api/utils.py

from datetime import datetime
import json
import uuid
import asyncio
import random
import os
from typing import Any, Dict, Optional

import httpx
from fastapi import HTTPException
from api.config import (
    MODEL_MAPPING,
    get_headers_api_chat,
    BASE_URL,
    AGENT_MODE,
    TRENDING_AGENT_MODE,
    MODEL_PREFIXES,
)
from api.models import ChatRequest
from api.logger import setup_logger
from api.validate import getHid

logger = setup_logger(__name__)

BLOCKED_MESSAGE = "Generated by BLACKBOX.AI, try unlimited chat https://www.blackbox.ai"
ADVERTISEMENT_TEXT = os.getenv("ADVERTISEMENT_TEXT", "")

def create_chat_completion_data(content: str, model: str, timestamp: int, finish_reason: Optional[str] = None) -> Dict[str, Any]:
    return {
        "id": f"chatcmpl-{uuid.uuid4()}",
        "object": "chat.completion.chunk",
        "created": timestamp,
        "model": model,
        "choices": [
            {
                "index": 0,
                "delta": {"content": content, "role": "assistant"},
                "finish_reason": finish_reason,
            }
        ],
        "usage": None,
    }

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}"
    return {"role": message.role, "content": content}

def strip_model_prefix(content: str, model_prefix: Optional[str] = None) -> str:
    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 with headers from config.py
async def process_streaming_response(request: ChatRequest):
    request_id = f"chatcmpl-{uuid.uuid4()}"
    logger.info(f"Processing request with ID: {request_id} - Model: {request.model}")

    agent_mode = AGENT_MODE.get(request.model, {})
    trending_agent_mode = TRENDING_AGENT_MODE.get(request.model, {})
    model_prefix = MODEL_PREFIXES.get(request.model, "")

    headers_api_chat = get_headers_api_chat(BASE_URL)

    # Delay for 'o1-preview' model if necessary
    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' (Request ID: {request_id})")
        await asyncio.sleep(delay_seconds)

    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.")

    json_data = {
        "agentMode": agent_mode,
        "clickedAnswer2": False,
        "clickedAnswer3": False,
        "clickedForceWebSearch": False,
        "codeModelMode": True,
        "githubToken": None,
        "id": None,
        "isChromeExt": False,
        "isMicMode": False,
        "maxTokens": request.max_tokens,
        "messages": [message_to_dict(msg, model_prefix=model_prefix) for msg in request.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,
        "visitFromDelta": False,
        "webSearchModePrompt": False,
        "imageGenerationMode": False,
    }

    advertisement_added = False  # Track if advertisement is already added

    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()

                timestamp = int(datetime.now().timestamp())
                async for chunk in response.aiter_text():
                    if chunk:
                        content = chunk
                        if content.startswith("$@$v=undefined-rv1$@$"):
                            content = content[21:]  # Remove unwanted prefix

                        if BLOCKED_MESSAGE in content:
                            logger.info(f"Blocked message detected in response for Request ID {request_id}.")
                            content = content.replace(BLOCKED_MESSAGE, '').strip()

                        if not content:
                            continue  # Skip empty content

                        cleaned_content = strip_model_prefix(content, model_prefix)

                        # Yield cleaned chunk as part of the stream
                        yield f"data: {json.dumps(create_chat_completion_data(cleaned_content, request.model, timestamp))}\n\n"

                # Append markdown separator and advertisement as a separate chunk
                if ADVERTISEMENT_TEXT and not advertisement_added:
                    advertisement_with_line_break = "\n\n---\n" + ADVERTISEMENT_TEXT  # Add space before the markdown horizontal rule
                    yield f"data: {json.dumps(create_chat_completion_data(advertisement_with_line_break, request.model, timestamp))}\n\n"
                    advertisement_added = True

                # Yield final chunk indicating the end of the stream
                yield f"data: {json.dumps(create_chat_completion_data('', request.model, timestamp, '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}")
            raise HTTPException(status_code=e.response.status_code, detail=str(e))
        except httpx.RequestError as e:
            logger.error(f"Request error occurred for Request ID {request_id}: {e}")
            raise HTTPException(status_code=500, detail=str(e))

# Process non-streaming response with headers from config.py
async def process_non_streaming_response(request: ChatRequest):
    request_id = f"chatcmpl-{uuid.uuid4()}"
    logger.info(f"Processing request with ID: {request_id} - Model: {request.model}")

    agent_mode = AGENT_MODE.get(request.model, {})
    trending_agent_mode = TRENDING_AGENT_MODE.get(request.model, {})
    model_prefix = MODEL_PREFIXES.get(request.model, "")

    headers_api_chat = get_headers_api_chat(BASE_URL)

    # Delay for 'o1-preview' model if necessary
    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' (Request ID: {request_id})")
        await asyncio.sleep(delay_seconds)

    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.")

    json_data = {
        "agentMode": agent_mode,
        "clickedAnswer2": False,
        "clickedAnswer3": False,
        "clickedForceWebSearch": False,
        "codeModelMode": True,
        "githubToken": None,
        "id": None,
        "isChromeExt": False,
        "isMicMode": False,
        "maxTokens": request.max_tokens,
        "messages": [message_to_dict(msg, model_prefix=model_prefix) for msg in request.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,
        "visitFromDelta": False,
        "webSearchModePrompt": False,
        "imageGenerationMode": False,
    }

    full_response = ""  # Store full non-streaming response content

    async with httpx.AsyncClient() as client:
        try:
            async with client.stream("POST", 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}")
            raise HTTPException(status_code=e.response.status_code, detail=str(e))
        except httpx.RequestError as e:
            logger.error(f"Request error occurred for Request ID {request_id}: {e}")
            raise HTTPException(status_code=500, detail=str(e))

    if full_response.startswith("$@$v=undefined-rv1$@$"):
        full_response = full_response[21:]

    if BLOCKED_MESSAGE in full_response:
        full_response = full_response.replace(BLOCKED_MESSAGE, '').strip()

    cleaned_full_response = strip_model_prefix(full_response, model_prefix)

    # Append advertisement once
    if ADVERTISEMENT_TEXT:
        cleaned_full_response += "\n\n" + ADVERTISEMENT_TEXT

    return {
        "id": f"chatcmpl-{uuid.uuid4()}",
        "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": None,
    }