Webscout-TTI / app.py
Abhaykoul's picture
Update app.py
6f99d02 verified
raw
history blame
27.9 kB
import os
import io
import uuid
import base64
from typing import Dict, List, Optional, Any, Union
from pathlib import Path
import aiohttp
from fastapi import FastAPI, HTTPException, BackgroundTasks, Depends, Header, Request
from fastapi.responses import StreamingResponse, JSONResponse
from fastapi.middleware.cors import CORSMiddleware
from pydantic import BaseModel, Field
import asyncio
import uvicorn
from datetime import datetime
import time
# Import all TTI providers
from webscout.Provider.TTI import (
# Import all image providers
BlackboxAIImager, AsyncBlackboxAIImager,
DeepInfraImager, AsyncDeepInfraImager,
AiForceimager, AsyncAiForceimager,
NexraImager, AsyncNexraImager,
FreeAIImager, AsyncFreeAIImager,
NinjaImager, AsyncNinjaImager,
TalkaiImager, AsyncTalkaiImager,
PiclumenImager, AsyncPiclumenImager,
ArtbitImager, AsyncArtbitImager,
HFimager, AsyncHFimager,
)
try:
from webscout.Provider.TTI import AIArtaImager, AsyncAIArtaImager
AIARTA_AVAILABLE = True
except ImportError:
AIARTA_AVAILABLE = False
# Create FastAPI instance
app = FastAPI(
title="WebScout TTI API Server",
description="API server for Text-to-Image generation using various providers with OpenAI-compatible interface",
version="1.0.0",
)
# Add CORS middleware to allow cross-origin requests
app.add_middleware(
CORSMiddleware,
allow_origins=["*"],
allow_credentials=True,
allow_methods=["*"],
allow_headers=["*"],
)
# Storage for generated images (in-memory for demo purposes)
# In a production environment, you might want to store these in a database or a file system
IMAGE_STORAGE = {}
# Simple API key verification (demo purposes only)
# In production, you'd want a more secure authentication system
API_KEYS = {"sk-demo-key": "demo"}
# Provider mapping
PROVIDER_MAP = {
"blackbox": {
"class": AsyncBlackboxAIImager,
"description": "High-performance image generation with advanced retry mechanisms"
},
"deepinfra": {
"class": AsyncDeepInfraImager,
"description": "Powerful image generation using FLUX-1-schnell and other models"
},
"aiforce": {
"class": AsyncAiForceimager,
"description": "Advanced AI image generation with 12 specialized models"
},
"nexra": {
"class": AsyncNexraImager,
"description": "Next-gen image creation with 19+ models"
},
"freeai": {
"class": AsyncFreeAIImager,
"description": "Premium image generation with DALL-E 3 and Flux series models"
},
"ninja": {
"class": AsyncNinjaImager,
"description": "Ninja-fast image generation with cyberpunk-themed logging"
},
"talkai": {
"class": AsyncTalkaiImager,
"description": "Fast and reliable image generation with comprehensive error handling"
},
"piclumen": {
"class": AsyncPiclumenImager,
"description": "Professional photorealistic image generation with advanced processing"
},
"artbit": {
"class": AsyncArtbitImager,
"description": "Bit-perfect AI art creation with precise control over parameters"
},
"huggingface": {
"class": AsyncHFimager,
"description": "Direct integration with HuggingFace's powerful models"
},
}
# Add AIArta provider if available
if AIARTA_AVAILABLE:
PROVIDER_MAP["aiarta"] = {
"class": AsyncAIArtaImager,
"description": "Generate stunning AI art with AI Arta with 45+ artistic styles"
}
# Provider model info
PROVIDER_MODEL_INFO = {
"blackbox": {
"default": "blackbox-default",
"models": ["blackbox-default"],
"default_params": {}
},
"deepinfra": {
"default": "flux-1-schnell",
"models": ["flux-1-schnell"],
"default_params": {
"num_inference_steps": 25,
"guidance_scale": 7.5,
"width": 1024,
"height": 1024
}
},
"aiforce": {
"default": "flux-1-pro",
"models": [
"stable-diffusion-xl-lightning",
"stable-diffusion-xl-base",
"flux-1-pro",
"ideogram",
"flux",
"flux-realism",
"flux-anime",
"flux-3d",
"flux-disney",
"flux-pixel",
"flux-4o",
"any-dark"
],
"default_params": {
"width": 768,
"height": 768
}
},
"nexra": {
"default": "midjourney",
"models": [
"emi",
"stablediffusion-1-5",
"stablediffusion-2-1",
"sdxl-lora",
"dalle",
"dalle2",
"dalle-mini",
"flux",
"midjourney",
"dreamshaper-xl",
"dynavision-xl",
"juggernaut-xl",
"realism-engine-sdxl",
"sd-xl-base-1-0",
