# For DALL-E and OpenAI TTS
export OPENAI_API_KEY="sk-..."

# For Fal.ai (Flux images, video, Whisper)
export FAL_KEY="..."

from agentfield import Agent, AIConfig

app = Agent(
    node_id="media-agent",
    ai_config=AIConfig(
        fal_api_key="...",  # Optional - falls back to FAL_KEY env var
        video_model="fal-ai/minimax-video/image-to-video"  # Default video model
    )
)

# Fal.ai - Flux (fast, high quality)
result = await app.ai_with_vision(
    "A cyberpunk city at night",
    model="fal-ai/flux/schnell"  # Fast
)
result.images[0].save("city.png")

# DALL-E 3 (via LiteLLM)
result = await app.ai_with_vision(
    "A serene mountain landscape",
    model="dall-e-3",
    size="1792x1024",
    quality="hd"
)

# OpenRouter
result = await app.ai_with_vision(
    "Abstract art",
    model="openrouter/google/gemini-2.5-flash-image-preview"
)

# OpenAI TTS
result = await app.ai_with_audio(
    "Hello, welcome to the presentation.",
    voice="nova",  # alloy, echo, fable, onyx, nova, shimmer
    model="tts-1-hd"
)
result.audio.save("greeting.mp3")
result.audio.play()  # Requires pygame

# Image-to-video (default model)
result = await app.ai_generate_video(
    "Camera slowly zooms in on the landscape",
    image_url="https://example.com/image.jpg"
)
result.files[0].save("video.mp4")

# Text-to-video
result = await app.ai_generate_video(
    "A cat playing with yarn",
    model="fal-ai/kling-video/v1/standard"
)

# Basic transcription
result = await app.ai_transcribe_audio(
    "https://example.com/recording.mp3"
)
print(result.text)

# Faster transcription with language hint
result = await app.ai_transcribe_audio(
    "https://example.com/spanish.mp3",
    model="fal-ai/wizper",  # 2x faster than whisper
    language="es"
)