Rodrigo Rodriguez (Pragmatismo)
5a43dc81c7
Replace Azure Functions architecture with a modern FastAPI-based REST API providing image, video, speech, and vision capabilities for General Bots. Key changes: - Add FastAPI app with versioned API endpoints and OpenAPI docs - Implement services for Stable Diffusion, Zeroscope, TTS/Whisper, BLIP2 - Add pydantic schemas for request/response validation - Configure structured logging with structlog - Support lazy model loading and GPU acceleration - Update dependencies from Azure/TensorFlow stack to PyTorch/diffusers
38 lines
789 B
Text
38 lines
789 B
Text
# Server Configuration
|
|
ENV=development
|
|
HOST=0.0.0.0
|
|
PORT=8085
|
|
LOG_LEVEL=INFO
|
|
|
|
# Security - IMPORTANT: Change this in production!
|
|
API_KEY=change-me-in-production
|
|
|
|
# Model Paths
|
|
# These can be local paths or model identifiers for HuggingFace Hub
|
|
IMAGE_MODEL_PATH=./models/stable-diffusion-v1-5
|
|
VIDEO_MODEL_PATH=./models/zeroscope-v2
|
|
SPEECH_MODEL_PATH=./models/tts
|
|
VISION_MODEL_PATH=./models/blip2
|
|
WHISPER_MODEL_PATH=./models/whisper
|
|
|
|
# Device Configuration
|
|
# Options: cuda, cpu, mps (for Apple Silicon)
|
|
DEVICE=cuda
|
|
|
|
# Image Generation Defaults
|
|
IMAGE_STEPS=4
|
|
IMAGE_WIDTH=512
|
|
IMAGE_HEIGHT=512
|
|
IMAGE_GPU_LAYERS=20
|
|
IMAGE_BATCH_SIZE=1
|
|
|
|
# Video Generation Defaults
|
|
VIDEO_FRAMES=24
|
|
VIDEO_FPS=8
|
|
VIDEO_WIDTH=320
|
|
VIDEO_HEIGHT=576
|
|
VIDEO_GPU_LAYERS=15
|
|
VIDEO_BATCH_SIZE=1
|
|
|
|
# Storage
|
|
OUTPUT_DIR=./outputs
|