Integration with LLM Catalog

[lbedner]

Scope

Query existing LLM catalog for audio-capable models.

Tasks

1. Add catalog query for audio models

def get_audio_models(session: Session, direction: str = "output") -> list[LargeLanguageModel]:
    """Get models that support audio modality."""
    return session.exec(
        select(LargeLanguageModel)
        .join(LLMModality)
        .where(LLMModality.modality == Modality.AUDIO)
        .where(LLMModality.direction == Direction(direction))
    ).all()

2. Display audio-capable models in Voice Settings Tab

• Show which models support real-time audio input/output
• Examples: GPT-4o-audio-preview, gpt-4o-realtime-preview

3. Link audio models to voice services

• Allow selection of audio-capable LLMs for voice chat
• Show capabilities (realtime, streaming, etc.)

Files

• app/services/ai/llm_catalog_context.py (add audio query methods)
• app/services/ai/etl/llm_sync_service.py (verify audio modality sync)

Dependencies

• Requires Voice Catalog Infrastructure #480 (Voice Catalog Infrastructure)
• Integrates with Voice Settings Tab Frontend #482 (Voice Settings Tab Frontend)

Verification

# Sync audio models from catalog
my-app ai llm-sync --mode=all

# Query should return audio-capable models
# Check database for models with AUDIO modality