server.py

from fastapi import FastAPI, HTTPException, Request
from fastapi.responses import StreamingResponse, JSONResponse
from pydantic import BaseModel, Field, ValidationError
from typing import List, Optional, Dict, Any, Union
import json
import time
import uuid
import logging
import traceback
from curl_cffi import CurlError
from curl_cffi.requests import Session
import asyncio
from threading import Lock
import os

# Configure logging
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)

class GithubChat:
    def __init__(self, cookie_path="cookies.json", model="gpt-4o"):
        self.api_url = "https://api.individual.githubcopilot.com"
        self.session = Session()
        self.session.headers.update({
            "Content-Type": "application/json",
            "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36",
            "Accept": "*/*",
            "Accept-Encoding": "gzip, deflate, br",
            "Accept-Language": "en-US,en;q=0.5",
            "Origin": "https://github.com",
            "Referer": "https://github.com/copilot",
            "GitHub-Verified-Fetch": "true",
            "X-Requested-With": "XMLHttpRequest",
            "Connection": "keep-alive",
            "Sec-Fetch-Dest": "empty",
            "Sec-Fetch-Mode": "cors",
            "Sec-Fetch-Site": "same-origin",
        })

        # Load cookies with better error handling
        self.cookie_path = cookie_path
        self._load_cookies()

        self.model = model
        self._access_token = None
        self._conversation_id = None
        self._token_lock = Lock()
        self._conversation_lock = Lock()
        self.max_retries = 3
        self.retry_delay = 1.0

    def _load_cookies(self):
        """Load cookies with robust error handling"""
        try:
            if not os.path.exists(self.cookie_path):
                logger.warning(f"Cookie file {self.cookie_path} not found")
                return

            with open(self.cookie_path, 'r', encoding='utf-8') as f:
                cookies_data = json.load(f)

            if not isinstance(cookies_data, list):
                logger.error("Invalid cookie format: expected list")
                return

            cookies = {}
            current_time = time.time()

            for cookie in cookies_data:
                if not isinstance(cookie, dict):
                    continue

                name = cookie.get('name')
                value = cookie.get('value')

                if not name or not value:
                    continue

                # Check expiration
                expiry = cookie.get('expirationDate')
                if expiry and expiry <= current_time:
                    logger.debug(f"Cookie {name} expired")
                    continue

                cookies[name] = value

            self.session.cookies.update(cookies)
            logger.info(f"Loaded {len(cookies)} valid cookies")

        except json.JSONDecodeError as e:
            logger.error(f"Invalid JSON in cookie file: {e}")
        except Exception as e:
            logger.error(f"Error loading cookies: {e}")

    def _retry_request(self, func, *args, **kwargs):
        """Retry mechanism for API requests"""
        last_exception = None

        for attempt in range(self.max_retries):
            try:
                return func(*args, **kwargs)
            except (CurlError, ConnectionError, TimeoutError) as e:
                last_exception = e
                if attempt < self.max_retries - 1:
                    wait_time = self.retry_delay * (2 ** attempt)
                    logger.warning(f"Request failed (attempt {attempt + 1}), retrying in {wait_time}s: {e}")
                    time.sleep(wait_time)
                else:
                    logger.error(f"Request failed after {self.max_retries} attempts: {e}")
            except Exception as e:
                logger.error(f"Unexpected error in request: {e}")
                last_exception = e
                break

        raise last_exception or Exception("Request failed after retries")

    def get_access_token(self):
        """Get access token with thread safety and retry logic"""
        with self._token_lock:
            if self._access_token:
                return self._access_token

            def _get_token():
                response = self.session.post(
                    "https://github.com/github-copilot/chat/token",
                    headers=self.session.headers,
                    timeout=30
                )
                if response.status_code == 200:
                    data = response.json()
                    token = data.get("token")
                    if token:
                        self._access_token = token
                        logger.info("Successfully obtained access token")
                        return token
                    else:
                        raise Exception("No token in response")
                else:
                    raise Exception(f"Token request failed: {response.status_code} - {response.text}")

            try:
                return self._retry_request(_get_token)
            except Exception as e:
                logger.error(f"Failed to get access token: {e}")
                # Reset token on failure
                self._access_token = None
                return None

    def create_conversation(self):
        """Create conversation with thread safety and retry logic"""
        with self._conversation_lock:
            if self._conversation_id:
                return self._conversation_id

            access_token = self.get_access_token()
            if not access_token:
                logger.error("Cannot create conversation: no access token")
                return None

            def _create_conv():
                headers = self.session.headers.copy()
                headers["Authorization"] = f"GitHub-Bearer {access_token}"

                response = self.session.post(
                    f"{self.api_url}/github/chat/threads",
                    headers=headers,
                    impersonate="chrome120",
                    timeout=30
                )

