Adding support for pronunciation_dict_id

[bpanahij]

Summary

Adds support for the pronunciation_dict_id parameter to TTS generation requests, enabling users to apply custom pronunciation dictionaries to their text-to-speech generations.

Changes

• Added pronunciation_dict_id parameter to GenerationRequestParams and GenerationRequest types
  • src/cartesia/tts/requests/generation_request.py:72-76
  • src/cartesia/tts/types/generation_request.py:77-79
• Added test coverage for both SSE and WebSocket interfaces
  • tests/custom/test_client.py:434-455 - SSE test
  • tests/custom/test_client.py:610-641 - WebSocket async test

Implementation Details

The pronunciation_dict_id parameter is:

• Optional (typing_extensions.NotRequired in TypedDict, typing.Optional[str] in Pydantic model)
• Available in both SSE and WebSocket generation methods
• Applied on a per-generation basis, allowing different pronunciation dictionaries for different requests

Test Plan

• ✅ Added test_sse_pronunciation_dict() to verify SSE endpoint accepts the parameter
• ✅ Added test_ws_pronunciation_dict() to verify WebSocket endpoint accepts the parameter
• ✅ Both tests validate audio generation works correctly with the parameter

---

This enables users of the official Cartesia Python client to leverage custom pronunciation dictionaries when generating speech.