"animagine-xl-v3",
"sd-xl-base-inpainting",
"turbovision-xl",
"devlish-photorealism-sdxl",
"realvis-xl-v4"
],
"default_params": {}
},
"freeai": {
"default": "dall-e-3",
"models": [
"dall-e-3",
"flux-pro-ultra",
"flux-pro",
"flux-pro-ultra-raw",
"flux-schnell",
"flux-realism",
"grok-2-aurora"
],
"default_params": {
"size": "1024x1024",
"quality": "standard",
"style": "vivid"
}
},
"ninja": {
"default": "flux-dev",
"models": ["stable-diffusion", "flux-dev"],
"default_params": {}
},
"talkai": {
"default": "talkai-default",
"models": ["talkai-default"],
"default_params": {}
},
"piclumen": {
"default": "piclumen-default",
"models": ["piclumen-default"],
"default_params": {}
},
"artbit": {
"default": "sdxl",
"models": ["sdxl", "sd"],
"default_params": {
"selected_ratio": "1024"
}
},
"huggingface": {
"default": "stable-diffusion-xl-base-1-0",
"models": ["stable-diffusion-xl-base-1-0", "stable-diffusion-v1-5"],
"default_params": {
"guidance_scale": 7.5,
"num_inference_steps": 30
}
}
}
# Normalize model names to OpenAI-like format
for provider, info in PROVIDER_MODEL_INFO.items():
info["models"] = [model.replace("/", "-").replace(".", "-").replace("_", "-").lower() for model in info["models"]]
info["default"] = info["default"].replace("/", "-").replace(".", "-").replace("_", "-").lower()
# Add AIArta model info if available
if AIARTA_AVAILABLE:
PROVIDER_MODEL_INFO["aiarta"] = {
"default": "flux",
"models": [
"flux", "medieval", "vincent-van-gogh", "f-dev", "low-poly",
"dreamshaper-xl", "anima-pencil-xl", "biomech", "trash-polka",
"no-style", "cheyenne-xl", "chicano", "embroidery-tattoo",
"red-and-black", "fantasy-art", "watercolor", "dotwork",
"old-school-colored", "realistic-tattoo", "japanese-2",
"realistic-stock-xl", "f-pro", "revanimated", "katayama-mix-xl",
"sdxl-l", "cor-epica-xl", "anime-tattoo", "new-school",
"death-metal", "old-school", "juggernaut-xl", "photographic",
"sdxl-1-0", "graffiti", "mini-tattoo", "surrealism",
"neo-traditional", "on-limbs-black", "yamers-realistic-xl",
"pony-xl", "playground-xl", "anything-xl", "flame-design",
"kawaii", "cinematic-art", "professional", "flux-black-ink"
],
"default_params": {
"negative_prompt": "blurry, deformed hands, ugly",
"guidance_scale": 7,
"num_inference_steps": 30,
"aspect_ratio": "1:1"
}
}
# Define Pydantic models for request and response validation (OpenAI-compatible)
class ImageSize(BaseModel):
width: int = Field(1024, description="Image width")
height: int = Field(1024, description="Image height")
class ImageGenerationRequest(BaseModel):
model: str = Field(..., description="The model to use for image generation")
prompt: str = Field(..., description="The prompt to generate images from")
n: Optional[int] = Field(1, description="Number of images to generate", ge=1, le=10)
size: Optional[str] = Field("1024x1024", description="Image size in format WIDTHxHEIGHT")
response_format: Optional[str] = Field("url", description="The format in which the generated images are returned", enum=["url", "b64_json"])
user: Optional[str] = Field(None, description="A unique identifier for the user")
style: Optional[str] = Field(None, description="Style for the generation")
quality: Optional[str] = Field(None, description="Quality level for the generation")
negative_prompt: Optional[str] = Field(None, description="What to avoid in the generated image")
class ImageData(BaseModel):
url: Optional[str] = Field(None, description="The URL of the generated image")
b64_json: Optional[str] = Field(None, description="Base64 encoded JSON string of the image")
revised_prompt: Optional[str] = Field(None, description="The prompt after any revisions")
class ImageGenerationResponse(BaseModel):
created: int = Field(..., description="Unix timestamp for when the request was created")
data: List[ImageData] = Field(..., description="List of generated images")
class ModelsListResponse(BaseModel):
object: str = Field("list", description="Object type")
data: List[Dict[str, Any]] = Field(..., description="List of available models")
class ErrorResponse(BaseModel):
error: Dict[str, Any] = Field(..., description="Error details")
# Error handling
class APIError(Exception):
