from datetime import datetime import json import uuid import asyncio import random import string from typing import Any, Dict, Optional, List import httpx from fastapi import HTTPException from api.config import ( MODEL_MAPPING, get_headers_api_chat, get_headers_chat, BASE_URL, AGENT_MODE, TRENDING_AGENT_MODE, MODEL_PREFIXES, MODEL_REFERERS ) from api.models import ChatRequest from api.logger import setup_logger from api.validate import getHid # Import the asynchronous getHid function import tiktoken # ---------------- NEW IMPORTS FOR CLOUDFLARE R2 UPLOAD ---------------- import boto3 import re logger = setup_logger(__name__) # Define the blocked message BLOCKED_MESSAGE = "Generated by BLACKBOX.AI, try unlimited chat https://www.blackbox.ai and for API requests replace https://www.blackbox.ai with https://api.blackbox.ai" # ---------------- R2 CONFIG EXAMPLE ---------------- R2_ACCESS_KEY_ID = "df9c9eb87e850a8eb27afd3968077b42" R2_SECRET_ACCESS_KEY = "14b08b0855263bb63d2618da3a6537e1b0446d89d51da03a568620b1e5342ea8" R2_ENDPOINT_URL = "https://f2f92ac53fae792c4155f6e93a514989.r2.cloudflarestorage.com" R2_BUCKET_NAME = "snapzion" R2_REPLACED_URLS_KEY_PREFIX = "nai" # or however you want to name your uploads # Initialize your R2 client s3 = boto3.client( "s3", endpoint_url=R2_ENDPOINT_URL, aws_access_key_id=R2_ACCESS_KEY_ID, aws_secret_access_key=R2_SECRET_ACCESS_KEY, ) # Function to upload replaced URLs into R2 as a text file def upload_replaced_urls_to_r2(urls: List[str], request_id: str) -> None: """ Given a list of replaced URLs, upload them as a .txt file to your R2 bucket. The file name includes the request_id so it's unique for each request. """ if not urls: logger.info(f"No replaced URLs found for request {request_id}. Skipping upload.") return # Join all replaced URLs with a newline content_body = "\n".join(urls) # Create an object key, for example: replaced-urls/chatcmpl-.txt object_key = f"{R2_REPLACED_URLS_KEY_PREFIX}/{request_id}.txt" try: s3.put_object( Bucket=R2_BUCKET_NAME, Key=object_key, Body=content_body.encode("utf-8"), ) logger.info(f"Uploaded replaced URLs to R2: {object_key}") except Exception as e: logger.error(f"Failed to upload replaced URLs to R2: {e}") # Function to calculate tokens using tiktoken def calculate_tokens(text: str, model: str) -> int: try: encoding = tiktoken.encoding_for_model(model) tokens = encoding.encode(text) return len(tokens) except KeyError: # Handle the case where the model is not supported by tiktoken logger.warning(f"Model '{model}' not supported by tiktoken for token counting. Using a generic method.") return len(text.split()) # Helper function to create chat completion data def create_chat_completion_data( content: str, model: str, timestamp: int, request_id: str, prompt_tokens: int = 0, completion_tokens: int = 0, finish_reason: Optional[str] = None ) -> Dict[str, Any]: if finish_reason == "stop": usage = { "prompt_tokens": prompt_tokens, "completion_tokens": completion_tokens, "total_tokens": prompt_tokens + completion_tokens, } else: usage = None return { "id": request_id, "object": "chat.completion.chunk", "created": timestamp, "model": model, "choices": [ { "index": 0, "delta": {"content": content, "role": "assistant"}, "finish_reason": finish_reason, } ], "usage": usage, } # Function to convert message to dictionary format, ensuring base64 data and optional model prefix def message_to_dict(message, model_prefix: Optional[str] = None): content = message.content if isinstance(message.content, str) else message.content[0]["text"] if model_prefix: content = f"{model_prefix} {content}" if isinstance(message.content, list) and len(message.content) == 2 and "image_url" in message.content[1]: # Ensure base64 images are always included for all models image_base64 = message.content[1]["image_url"]["url"] return { "role": message.role, "content": content, "data": { "imageBase64": image_base64, "fileText": "", "title": "snapshot", # Added imagesData field here "imagesData": [ { "filePath": f"MultipleFiles/{uuid.uuid4().hex}.jpg", "contents": image_base64 } ], }, } return {"role": message.role, "content": content} # Function to strip model prefix from content if present def strip_model_prefix(content: str, model_prefix: Optional[str] = None) -> str: """Remove the model prefix from the response content if present.""" if model_prefix and content.startswith(model_prefix): logger.debug(f"Stripping prefix '{model_prefix}' from content.") return content[len(model_prefix):].strip() return content # --------------------- PROCESS STREAMING RESPONSE --------------------- async def process_streaming_response(request: ChatRequest): # Generate a unique ID for this request request_id = f"chatcmpl-{uuid.uuid4()}" logger.info(f"Processing request with ID: {request_id} - Model: {request.model}") # Get the appropriate configuration for the requested model agent_mode = AGENT_MODE.get(request.model, {}) trending_agent_mode = TRENDING_AGENT_MODE.get(request.model, {}) model_prefix = MODEL_PREFIXES.get(request.model, "") # Adjust headers_api_chat since referer_url is removed headers_api_chat = get_headers_api_chat(BASE_URL) if request.model == 'o1-preview': delay_seconds = random.randint(1, 60) logger.info( f"Introducing a delay of {delay_seconds} seconds for model 'o1-preview' " f"(Request ID: {request_id})" ) await asyncio.sleep(delay_seconds) # Fetch the h-value for the 'validated' field h_value = await getHid() if not h_value: logger.error("Failed to retrieve h-value for validation.") raise HTTPException( status_code=500, detail="Validation failed due to missing h-value." ) messages = [ message_to_dict(msg, model_prefix=model_prefix) for msg in request.messages ] json_data = { "agentMode": agent_mode, "clickedAnswer2": False, "clickedAnswer3": False, "clickedForceWebSearch": False, "codeModelMode": True, "githubToken": None, "id": request_id, "isChromeExt": False, "isMicMode": False, "maxTokens": request.max_tokens, "messages": messages, "mobileClient": False, "playgroundTemperature": request.temperature, "playgroundTopP": request.top_p, "previewToken": None, "trendingAgentMode": trending_agent_mode, "userId": None, "userSelectedModel": MODEL_MAPPING.get(request.model, request.model), "userSystemPrompt": None, "validated": h_value, # Dynamically set the validated field "visitFromDelta": False, "webSearchModePrompt": False, "imageGenerationMode": False, # Added this line } prompt_tokens = 0 for message in messages: if 'content' in message: prompt_tokens += calculate_tokens(message['content'], request.model) if 'data' in message and 'imageBase64' in message['data']: prompt_tokens += calculate_tokens(message['data']['imageBase64'], request.model) completion_tokens = 0 # We'll keep track of any replaced URLs in this list replaced_urls = [] async with httpx.AsyncClient() as client: try: async with client.stream( "POST", f"{BASE_URL}/api/chat", headers=headers_api_chat, json=json_data, timeout=100, ) as response: response.raise_for_status() async for chunk in response.aiter_text(): timestamp = int(datetime.now().timestamp()) if chunk: # First handle any prefix from the server, if present if chunk.startswith("$@$v=undefined-rv1$@$"): chunk = chunk[21:] # Remove the blocked message if present if BLOCKED_MESSAGE in chunk: logger.info( f"Blocked message detected in response for Request ID {request_id}." ) chunk = chunk.replace(BLOCKED_MESSAGE, '').strip() if not chunk: continue # Skip if content is empty after removal # ---------------- REPLACE STORAGE URLS ---------------- # We'll search for "https://storage.googleapis.com" and replace # with "https://cdn.snapzion.com" # Also collect them for uploading to R2 found_urls = re.findall(r"https://storage\.googleapis\.com\S*", chunk) if found_urls: replaced_urls.extend(found_urls) chunk = chunk.replace("https://storage.googleapis.com", "https://cdn.snapzion.com") cleaned_content = strip_model_prefix(chunk, model_prefix) completion_tokens += calculate_tokens(cleaned_content, request.model) yield f"data: {json.dumps(create_chat_completion_data(cleaned_content, request.model, timestamp, request_id))}\n\n" # At the end of the stream, finalize the response # yield final piece with finish_reason="stop" yield f"data: {json.dumps(create_chat_completion_data('', request.model, timestamp, request_id, prompt_tokens, completion_tokens, 'stop'))}\n\n" yield "data: [DONE]\n\n" except httpx.HTTPStatusError as e: logger.error(f"HTTP error occurred for Request ID {request_id}: {e}") error_message = f"HTTP error occurred: {e}" try: error_details = e.response.json() error_message += f" Details: {error_details}" except ValueError: error_message += f" Response body: {e.response.text}" yield f"data: {json.dumps(create_chat_completion_data(error_message, request.model, timestamp, request_id, prompt_tokens, completion_tokens, 'error'))}\n\n" yield "data: [DONE]\n\n" except httpx.RequestError as e: logger.error( f"Error occurred during request for Request ID {request_id}: {e}" ) error_message = f"Request error occurred: {e}" yield f"data: {json.dumps(create_chat_completion_data(error_message, request.model, timestamp, request_id, prompt_tokens, completion_tokens, 'error'))}\n\n" yield "data: [DONE]\n\n" except Exception as e: logger.error(f"An unexpected error occurred for Request ID {request_id}: {e}") error_message = f"An unexpected error occurred: {e}" yield f"data: {json.dumps(create_chat_completion_data(error_message, request.model, timestamp, request_id, prompt_tokens, completion_tokens, 'error'))}\n\n" yield "data: [DONE]\n\n" # Once the entire stream is done, upload any replaced URLs to R2 upload_replaced_urls_to_r2(replaced_urls, request_id) # --------------------- PROCESS NON-STREAMING RESPONSE --------------------- async def process_non_streaming_response(request: ChatRequest): # Generate a unique ID for this request request_id = f"chatcmpl-{uuid.uuid4()}" logger.info(f"Processing request with ID: {request_id} - Model: {request.model}") # Get the appropriate configuration for the requested model agent_mode = AGENT_MODE.get(request.model, {}) trending_agent_mode = TRENDING_AGENT_MODE.get(request.model, {}) model_prefix = MODEL_PREFIXES.get(request.model, "") # Adjust headers_api_chat and headers_chat since referer_url is removed headers_api_chat = get_headers_api_chat(BASE_URL) headers_chat = get_headers_chat( BASE_URL, next_action=str(uuid.uuid4()), next_router_state_tree=json.dumps([""]), ) if request.model == 'o1-preview': delay_seconds = random.randint(20, 60) logger.info( f"Introducing a delay of {delay_seconds} seconds for model 'o1-preview' " f"(Request ID: {request_id})" ) await asyncio.sleep(delay_seconds) # Fetch the h-value for the 'validated' field # Hard-coded for demonstration h_value = "00f37b34-a166-4efb-bce5-1312d87f2f94" if not h_value: logger.error("Failed to retrieve h-value for validation.") raise HTTPException( status_code=500, detail="Validation failed due to missing h-value." ) messages = [ message_to_dict(msg, model_prefix=model_prefix) for msg in request.messages ] json_data = { "agentMode": agent_mode, "clickedAnswer2": False, "clickedAnswer3": False, "clickedForceWebSearch": False, "codeModelMode": True, "githubToken": None, "id": request_id, "isChromeExt": False, "isMicMode": False, "maxTokens": request.max_tokens, "messages": messages, "mobileClient": False, "playgroundTemperature": request.temperature, "playgroundTopP": request.top_p, "previewToken": None, "trendingAgentMode": trending_agent_mode, "userId": None, "userSelectedModel": MODEL_MAPPING.get(request.model, request.model), "userSystemPrompt": None, "validated": h_value, # Dynamically set the