coverage for transcription typing
   Extend tests/test_client.py to import the new AudioTranscription type and verify client.audio.transcription(...) returns an AudioTranscription with text == "hello". Add an annotation regression check, such as typing.get_type_hints(AudioResource.transcription)["return"] and typing.get_type_hints(AudioResource.transcription_async)["return"], to ensure both methods are no longer typed as APIResponse.
   Files: tests/test_client.py
   Tests: python -m pytest tests/test_client.py::test_native_surface_and_sse_streaming, python -m mypy src tests