                if response.status_code in [200, 201]:
                    data = response.json()
                    thread_id = data.get("thread_id")
                    if thread_id:
                        self._conversation_id = thread_id
                        logger.info(f"Created conversation: {thread_id}")
                        return thread_id
                    else:
                        raise Exception("No thread_id in response")
                else:
                    raise Exception(f"Conversation creation failed: {response.status_code} - {response.text}")

            try:
                return self._retry_request(_create_conv)
            except Exception as e:
                logger.error(f"Failed to create conversation: {e}")
                # Reset conversation on failure
                self._conversation_id = None
                return None

    def chat(self, prompt, stream=False):
        """Chat with robust error handling and validation"""
        if not prompt or not prompt.strip():
            logger.error("Empty prompt provided")
            return None

        conversation_id = self.create_conversation()
        if not conversation_id:
            logger.error("Failed to create conversation")
            return None

        access_token = self.get_access_token()
        if not access_token:
            logger.error("Failed to get access token")
            return None

        def _send_message():
            headers = self.session.headers.copy()
            headers["Authorization"] = f"GitHub-Bearer {access_token}"

            data = {
                "content": prompt,
                "intent": "conversation",
                "references": [],
                "context": [],
                "currentURL": f"https://github.com/copilot/c/{conversation_id}",
                "streaming": True,  # GitHub Copilot API always uses streaming
                "confirmations": [],
                "customInstructions": [],
                "model": self.model,
                "mode": "immersive"
            }

            response = self.session.post(
                f"{self.api_url}/github/chat/threads/{conversation_id}/messages",
                json=data,
                headers=headers,
                impersonate="chrome120",
                stream=True,
                timeout=60  # Longer timeout for chat
            )

            if response.status_code not in [200, 201]:
                raise Exception(f"Chat request failed: {response.status_code} - {response.text}")

            return response

        try:
            response = self._retry_request(_send_message)

            if stream:
                # Return generator for streaming
                return self._stream_response(response)
            else:
                # Collect all chunks for non-streaming response
                response_text = ""
                try:
                    for chunk in self._stream_response(response):
                        if chunk:
                            response_text += chunk
                except Exception as e:
                    logger.error(f"Error collecting streaming response: {e}")
                    return None
                return response_text

        except Exception as e:
            logger.error(f"Chat request failed: {e}")
            # Reset conversation on failure to force new one next time
            self._conversation_id = None
            return None
    
    def _stream_response(self, response):
        """Helper method to parse streaming response with robust error handling"""
        try:
            for line in response.iter_lines():
                if not line:
                    continue

                try:
                    # Handle different line formats
                    if line.startswith(b'data: '):
                        data_str = line[6:].decode('utf-8')
                    elif line.startswith(b'data:'):
                        data_str = line[5:].decode('utf-8')
                    else:
                        continue

                    # Skip empty data or [DONE]
                    if not data_str.strip() or data_str.strip() == '[DONE]':
                        continue

                    try:
                        data = json.loads(data_str)
                        if isinstance(data, dict) and data.get("type") == "content":
                            body = data.get("body", "")
                            if body and isinstance(body, str):  # Only yield non-empty string content
                                yield body
                    except json.JSONDecodeError as e:
                        logger.debug(f"JSON decode error for line: {data_str[:100]}... Error: {e}")
                        continue

                except UnicodeDecodeError as e:
                    logger.debug(f"Unicode decode error: {e}")
                    continue
                except Exception as e:
                    logger.debug(f"Unexpected error processing line: {e}")
                    continue

        except Exception as e:
            logger.error(f"Error in stream response: {e}")
            raise

# OpenAI Compatible Models
class OpenAIModel(BaseModel):
    id: str
    object: str = "model"
    created: int = int(time.time())
    owned_by: str = "github-copilot"

class FunctionCall(BaseModel):
    name: str
    arguments: str

class ToolCall(BaseModel):
    id: str
    type: str = "function"
    function: FunctionCall

class ChatMessage(BaseModel):
    role: str
    content: Optional[Union[str, List[Dict[str, Any]]]] = None
    name: Optional[str] = None
    function_call: Optional[FunctionCall] = None
    tool_calls: Optional[List[ToolCall]] = None
    tool_call_id: Optional[str] = None

class Function(BaseModel):
    name: str
    description: Optional[str] = None
    parameters: Optional[Dict[str, Any]] = None