validated field "visitFromDelta": False, "webSearchModePrompt": False, "imageGenerationMode": False, # Added this line } prompt_tokens = 0 for message in messages: if 'content' in message: prompt_tokens += calculate_tokens(message['content'], request.model) if 'data' in message and 'imageBase64' in message['data']: prompt_tokens += calculate_tokens(message['data']['imageBase64'], request.model) full_response = "" # We'll keep track of any replaced URLs here replaced_urls = [] async with httpx.AsyncClient() as client: try: async with client.stream( method="POST", url=f"{BASE_URL}/api/chat", headers=headers_api_chat, json=json_data, ) as response: response.raise_for_status() async for chunk in response.aiter_text(): full_response += chunk except httpx.HTTPStatusError as e: logger.error(f"HTTP error occurred for Request ID {request_id}: {e}") error_message = f"HTTP error occurred: {e}" try: error_details = e.response.json() error_message += f" Details: {error_details}" except ValueError: error_message += f" Response body: {e.response.text}" return { "id": request_id, "object": "chat.completion", "created": int(datetime.now().timestamp()), "model": request.model, "choices": [ { "index": 0, "message": {"role": "assistant", "content": error_message}, "finish_reason": "error", } ], "usage": { "prompt_tokens": prompt_tokens, "completion_tokens": 0, "total_tokens": prompt_tokens, }, } except httpx.RequestError as e: logger.error( f"Error occurred during request for Request ID {request_id}: {e}" ) error_message = f"Request error occurred: {e}" return { "id": request_id, "object": "chat.completion", "created": int(datetime.now().timestamp()), "model": request.model, "choices": [ { "index": 0, "message": {"role": "assistant", "content": error_message}, "finish_reason": "error", } ], "usage": { "prompt_tokens": prompt_tokens, "completion_tokens": 0, "total_tokens": prompt_tokens, }, } except Exception as e: logger.error(f"An unexpected error occurred for Request ID {request_id}: {e}") error_message = f"An unexpected error occurred: {e}" return { "id": request_id, "object": "chat.completion", "created": int(datetime.now().timestamp()), "model": request.model, "choices": [ { "index": 0, "message": {"role": "assistant", "content": error_message}, "finish_reason": "error", } ], "usage": { "prompt_tokens": prompt_tokens, "completion_tokens": 0, "total_tokens": prompt_tokens, }, } # Clean up server prefix if needed if full_response.startswith("$@$v=undefined-rv1$@$"): full_response = full_response[21:] # Remove blocked message if BLOCKED_MESSAGE in full_response: logger.info( f"Blocked message detected in response for Request ID {request_id}." ) full_response = full_response.replace(BLOCKED_MESSAGE, '').strip() if not full_response: raise HTTPException( status_code=500, detail="Blocked message detected in response." ) # ---------------- REPLACE STORAGE URLS ---------------- # We'll search for "https://storage.googleapis.com" and replace # with "https://cdn.snapzion.com" found_urls = re.findall(r"https://storage\.googleapis\.com\S*", full_response) if found_urls: replaced_urls.extend(found_urls) full_response = full_response.replace("https://storage.googleapis.com", "https://cdn.snapzion.com") # Strip model prefix if present cleaned_full_response = strip_model_prefix(full_response, model_prefix) completion_tokens = calculate_tokens(cleaned_full_response, request.model) # Upload replaced URLs to R2 upload_replaced_urls_to_r2(replaced_urls, request_id) return { "id": request_id, "object": "chat.completion", "created": int(datetime.now().timestamp()), "model": request.model, "choices": [ { "index": 0, "message": {"role": "assistant", "content": cleaned_full_response}, "finish_reason": "stop", } ], "usage": { "prompt_tokens": prompt_tokens, "completion_tokens": completion_tokens, "total_tokens": prompt_tokens + completion_tokens, }, }