import re import random import string import uuid import json import logging import asyncio import base64 from aiohttp import ClientSession, ClientTimeout, ClientError from fastapi import FastAPI, HTTPException, Request from pydantic import BaseModel from typing import List, Dict, Any, Optional, AsyncGenerator, Union from datetime import datetime from fastapi.responses import StreamingResponse # Configure logging logging.basicConfig( level=logging.INFO, format="%(asctime)s [%(levelname)s] %(name)s: %(message)s", handlers=[ logging.StreamHandler() ] ) logger = logging.getLogger(__name__) # Custom exception for model not working class ModelNotWorkingException(Exception): def __init__(self, model: str): self.model = model self.message = f"The model '{model}' is currently not working. Please try another model or wait for it to be fixed." super().__init__(self.message) # Proper implementation for ImageResponse and to_data_uri class ImageResponse: def __init__(self, data_uri: str, alt: str): self.data_uri = data_uri self.alt = alt def to_data_uri(image: bytes, mime_type: str = "image/png") -> str: encoded = base64.b64encode(image).decode('utf-8') return f"data:{mime_type};base64,{encoded}" def decode_base64_image(data_uri: str) -> bytes: try: header, encoded = data_uri.split(",", 1) return base64.b64decode(encoded) except Exception as e: logger.error(f"Error decoding base64 image: {e}") raise e class Blackbox: url = "https://www.blackbox.ai" api_endpoint = "https://www.blackbox.ai/api/chat" working = True supports_stream = True supports_system_message = True supports_message_history = True default_model = 'blackboxai' image_models = ['ImageGeneration'] models = [ default_model, 'blackboxai-pro', "llama-3.1-8b", 'llama-3.1-70b', 'llama-3.1-405b', 'gpt-4o', 'gemini-pro', 'gemini-1.5-flash', 'claude-sonnet-3.5', 'PythonAgent', 'JavaAgent', 'JavaScriptAgent', 'HTMLAgent', 'GoogleCloudAgent', 'AndroidDeveloper', 'SwiftDeveloper', 'Next.jsAgent', 'MongoDBAgent', 'PyTorchAgent', 'ReactAgent', 'XcodeAgent', 'AngularJSAgent', *image_models, 'Niansuh', ] agentMode = { 'ImageGeneration': {'mode': True, 'id': "ImageGenerationLV45LJp", 'name': "Image Generation"}, 'Niansuh': {'mode': True, 'id': "NiansuhAIk1HgESy", 'name': "Niansuh"}, } trendingAgentMode = { "blackboxai": {}, "gemini-1.5-flash": {'mode': True, 'id': 'Gemini'}, "llama-3.1-8b": {'mode': True, 'id': "llama-3.1-8b"}, 'llama-3.1-70b': {'mode': True, 'id': "llama-3.1-70b"}, 'llama-3.1-405b': {'mode': True, 'id': "llama-3.1-405b"}, 'blackboxai-pro': {'mode': True, 'id': "BLACKBOXAI-PRO"}, 'PythonAgent': {'mode': True, 'id': "Python Agent"}, 'JavaAgent': {'mode': True, 'id': "Java Agent"}, 'JavaScriptAgent': {'mode': True, 'id': "JavaScript Agent"}, 'HTMLAgent': {'mode': True, 'id': "HTML Agent"}, 'GoogleCloudAgent': {'mode': True, 'id': "Google Cloud Agent"}, 'AndroidDeveloper': {'mode': True, 'id': "Android Developer"}, 'SwiftDeveloper': {'mode': True, 'id': "Swift Developer"}, 'Next.jsAgent': {'mode': True, 'id': "Next.js Agent"}, 'MongoDBAgent': {'mode': True, 'id': "MongoDB Agent"}, 'PyTorchAgent': {'mode': True, 'id': "PyTorch Agent"}, 'ReactAgent': {'mode': True, 'id': "React Agent"}, 'XcodeAgent': {'mode': True, 'id': "Xcode Agent"}, 'AngularJSAgent': {'mode': True, 'id': "AngularJS Agent"}, } userSelectedModel = { "gpt-4o": "gpt-4o", "gemini-pro": "gemini-pro", 'claude-sonnet-3.5': "claude-sonnet-3.5", } model_prefixes = { 'gpt-4o': '@GPT-4o', 'gemini-pro': '@Gemini-PRO', 'claude-sonnet-3.5': '@Claude-Sonnet-3.5', 'PythonAgent': '@Python Agent', 'JavaAgent': '@Java Agent', 'JavaScriptAgent': '@JavaScript Agent', 'HTMLAgent': '@HTML Agent', 'GoogleCloudAgent': '@Google Cloud Agent', 'AndroidDeveloper': '@Android Developer', 