def __init__(self, message, code=400, param=None, type="invalid_request_error"):
self.message = message
self.code = code
self.param = param
self.type = type
# Authentication dependency
async def verify_api_key(authorization: Optional[str] = Header(None)):
if authorization is None:
raise HTTPException(
status_code=401,
detail={
"error": {
"message": "No API key provided",
"type": "authentication_error",
"param": None,
"code": "no_api_key"
}
}
)
# Extract the key from the Authorization header
parts = authorization.split()
if len(parts) != 2 or parts[0].lower() != "bearer":
raise HTTPException(
status_code=401,
detail={
"error": {
"message": "Invalid authentication format. Use 'Bearer YOUR_API_KEY'",
"type": "authentication_error",
"param": None,
"code": "invalid_auth_format"
}
}
)
api_key = parts[1]
# Check if the API key is valid
# In production, you'd want to use a more secure method
if api_key not in API_KEYS:
raise HTTPException(
status_code=401,
detail={
"error": {
"message": "Invalid API key",
"type": "authentication_error",
"param": None,
"code": "invalid_api_key"
}
}
)
return api_key
# Find provider from model ID - updating this function to support provider/model format
def get_provider_for_model(model: str):
model = model.lower()
# Check if it's in the format 'provider/model'
if "/" in model:
provider_name, model_name = model.split("/", 1)
model_name = model_name.replace("/", "-").replace(".", "-").replace("_", "-").lower()
# Check if provider exists
if provider_name not in PROVIDER_MAP:
raise APIError(
message=f"Provider '{provider_name}' not found",
code=404,
type="provider_not_found"
)
# Check if model exists for this provider
provider_models = PROVIDER_MODEL_INFO[provider_name]["models"]
if model_name not in provider_models:
# Try searching with less normalization - some providers might use underscore variants
original_model_name = model_name.replace("-", "_")
if original_model_name not in [m.replace("-", "_") for m in provider_models]:
raise APIError(
message=f"Model '{model_name}' not found for provider '{provider_name}'",
code=404,
type="model_not_found"
)
return provider_name, model_name
# If not in provider/model format, search all providers (original behavior)
for provider_name, provider_info in PROVIDER_MODEL_INFO.items():
# Check if this model belongs to this provider
if model in provider_info["models"] or model == provider_info["default"]:
return provider_name, model
# If no provider found, return error
raise APIError(
message=f"Model '{model}' not found",
code=404,
type="model_not_found"
)
# Health check endpoint
@app.get("/health", response_model=Dict[str, str])
async def health_check():
return {"status": "ok"}
# OpenAI-compatible endpoints
# List available models
@app.get("/v1/models", response_model=ModelsListResponse, dependencies=[Depends(verify_api_key)])
async def list_models():
models_data = []
for provider_name, provider_info in PROVIDER_MODEL_INFO.items():
provider_description = PROVIDER_MAP.get(provider_name, {}).get("description", "")
for model_name in provider_info["models"]:
is_default = model_name == provider_info["default"]
models_data.append({
"id": model_name,
"object": "model",
"created": int(time.time()),
"owned_by": provider_name,
"permission": [],
"root": model_name,
"parent": None,
"description": f"{provider_description} - {'Default model' if is_default else 'Alternative model'}",
})
return {
"object": "list",
"data": models_data
}
# Get model information
@app.get("/v1/models/{model_id}", dependencies=[Depends(verify_api_key)])
async def get_model(model_id: str):
try:
provider_name, model = get_provider_for_model(model_id)
provider_description = PROVIDER_MAP.get(provider_name, {}).get("description", "")
return {
"id": model,
"object": "model",
"created": int(time.time()),
"owned_by": provider_name,
"permission": [],
"root": model,
"parent": None,
"description": provider_description
}
except APIError as e:
return JSONResponse(
status_code=e.code,
content={"error": {"message": e.message, "type": e.type, "param": e.param, "code": e.code}}
)
# Generate images
@app.post("/v1/images/generations", response_model=ImageGenerationResponse, dependencies=[Depends(verify_api_key)])
