v0.5.0

Breaking

• Replace Grok Imagine image-to-video source_image_urls with scalar source_image_url.
  Migration: Pass --source-image-url URL or provide source_image_url in JSON input.

Added

• Add the synchronous runapi midjourney shorten-prompt command with generated input validation and help.
• Add synchronous text-to-speech commands for OpenAI TTS and Fish Audio.
• Expose model and text input help from the generated contract.
• Add Gemini Omni Flash Preview model selection and model-specific text-to-video help.
• Add the gemini-tts text-to-speech command with generated input help and validation.
• Add Seedream 5 Pro text-to-image and edit-image models to CLI help and validation.
• Expose advanced stem separation mode and supported stem names in CLI help and validation.
• Add the Producer text-to-music command with generated contract help and typed SDK wiring.

Changed

• Expose Seedream 5-Lite output_format values and default in CLI help and validation.
• Include minimum and maximum item counts for constrained array fields in generated CLI metadata and help.

Fixed

• Preserve API-provided error codes in JSON output and omit the code when the response does not provide one.
• Classify command exit and retry behavior by HTTP status or SDK error type instead of generating substitute codes.