""" Utility functions and constants for AI-Inferoxy AI Hub. Contains configuration constants and helper functions. """ import os import re # Configuration constants DEFAULT_CHAT_MODEL = "openai/gpt-oss-20b" DEFAULT_IMAGE_MODEL = "Qwen/Qwen-Image" DEFAULT_IMAGE_TO_IMAGE_MODEL = "Qwen/Qwen-Image-Edit" DEFAULT_TTS_MODEL = "hexgrad/Kokoro-82M" DEFAULT_VIDEO_MODEL = "Wan-AI/Wan2.2-T2V-A14B" # Unified default provider used by all non-chat tasks DEFAULT_PROVIDER = "auto" # Chat configuration CHAT_CONFIG = { "max_tokens": 1024, "temperature": 0.7, "top_p": 0.95, "system_message": "You are a helpful and friendly AI assistant. Provide clear, accurate, and helpful responses." } # Image generation configuration IMAGE_CONFIG = { "width": 1024, "height": 1024, "num_inference_steps": 20, "guidance_scale": 7.5, "seed": -1, "negative_prompt": "blurry, low quality, distorted, deformed, ugly, bad anatomy" } # Supported providers (unified across tasks) PROVIDERS_UNIFIED = [ "auto", "cerebras", "cohere", "fal-ai", "featherless-ai", "fireworks-ai", "groq", "hf-inference", "hyperbolic", "nebius", "novita", "nscale", "replicate", "sambanova", "together", ] # Backwards compatibility exported lists CHAT_PROVIDERS = PROVIDERS_UNIFIED IMAGE_PROVIDERS = PROVIDERS_UNIFIED # Popular models for quick access POPULAR_CHAT_MODELS = [ "openai/gpt-oss-20b", "meta-llama/Llama-2-7b-chat-hf", "microsoft/DialoGPT-medium", "google/flan-t5-base" ] POPULAR_IMAGE_MODELS = [ "Qwen/Qwen-Image", "black-forest-labs/FLUX.1-dev", "stabilityai/stable-diffusion-xl-base-1.0", "runwayml/stable-diffusion-v1-5" ] # Suggested model lists (users can still input any model id) SUGGESTED_CHAT_MODELS = [ "openai/gpt-oss-20b", "openai/gpt-oss-120b", "deepseek-ai/DeepSeek-V3.1", "zai-org/GLM-4.5", "Qwen/Qwen3-8B", "meta-llama/Llama-3.1-8B-Instruct", "deepseek-ai/DeepSeek-R1", "moonshotai/Kimi-K2-Instruct", "Qwen/Qwen3-Coder-30B-A3B-Instruct", "CohereLabs/command-a-reasoning-08-2025", ] SUGGESTED_IMAGE_MODELS = [ "Qwen/Qwen-Image", "black-forest-labs/FLUX.1-dev", "black-forest-labs/FLUX.1-Krea-dev", "stabilityai/stable-diffusion-xl-base-1.0", "black-forest-labs/FLUX.1-schnell", "UmeAiRT/FLUX.1-dev-LoRA-Modern_Pixel_art", "xey/sldr_flux_nsfw_v2-studio", "HiDream-ai/HiDream-I1-Full", "Kwai-Kolors/Kolors", ] SUGGESTED_IMAGE_TO_IMAGE_MODELS = [ "Qwen/Qwen-Image-Edit", "Kontext-Style/Ghibli_lora", "black-forest-labs/FLUX.1-Kontext-dev", "fofr/kontext-make-person-real", "jerrrycans/watermark20000", "fal/Pencil-Drawing-Kontext-Dev-LoRA", ] SUGGESTED_VIDEO_MODELS = [ "Wan-AI/Wan2.2-T2V-A14B", "Wan-AI/Wan2.2-TI2V-5B", "tencent/HunyuanVideo", "Wan-AI/Wan2.2-T2V-A14B-Diffusers", "zai-org/CogVideoX-5b", "Wan-AI/Wan2.1-T2V-14B", "genmo/mochi-1-preview", "Wan-AI/Wan2.1-T2V-1.3B", "Lightricks/LTX-Video-0.9.7-dev", "Lightricks/LTX-Video-0.9.5", "Lightricks/LTX-Video-0.9.7-distilled", ] # Model-specific configurations for TTS TTS_MODEL_CONFIGS = { "hexgrad/Kokoro-82M": { "type": "kokoro", "supports_voice": True, "supports_speed": True, "extra_body_params": ["voice", "speed"] }, "ResembleAI/chatterbox": { "type": "chatterbox", "supports_voice": False, "supports_speed": False, "extra_body_params": ["audio_url", "exaggeration", "temperature", "cfg"] }, "nari-labs/Dia-1.6B": { "type": "dia", "supports_voice": False, "supports_speed": False, "extra_body_params": [] } } # ----------------------------- # Text-to-Video configuration # ----------------------------- # Example prompts for text-to-video generation VIDEO_EXAMPLE_PROMPTS = [ "A young man walking on the street", "A corgi puppy running through a field of flowers, cinematic", "A futuristic city skyline at sunset with flying cars, 4k", "A serene beach with gentle waves and palm trees swaying", ] # Voice options for