class Tool(BaseModel):
    type: str = "function"
    function: Function

class ChatCompletionRequest(BaseModel):
    model: str
    messages: List[ChatMessage]
    max_tokens: Optional[int] = None
    temperature: Optional[float] = 0.7
    top_p: Optional[float] = 1.0
    stream: Optional[bool] = True  # Default to True for streaming
    stop: Optional[Union[str, List[str]]] = None
    tools: Optional[List[Tool]] = None
    tool_choice: Optional[Union[str, Dict[str, Any]]] = None
    functions: Optional[List[Function]] = None
    function_call: Optional[Union[str, Dict[str, Any]]] = None
    presence_penalty: Optional[float] = 0.0
    frequency_penalty: Optional[float] = 0.0
    logit_bias: Optional[Dict[str, float]] = None
    user: Optional[str] = None
    seed: Optional[int] = None
    logprobs: Optional[bool] = None
    top_logprobs: Optional[int] = None
    n: Optional[int] = 1

class ChatCompletionChoice(BaseModel):
    index: int = 0
    message: ChatMessage
    finish_reason: Optional[str] = None
    logprobs: Optional[Dict[str, Any]] = None

class ChatCompletionResponse(BaseModel):
    id: str
    object: str = "chat.completion"
    created: int
    model: str
    choices: List[ChatCompletionChoice]
    usage: Optional[Dict[str, int]] = None
    system_fingerprint: Optional[str] = None

class ChatCompletionStreamChoice(BaseModel):
    index: int = 0
    delta: Dict[str, Any]
    finish_reason: Optional[str] = None
    logprobs: Optional[Dict[str, Any]] = None

class ChatCompletionStreamResponse(BaseModel):
    id: str
    object: str = "chat.completion.chunk"
    created: int
    model: str
    choices: List[ChatCompletionStreamChoice]
    system_fingerprint: Optional[str] = None

def format_prompt(messages: List[Dict[str, Any]], add_special_tokens: bool = False,
                 do_continue: bool = False, include_system: bool = True) -> str:
    """
    Format a series of messages into a single string, optionally adding special tokens.

    Args:
        messages: A list of message dictionaries, each containing 'role' and 'content'.
        add_special_tokens: Whether to add special formatting tokens.
        do_continue: If True, don't add the final "Assistant:" prompt.
        include_system: Whether to include system messages in the formatted output.

    Returns:
        A formatted string containing all messages.
    """
    # Helper function to convert content to string
    def to_string(value) -> str:
        if isinstance(value, str):
            return value
        elif isinstance(value, dict):
            if "text" in value:
                return value.get("text", "")
            return ""
        elif isinstance(value, list):
            # Handle array content (like images + text)
            text_parts = []
            for item in value:
                if isinstance(item, dict):
                    if item.get("type") == "text":
                        text_parts.append(item.get("text", ""))
                    elif item.get("type") == "image_url":
                        text_parts.append("[Image]")
                elif isinstance(item, str):
                    text_parts.append(item)
            return "".join(text_parts)
        return str(value) if value is not None else ""

    # If there's only one message and no special tokens needed, just return its content
    if not add_special_tokens and len(messages) <= 1 and messages:
        return to_string(messages[0].get("content", ""))

    # Filter and process messages
    processed_messages = []
    for message in messages:
        role = message.get("role", "")
        content = message.get("content")

        if include_system or role != "system":
            content_str = to_string(content)
            if content_str.strip():
                processed_messages.append((role, content_str))

    # Format each message as "Role: Content"
    formatted = "\n".join([
        f'{role.capitalize()}: {content}'
        for role, content in processed_messages
    ])

    # Add final prompt for assistant if needed
    if do_continue:
        return formatted

    return f"{formatted}\nAssistant:" if formatted else "Assistant:"

app = FastAPI(
    title="GitHub Copilot OpenAI Compatible API",
    version="1.0.0",
    description="OpenAI-compatible API for GitHub Copilot with full streaming and tool support"
)

# Global variables for health monitoring
chat_client = None
startup_time = time.time()
request_count = 0
error_count = 0

@app.on_event("startup")
async def startup_event():
    """Initialize the chat client on startup"""
    global chat_client
    try:
        logger.info("Initializing GitHub Copilot chat client...")
        chat_client = GithubChat()
        logger.info("Chat client initialized successfully")
    except Exception as e:
        logger.error(f"Failed to initialize chat client: {e}")
        # Don't fail startup, but log the error
        chat_client = None

# Add CORS middleware for web clients
try:
    from fastapi.middleware.cors import CORSMiddleware
    app.add_middleware(
        CORSMiddleware,
        allow_origins=["*"],
        allow_credentials=True,
        allow_methods=["*"],
        allow_headers=["*"],
    )
except ImportError:
    pass  # CORS middleware is optional

# Add comprehensive error handling
@app.exception_handler(ValidationError)
async def validation_exception_handler(request, exc: ValidationError):
    logger.error(f"Validation error: {exc}")
    return JSONResponse(
        status_code=422,
        content={
            "error": {
                "message": "Validation error",
                "type": "invalid_request_error",
                "details": exc.errors()
            }
        }
    )