'SwiftDeveloper': '@Swift Developer', 'Next.jsAgent': '@Next.js Agent', 'MongoDBAgent': '@MongoDB Agent', 'PyTorchAgent': '@PyTorch Agent', 'ReactAgent': '@React Agent', 'XcodeAgent': '@Xcode Agent', 'AngularJSAgent': '@AngularJS Agent', 'blackboxai-pro': '@BLACKBOXAI-PRO', 'ImageGeneration': '@Image Generation', 'Niansuh': '@Niansuh', } model_referers = { "blackboxai": f"{url}/?model=blackboxai", "gpt-4o": f"{url}/?model=gpt-4o", "gemini-pro": f"{url}/?model=gemini-pro", "claude-sonnet-3.5": f"{url}/?model=claude-sonnet-3.5" } model_aliases = { "gemini-flash": "gemini-1.5-flash", "claude-3.5-sonnet": "claude-sonnet-3.5", "flux": "ImageGeneration", "niansuh": "Niansuh", } @classmethod def get_model(cls, model: str) -> str: if model in cls.models: return model elif model in cls.userSelectedModel: return cls.userSelectedModel[model] elif model in cls.model_aliases: return cls.model_aliases[model] else: return cls.default_model @classmethod async def create_async_generator( cls, model: str, messages: List[Dict[str, Any]], proxy: Optional[str] = None, image: Optional[str] = None, # Expecting a base64 string image_name: Optional[str] = None, webSearchMode: bool = False, **kwargs ) -> AsyncGenerator[Any, None]: model = cls.get_model(model) logger.info(f"Selected model: {model}") if not cls.working or model not in cls.models: logger.error(f"Model {model} is not working or not supported.") raise ModelNotWorkingException(model) headers = { "accept": "*/*", "accept-language": "en-US,en;q=0.9", "cache-control": "no-cache", "content-type": "application/json", "origin": cls.url, "pragma": "no-cache", "priority": "u=1, i", "referer": cls.model_referers.get(model, cls.url), "sec-ch-ua": '"Chromium";v="129", "Not=A?Brand";v="8"', "sec-ch-ua-mobile": "?0", "sec-ch-ua-platform": '"Linux"', "sec-fetch-dest": "empty", "sec-fetch-mode": "cors", "sec-fetch-site": "same-origin", "user-agent": "Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/129.0.0.0 Safari/537.36", } if model in cls.model_prefixes: prefix = cls.model_prefixes[model] if not messages[0]['content'].startswith(prefix): logger.debug(f"Adding prefix '{prefix}' to the first message.") messages[0]['content'] = f"{prefix} {messages[0]['content']}" random_id = ''.join(random.choices(string.ascii_letters + string.digits, k=7)) user_message = { "id": random_id, "role": 'user', "content": 'Hi' # This should be dynamically set based on input } if image is not None: try: image_bytes = decode_base64_image(image) data_uri = to_data_uri(image_bytes) user_message['data'] = { 'fileText': '', 'imageBase64': data_uri, 'title': image_name or "Uploaded Image" } user_message['content'] = 'FILE:BB\n$#$\n\n$#$\n' + user_message['content'] logger.debug("Image data added to the message.") except Exception as e: logger.error(f"Failed to decode base64 image: {e}") raise HTTPException(status_code=400, detail="Invalid image data provided.") # Update the last message with user_message if messages: messages[-1] = user_message else: messages.append(user_message) data = { "messages": messages, "id": random_id, "previewToken": None, "userId": None, "codeModelMode": True, "agentMode": {}, "trendingAgentMode": {}, "isMicMode": False, "userSystemPrompt": None, "maxTokens": 99999999, "playgroundTopP": 0.9, "playgroundTemperature": 0.5, "isChromeExt": False, "githubToken": None, "clickedAnswer2": False, "clickedAnswer3": False, "clickedForceWebSearch": False, "visitFromDelta": False, "mobileClient": False, "userSelectedModel": None, "webSearchMode": webSearchMode, } if model in cls.agentMode: data["agentMode"] = cls.agentMode[model] elif model in cls.trendingAgentMode: data["trendingAgentMode"] = cls.trendingAgentMode[model] elif model in cls.userSelectedModel: data["userSelectedModel"] = cls.userSelectedModel[model] logger.info(f"Sending request to {cls.api_endpoint} with data: {data}") timeout = ClientTimeout(total=60) # Set an appropriate timeout retry_attempts = 10 # Set the number of retry attempts for attempt in range(retry_attempts): try: async with ClientSession(headers=headers, timeout=timeout) as session: async with session.post(cls.api_endpoint, json=data, proxy=proxy) as response: response.raise_for_status() logger.info(f"Received response with status {response.status}") if model == 'ImageGeneration': response_text = await response.text() url_match = re.search(r'https://storage\.googleapis\.com/[^\s\)]+', response_text) if url_match: image_url = url_match.group(0) logger.info(f"Image URL found: {image_url}") # Fetch the image data async with session.get(image_url) as img_response: img_response.raise_for_status() image_bytes = await img_response.read() data_uri = to_data_uri(image_bytes) logger.info("Image converted to base64 data URI.") yield ImageResponse(data_uri, alt=messages[-1]['content']) else: logger.error("Image URL not found in the response.") raise Exception("Image URL not found in the response") else: full_response = "" search_results_json = "" try: async for chunk, _ in response.content.iter_chunks(): if chunk: decoded_chunk = chunk.decode(errors='ignore') decoded_chunk = re.sub(r'\$@\$v=[^$]+\$@\$', '', decoded_chunk) if decoded_chunk.strip(): if '$~~~$' in decoded_chunk: search_results_json += decoded_chunk else: full_response += decoded_chunk yield decoded_chunk logger.info("Finished streaming response chunks.") except Exception as e: logger.exception("Error while iterating over response chunks.") raise e if data["webSearchMode"] and search_results_json: match = re.search(r'\$~~~\$(.*?)\$~~~\$', search_results_json, re.DOTALL) if match: try: search_results = json.loads(match.group(1)) formatted_results = "\n\n**Sources:**\n" for i, result in enumerate(search_results[:5], 1): formatted_results += f"{i}. [{result['title']}]({result['link']})\n" logger.info("Formatted search results.") yield formatted_results except json.JSONDecodeError as je: logger.error("Failed to parse search results JSON.") raise je break # Exit the retry loop if successful except ClientError as ce: logger.error(f"Client error occurred: {ce}. Retrying attempt {attempt + 1}/{retry_attempts}") if attempt == retry_attempts - 1: raise HTTPException(status_code=502, detail="Error communicating with the external API. | NiansuhAI") except asyncio.TimeoutError: logger.error(f"Request timed out. Retrying attempt {attempt + 1}/{retry_attempts}") if attempt == retry_attempts - 1: raise HTTPException(status_code=504, detail="External API request timed out. | NiansuhAI") except Exception as e: logger.error(f"Unexpected error: {e}. Retrying attempt {attempt + 1}/{retry_attempts}") if attempt == retry_attempts - 1: raise HTTPException(status_code=500, detail=str(e)) # FastAPI app setup app = FastAPI() class Message(BaseModel): role: str content: str class ChatRequest(BaseModel): model: str messages: List[Message] stream: Optional[bool] = False webSearchMode: Optional[bool] = False image: Optional[str] = None # Add image field for base64 data def create_response(content: str, model: str, finish_reason: Optional[str] = None) -> Dict[str, Any]: return { "id": f"chatcmpl-{uuid.uuid4()}", "object": "chat.completion.chunk", "created": int(datetime.now().timestamp()), "model": model, "choices": [ { "index": 0, "delta": {"content": content, "role": "assistant"}, "finish_reason": finish_reason, } ], "usage": None, } @app.post("/niansuhai/v1/chat/completions") async def chat_completions(request: ChatRequest, req: Request): logger.info(f"Received