api/utils.py

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

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 (
    log_generated_chat_id_with_referer,
    log_model_delay,
    log_http_error,
    log_request_error,
    log_strip_prefix,
)

# Generate a random alphanumeric chat ID
def generate_chat_id(length: int = 7) -> str:
    characters = string.ascii_letters + string.digits
    return ''.join(random.choices(characters, k=length))

# Function to create chat completion data
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,
    }

# Convert message to dictionary format with optional model prefix
def message_to_dict(message, model_prefix: Optional[str] = None):
    content = message.content
    if model_prefix:
        content = f"{model_prefix} {content}"
    return {"role": message.role, "content": content}

# Remove model prefix from content if present
def strip_model_prefix(content: str, model_prefix: Optional[str] = None) -> str:
    if model_prefix and content.startswith(model_prefix):
        log_strip_prefix(model_prefix, content)
        return content[len(model_prefix):].strip()
    return content

# Streaming response with headers from config.py
async def process_streaming_response(request: ChatRequest):
    # Generate chat_id only if model is in MODEL_REFERERS
    chat_id = generate_chat_id() if request.model in MODEL_REFERERS else None
    referer_path = MODEL_REFERERS.get(request.model, "")
    referer_url = f"{BASE_URL}/chat/{chat_id}?model={request.model}" if chat_id else BASE_URL

    # Log chat ID, model, and referer URL if applicable
    log_generated_chat_id_with_referer(chat_id, request.model, referer_url)

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

    # Dynamic headers based on referer
    headers_api_chat = get_headers_api_chat(referer_url)

    # Delay for 'o1-preview' model
    if request.model == 'o1-preview':
        delay_seconds = random.randint(20, 60)
        log_model_delay(delay_seconds, request.model, chat_id)
        await asyncio.sleep(delay_seconds)

    json_data = {
        "agentMode": agent_mode,
        "clickedAnswer2": False,
        "clickedAnswer3": False,
        "clickedForceWebSearch": False,
        "codeModelMode": True,
        "githubToken": None,
        "id": chat_id,
        "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": "69783381-2ce4-4dbd-ac78-35e9063feabc",
        "visitFromDelta": False,
    }

    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 line in response.aiter_lines():
                    timestamp = int(datetime.now().timestamp())
                    if line:
                        content = line.lstrip("$@$v=undefined-rv1$@$")
                        cleaned_content = strip_model_prefix(content, model_prefix)
                        yield f"data: {json.dumps(create_chat_completion_data(cleaned_content, request.model, timestamp))}\n\n"

                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:
            log_http_error(e, chat_id)
            raise HTTPException(status_code=e.response.status_code, detail=str(e))
        except httpx.RequestError as e:
            log_request_error(e, chat_id)
            raise HTTPException(status_code=500, detail=str(e))

# Non-streaming response with headers from config.py
async def process_non_streaming_response(request: ChatRequest):
    chat_id = generate_chat_id() if request.model in MODEL_REFERERS else None
    referer_path = MODEL_REFERERS.get(request.model, "")
    referer_url = f"{BASE_URL}/chat/{chat_id}?model={request.model}" if chat_id else BASE_URL

    # Log with chat ID, model, and referer URL if applicable
    log_generated_chat_id_with_referer(chat_id, request.model, referer_url)

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

    # Headers for API chat and secondary request
    headers_api_chat = get_headers_api_chat(referer_url)
    headers_chat = get_headers_chat(referer_url, next_action=str(uuid.uuid4()), next_router_state_tree=json.dumps([""]))

    # Delay for 'o1-preview' model
    if request.model == 'o1-preview':
        delay_seconds = random.randint(20, 60)
        log_model_delay(delay_seconds, request.model, chat_id)
        await asyncio.sleep(delay_seconds)

    json_data = {
        "agentMode": agent_mode,
        "clickedAnswer2": False,
        "clickedAnswer3": False,
        "clickedForceWebSearch": False,
        "codeModelMode": True,
        "githubToken": None,
        "id": chat_id,
        "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": "69783381-2ce4-4dbd-ac78-35e9063feabc",
        "visitFromDelta": False,
    }

    full_response = ""
    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:
            log_http_error(e, chat_id)
            raise HTTPException(status_code=e.response.status_code, detail=str(e))
        except httpx.RequestError as e:
            log_request_error(e, chat_id)
            raise HTTPException(status_code=500, detail=str(e))
    
    if full_response.startswith("$@$v=undefined-rv1$@$"):
        full_response = full_response[21:]

    # Strip the model prefix
    cleaned_full_response = strip_model_prefix(full_response, model_prefix)

    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,
    }