Kokoro TTS (based on the reference app) TTS_VOICES = { 'πŸ‡ΊπŸ‡Έ 🚺 Heart ❀️': 'af_heart', 'πŸ‡ΊπŸ‡Έ 🚺 Bella πŸ”₯': 'af_bella', 'πŸ‡ΊπŸ‡Έ 🚺 Nicole 🎧': 'af_nicole', 'πŸ‡ΊπŸ‡Έ 🚺 Aoede': 'af_aoede', 'πŸ‡ΊπŸ‡Έ 🚺 Kore': 'af_kore', 'πŸ‡ΊπŸ‡Έ 🚺 Sarah': 'af_sarah', 'πŸ‡ΊπŸ‡Έ 🚺 Nova': 'af_nova', 'πŸ‡ΊπŸ‡Έ 🚺 Sky': 'af_sky', 'πŸ‡ΊπŸ‡Έ 🚺 Alloy': 'af_alloy', 'πŸ‡ΊπŸ‡Έ 🚺 Jessica': 'af_jessica', 'πŸ‡ΊπŸ‡Έ 🚺 River': 'af_river', 'πŸ‡ΊπŸ‡Έ 🚹 Michael': 'am_michael', 'πŸ‡ΊπŸ‡Έ 🚹 Fenrir': 'am_fenrir', 'πŸ‡ΊπŸ‡Έ 🚹 Puck': 'am_puck', 'πŸ‡ΊπŸ‡Έ 🚹 Echo': 'am_echo', 'πŸ‡ΊπŸ‡Έ 🚹 Eric': 'am_eric', 'πŸ‡ΊπŸ‡Έ 🚹 Liam': 'am_liam', 'πŸ‡ΊπŸ‡Έ 🚹 Onyx': 'am_onyx', 'πŸ‡ΊπŸ‡Έ 🚹 Santa': 'am_santa', 'πŸ‡ΊπŸ‡Έ 🚹 Adam': 'am_adam', 'πŸ‡¬πŸ‡§ 🚺 Emma': 'bf_emma', 'πŸ‡¬πŸ‡§ 🚺 Isabella': 'bf_isabella', 'πŸ‡¬πŸ‡§ 🚺 Alice': 'bf_alice', 'πŸ‡¬πŸ‡§ 🚺 Lily': 'bf_lily', 'πŸ‡¬πŸ‡§ 🚹 George': 'bm_george', 'πŸ‡¬πŸ‡§ 🚹 Fable': 'bm_fable', 'πŸ‡¬πŸ‡§ 🚹 Lewis': 'bm_lewis', 'πŸ‡¬πŸ‡§ 🚹 Daniel': 'bm_daniel', } # Example prompts for chat CHAT_EXAMPLE_PROMPTS = [ "What's a polite way to introduce myself at a networking event?", "Can you suggest a fun icebreaker question for a group chat?", "Explain the concept of entropy in simple terms suitable for a high school student.", "Summarize the main differences between classical and operant conditioning.", "Is it possible for artificial intelligence to possess consciousness? Discuss briefly.", "What does 'the map is not the territory' mean in philosophy?", "Write a Python function to reverse a linked list.", "How can I optimize a SQL query for faster performance?", "Suggest 3 imaginative prompts for generating images of futuristic cities.", "Give me 3 creative prompts for generating surreal animal portraits.", ] # Example prompts for image generation IMAGE_EXAMPLE_PROMPTS = [ "A majestic dragon flying over a medieval castle, epic fantasy art, detailed, 8k", "A serene Japanese garden with cherry blossoms, zen atmosphere, peaceful, high quality", "A futuristic cityscape with flying cars and neon lights, cyberpunk style, cinematic", "A cute robot cat playing with yarn, adorable, cartoon style, vibrant colors", "A magical forest with glowing mushrooms and fairy lights, fantasy, ethereal beauty", "Portrait of a wise old wizard with flowing robes, magical aura, fantasy character art", "A cozy coffee shop on a rainy day, warm lighting, peaceful atmosphere, detailed", "An astronaut floating in space with Earth in background, photorealistic, stunning" ] # Example prompts for image-to-image generation IMAGE_TO_IMAGE_EXAMPLE_PROMPTS = [ "Turn the cat into a tiger with stripes and fierce expression", "Turn this image into the Ghibli style.", "Make the background a magical forest with glowing mushrooms", "Change the style to vintage comic book with bold colors", "Add a superhero cape and mask to the person", "Transform the building into a futuristic skyscraper", "Make the flowers bloom and add butterflies around them", "Change the weather to a stormy night with lightning", "Add a magical portal in the background with sparkles" ] # Example texts for text-to-speech generation TTS_EXAMPLE_TEXTS = [ "Hello! Welcome to the amazing world of AI-powered text-to-speech technology.", "The quick brown fox jumps over the lazy dog. This pangram contains every letter of the alphabet.", "In a world where technology advances at lightning speed, artificial intelligence continues to reshape our future.", "Imagine a world where machines can understand and respond to human emotions with perfect clarity.", "The future belongs to those who believe in the beauty of their dreams and have the courage to pursue them.", "Science is not only compatible with spirituality; it is a profound source of spirituality.", "The only way to do great work