chat completions request: model='{request.model}' messages={request.messages} stream={request.stream} webSearchMode={request.webSearchMode} image={request.image}") try: # Validate that all messages have string content for idx, msg in enumerate(request.messages): if not isinstance(msg.content, str): logger.error(f"Message at index {idx} has invalid content type: {type(msg.content)}") raise HTTPException( status_code=422, detail=[{ "type": "string_type", "loc": ["body", "messages", idx, "content"], "msg": "Input should be a valid string", "input": msg.content }] ) # Convert Pydantic messages to dicts messages = [{"role": msg.role, "content": msg.content} for msg in request.messages] async_generator = Blackbox.create_async_generator( model=request.model, messages=messages, proxy=None, # Pass proxy if needed image=request.image, # Pass the base64 image image_name=None, webSearchMode=request.webSearchMode ) if request.stream: async def generate(): try: async for chunk in async_generator: if isinstance(chunk, ImageResponse): # Use the base64 data URI directly image_markdown = f"![{chunk.alt}]({chunk.data_uri})" response_chunk = create_response(image_markdown, request.model) else: response_chunk = create_response(chunk, request.model) # Yield each chunk in SSE format yield f"data: {json.dumps(response_chunk)}\n\n" # Signal the end of the stream yield "data: [DONE]\n\n" except HTTPException as he: error_response = {"error": he.detail} yield f"data: {json.dumps(error_response)}\n\n" except Exception as e: logger.exception("Error during streaming response generation.") error_response = {"error": str(e)} yield f"data: {json.dumps(error_response)}\n\n" return StreamingResponse(generate(), media_type="text/event-stream") else: response_content = "" async for chunk in async_generator: if isinstance(chunk, ImageResponse): response_content += f"![{chunk.alt}]({chunk.data_uri})\n" else: response_content += chunk logger.info("Completed non-streaming response generation.") return { "id": f"chatcmpl-{uuid.uuid4()}", "object": "chat.completion", "created": int(datetime.now().timestamp()), "model": request.model, "choices": [ { "message": { "role": "assistant", "content": response_content }, "finish_reason": "stop", "index": 0 } ], "usage": { "prompt_tokens": sum(len(msg['content'].split()) for msg in messages), "completion_tokens": len(response_content.split()), "total_tokens": sum(len(msg['content'].split()) for msg in messages) + len(response_content.split()) }, } except ModelNotWorkingException as e: logger.warning(f"Model not working: {e}") raise HTTPException(status_code=503, detail=str(e)) except HTTPException as he: logger.warning(f"HTTPException: {he.detail}") raise he except Exception as e: logger.exception("An unexpected error occurred while processing the chat completions request.") raise HTTPException(status_code=500, detail=str(e)) @app.get("/niansuhai/v1/models") async def get_models(): logger.info("Fetching available models.") return {"data": [{"id": model} for model in Blackbox.models]} # Additional endpoints for better functionality @app.get("/niansuhai/v1/health") async def health_check(): """Health check endpoint to verify the service is running.""" return {"status": "ok"} @app.get("/niansuhai/v1/models/{model}/status") async def model_status(model: str): """Check if a specific model is available.""" if model in Blackbox.models: return {"model": model, "status": "available"} elif model in Blackbox.model_aliases: actual_model = Blackbox.model_aliases[model] return {"model": actual_model, "status": "available via alias"} else: raise HTTPException(status_code=404, detail="Model not found") if __name__ == "__main__": import uvicorn uvicorn.run(app, host="0.0.0.0", port=8000)