async def create_image(request: ImageGenerationRequest, background_tasks: BackgroundTasks):
try:
# Get provider for the requested model
provider_name, model = get_provider_for_model(request.model)
provider_class = PROVIDER_MAP[provider_name]["class"]
# Parse size
width, height = 1024, 1024
if request.size:
try:
size_parts = request.size.split("x")
if len(size_parts) == 2:
width, height = int(size_parts[0]), int(size_parts[1])
else:
width = height = int(size_parts[0])
except:
pass
# Create task ID
task_id = str(uuid.uuid4())
IMAGE_STORAGE[task_id] = {"status": "processing", "images": []}
# Get default params and update with user-provided values
default_params = PROVIDER_MODEL_INFO[provider_name].get("default_params", {}).copy()
# Add additional parameters from the request
if request.negative_prompt:
default_params["negative_prompt"] = request.negative_prompt
if request.quality:
default_params["quality"] = request.quality
if request.style:
default_params["style"] = request.style
# Update size parameters
default_params["width"] = width
default_params["height"] = height
# Function to generate images in the background
async def generate_images():
try:
# Initialize provider based on the provider name
if provider_name == "freeai":
provider_instance = provider_class(model=model)
elif provider_name == "deepinfra" and "-flux-" in model:
# Convert back to model format expected by provider
original_model = "black-forest-labs/FLUX-1-schnell"
provider_instance = provider_class(model=original_model)
else:
provider_instance = provider_class()
# Generate images with provider-specific parameters
# Each provider may have different parameter requirements
if provider_name == "aiforce":
images = await provider_instance.generate(
prompt=request.prompt,
amount=request.n,
model=model.replace("-", "_"), # Convert back to format used by provider
width=default_params.get("width", 768),
height=default_params.get("height", 768),
seed=default_params.get("seed", None)
)
elif provider_name == "deepinfra":
images = await provider_instance.generate(
prompt=request.prompt,
amount=request.n,
num_inference_steps=default_params.get("num_inference_steps", 25),
guidance_scale=default_params.get("guidance_scale", 7.5),
width=default_params.get("width", 1024),
height=default_params.get("height", 1024),
seed=default_params.get("seed", None)
)
elif provider_name == "nexra":
# Convert back to original model format
original_model = model.replace("-", "_")
images = await provider_instance.generate(
prompt=request.prompt,
amount=request.n,
model=original_model,
additional_params=default_params
)
elif provider_name == "freeai":
images = await provider_instance.generate(
prompt=request.prompt,
amount=request.n,
size=f"{width}x{height}",
quality=default_params.get("quality", "standard"),
style=default_params.get("style", "vivid")
)
elif provider_name == "ninja":
images = await provider_instance.generate(
prompt=request.prompt,
amount=request.n,
model=model.replace("-", "_")
)
elif provider_name == "artbit":
images = await provider_instance.generate(
prompt=request.prompt,
amount=request.n,
caption_model=model,
selected_ratio=default_params.get("selected_ratio", "1024"),
negative_prompt=default_params.get("negative_prompt", "")
)
elif provider_name == "huggingface":
# Convert from dash format to slash format for HF
original_model = model.replace("-", "/")
images = await provider_instance.generate(
prompt=request.prompt,
amount=request.n,
model=original_model,
guidance_scale=default_params.get("guidance_scale", 7.5),
negative_prompt=default_params.get("negative_prompt", None),
num_inference_steps=default_params.get("num_inference_steps", 30),
width=width,
height=height
)
elif provider_name == "aiarta" and AIARTA_AVAILABLE:
images = await provider_instance.generate(
prompt=request.prompt,
amount=request.n,
model=model,
negative_prompt=default_params.get("negative_prompt", "blurry, deformed hands, ugly"),
guidance_scale=default_params.get("guidance_scale", 7),
num_inference_steps=default_params.get("num_inference_steps", 30),
aspect_ratio=default_params.get("aspect_ratio", "1:1")
)
else:
# Default case for providers with simpler interfaces
images = await provider_instance.generate(
prompt=request.prompt,
amount=request.n
)
# Process and store the generated images
for i, img in enumerate(images):
# Handle both URL strings and binary data
if isinstance(img, str):
# For providers that return URLs instead of binary data
async with aiohttp.ClientSession() as session:
async with session.get(img) as resp:
resp.raise_for_status()
img_data = await resp.read()
else:
img_data = img
# Generate a unique URL for the image
image_id = f"{i}"
image_url = f"/v1/images/{task_id}/{image_id}"
# Store image data based on requested format
if request.response_format == "b64_json":
encoded = base64.b64encode(img_data).decode('utf-8')
IMAGE_STORAGE[task_id]["images"].append({
"image_id": image_id,
"data": encoded,
"url": image_url,
})
else: # Default to URL
IMAGE_STORAGE[task_id]["images"].append({
"image_id": image_id,
"data": img_data,
"url": image_url,
})
# Update task status
IMAGE_STORAGE[task_id]["status"] = "completed"
except Exception as e:
# Handle errors
IMAGE_STORAGE[task_id]["status"] = "failed"
IMAGE_STORAGE[task_id]["error"] = str(e)
# Start background task
background_tasks.add_task(generate_images)
# Immediate response with task details
# For compatibility, we need to structure this like OpenAI's response
created_timestamp = int(time.time())
# Wait briefly to allow the background task to start
await asyncio.sleep(0.1)
# Check if the task failed immediately
if IMAGE_STORAGE[task_id]["status"] == "failed":
error_message = IMAGE_STORAGE[task_id].get("error", "Unknown error")
raise APIError(message=f"Image generation failed: {error_message}", code=500)
# Prepare response data
image_data = []
for i in range(request.n):
if request.response_format == "b64_json":
image_data.append({
"b64_json": "", # Will be filled in by the background task
"revised_prompt": request.prompt
})
else:
image_data.append({
"url": f"/v1/images/{task_id}/{i}",
"revised_prompt": request.prompt
})
return {
"created": created_timestamp,
"data": image_data
}
except APIError as e:
return JSONResponse(
status_code=e.code,
content={"error": {"message": e.message, "type": e.type, "param": e.param, "code": e.code}}
)
except Exception as e:
return JSONResponse(
status_code=500,
content={"error": {"message": str(e), "type": "server_error", "param": None, "code": 500}}
)
# Image retrieval endpoint
@app.get("/v1/images/{task_id}/{image_id}", dependencies=[Depends(verify_api_key)])
async def get_image(task_id: str, image_id: str):
if task_id not in IMAGE_STORAGE:
return JSONResponse(
status_code=404,
content={"error": {"message": f"Image not found", "type": "not_found_error"}}
)
task_data = IMAGE_STORAGE[task_id]
if task_data["status"] == "failed":
return JSONResponse(
status_code=500,
content={"error": {"message": f"Image generation failed: {task_data.get('error', 'Unknown error')}", "type": "processing_error"}}
)
if task_data["status"] == "processing":
return JSONResponse(
status_code=202,
content={"status": "processing", "message": "Image is still being generated"}
)
# Find the requested image
for img in task_data["images"]:
if img["image_id"] == image_id:
# If it's stored as base64, it's already in the right format
if isinstance(img["data"], str):
return JSONResponse(content={"b64_json": img["data"]})
# If it's binary data, return as an image stream
return StreamingResponse(
io.BytesIO(img["data"]),
media_type="image/png"
)
return JSONResponse(
status_code=404,
content={"error": {"message": f"Image not found", "type": "not_found_error"}}
)
# Legacy endpoints for backward compatibility
@app.get("/providers")
async def list_providers_legacy():
providers = {}
for provider_name, provider_info in PROVIDER_MAP.items():
model_info = PROVIDER_MODEL_INFO.get(provider_name, {})
providers[provider_name] = {
"description": provider_info.get("description", ""),
"default_model": model_info.get("default", "default"),
"models": model_info.get("models", ["default"]),
"default_params": model_info.get("default_params", {})
}
return providers
# Main entry point
if __name__ == "__main__":
uvicorn.run(
"app:app",
host="0.0.0.0",
port=8000,
reload=True
)