is to love what you do. If you haven't found it yet, keep looking.", "Life is what happens when you're busy making other plans. Embrace every moment with gratitude.", "[S1] Dia is an open weights text to dialogue model. [S2] You get full control over scripts and voices. [S1] Wow. Amazing. (laughs) [S2] Try it now." ] # Example audio URLs for Chatterbox TTS TTS_EXAMPLE_AUDIO_URLS = [ "https://github.com/nazdridoy/kokoro-tts/raw/main/previews/demo.mp3", "https://storage.googleapis.com/chatterbox-demo-samples/prompts/male_rickmorty.mp3" ] def get_proxy_key(): """Get the proxy API key from environment variables.""" return os.getenv("PROXY_KEY") def validate_proxy_key(): """Validate that the proxy key is available.""" proxy_key = get_proxy_key() if not proxy_key: return False, "❌ Error: PROXY_KEY not found in environment variables. Please set it in your HuggingFace Space secrets." return True, "" def get_proxy_url(): """Get the proxy URL from environment variables.""" return os.getenv("PROXY_URL") def validate_proxy_url(): """Validate that the proxy URL is available.""" proxy_url = get_proxy_url() if not proxy_url: return False, "❌ Error: PROXY_URL not found in environment variables. Please set it in your HuggingFace Space secrets." return True, "" def format_error_message(error_type, error_message): """Format error messages consistently.""" return f"❌ {error_type}: {error_message}" def format_success_message(operation, details=""): """Format success messages consistently.""" base_message = f"βœ… {operation} completed successfully" if details: return f"{base_message}: {details}" return f"{base_message}!" def get_gradio_theme(): """Get the default Gradio theme for the application.""" try: import gradio as gr return gr.themes.Soft() except ImportError: return None # ----------------------------- # Reasoning () utilities # ----------------------------- def render_with_reasoning_toggle(text: str, show_reasoning: bool) -> str: """Render assistant text while optionally revealing content inside .... Behavior: - When show_reasoning is True: * Replace the opening tag with a collapsible HTML
block and an opening fenced code block. Stream reasoning tokens inside this block as they arrive. * Replace the closing tag with the closing fence and
when it appears. - When show_reasoning is False: * Remove complete ... blocks. * For partial streams (no closing tag yet), trim everything from the first onward. Safe to call on every streamed chunk; conversions are idempotent. """ if not isinstance(text, str): return text # If we are NOT showing reasoning, remove it entirely. For partial streams, hide from onwards. if not show_reasoning: if "" not in text: return text if "" not in text: return text.split("", 1)[0] # Remove complete ... blocks pattern_strip = re.compile(r"[\s\S]*?", re.IGNORECASE) return pattern_strip.sub("", text) # Show reasoning: stream it as it arrives by converting tags into a collapsible details block open_block_open = "
Reasoning\n\n```text\n" open_block_closed = "
Reasoning\n\n```text\n" close_block = "\n```\n
\n" # If the closing tag is not present yet, keep the block expanded while streaming if "" not in text: # Replace any raw with an expanded details block text = re.sub(r"", open_block_open, text, flags=re.IGNORECASE) # If for any reason a closed details opening exists, switch it to open (expanded) text = text.replace(open_block_closed, open_block_open) return text # If the closing tag is present, render a collapsed block by default # 1) Ensure opening is the closed variant text = re.sub(r"", open_block_closed, text, flags=re.IGNORECASE) text = text.replace(open_block_open, open_block_closed) # 2) Close the block text = re.sub(r"", close_block, text, flags=re.IGNORECASE) return text