@app.exception_handler(HTTPException)
async def http_exception_handler(request, exc: HTTPException):
    logger.error(f"HTTP error: {exc.status_code} - {exc.detail}")
    return JSONResponse(
        status_code=exc.status_code,
        content={
            "error": {
                "message": exc.detail,
                "type": "api_error",
                "code": exc.status_code
            }
        }
    )

@app.exception_handler(Exception)
async def global_exception_handler(request, exc):
    logger.error(f"Unexpected error: {exc}\n{traceback.format_exc()}")
    return JSONResponse(
        status_code=500,
        content={
            "error": {
                "message": "Internal server error",
                "type": "server_error",
                "code": 500
            }
        }
    )

@app.get("/")
async def root():
    return {"message": "GitHub Copilot OpenAI Compatible API", "version": "1.0.0", "status": "running"}

@app.get("/health")
async def health_check():
    """Enhanced health check with system status"""
    global chat_client, startup_time, request_count, error_count

    uptime = time.time() - startup_time
    status = "healthy"

    # Check chat client status
    client_status = "unknown"
    if chat_client is None:
        client_status = "not_initialized"
        status = "degraded"
    else:
        try:
            # Quick test of token retrieval
            token = chat_client.get_access_token()
            client_status = "ready" if token else "auth_failed"
            if not token:
                status = "degraded"
        except Exception as e:
            client_status = f"error: {str(e)[:50]}"
            status = "degraded"

    return {
        "status": status,
        "timestamp": int(time.time()),
        "uptime_seconds": int(uptime),
        "client_status": client_status,
        "stats": {
            "total_requests": request_count,
            "total_errors": error_count,
            "error_rate": error_count / max(request_count, 1)
        }
    }

@app.get("/v1/models")
async def list_models():
    models = [
        OpenAIModel(id="gpt-4o"),
        OpenAIModel(id="o3-mini"),
        OpenAIModel(id="o1"),
        OpenAIModel(id="claude-3.5-sonnet"),
        OpenAIModel(id="claude-3.7-sonnet"),
        OpenAIModel(id="claude-3.7-sonnet-thought"),
        OpenAIModel(id="claude-sonnet-4"),
        OpenAIModel(id="gemini-2.0-flash-001"),
        OpenAIModel(id="gemini-2.5-pro"),
        OpenAIModel(id="gpt-4.1"),
        OpenAIModel(id="o4-mini"),
    ]
    return {"object": "list", "data": models}

@app.get("/models")
async def list_models_alt():
    """Alternative endpoint for models"""
    return await list_models()

@app.post("/v1/chat/validate")
async def validate_chat_request(request: ChatCompletionRequest):
    """Validate chat completion request format without processing"""
    try:
        # Validate messages
        if not request.messages:
            return {"valid": False, "error": "Messages cannot be empty"}

        validation_results = []
        for i, msg in enumerate(request.messages):
            msg_validation = {
                "index": i,
                "role": repr(msg.role),
                "role_type": type(msg.role).__name__,
                "content_type": type(msg.content).__name__ if msg.content is not None else "None",
                "valid": True,
                "errors": []
            }

            # Check role
            role = getattr(msg, 'role', None)
            if not role:
                msg_validation["valid"] = False
                msg_validation["errors"].append("Missing role")
            else:
                role_str = str(role).lower().strip()
                valid_roles = ["system", "user", "assistant", "function", "tool"]
                if role_str not in valid_roles:
                    msg_validation["valid"] = False
                    msg_validation["errors"].append(f"Invalid role '{role_str}'. Valid: {valid_roles}")

            validation_results.append(msg_validation)

        all_valid = all(result["valid"] for result in validation_results)

        return {
            "valid": all_valid,
            "model": request.model,
            "message_count": len(request.messages),
            "messages": validation_results
        }

    except Exception as e:
        return {"valid": False, "error": f"Validation error: {str(e)}"}

@app.post("/v1/chat/completions")
async def create_chat_completion(request: ChatCompletionRequest):
    """Enhanced chat completions endpoint with robust error handling"""
    global request_count, error_count, chat_client

    request_count += 1
    request_id = f"req-{uuid.uuid4().hex[:8]}"
    logger.info(f"[{request_id}] Chat completion request: model={request.model}, messages={len(request.messages)}, stream={request.stream}")

    # Debug log the first message for troubleshooting
    if request.messages:
        first_msg = request.messages[0]
        logger.debug(f"[{request_id}] First message - role: {repr(first_msg.role)}, content type: {type(first_msg.content)}")

    # Check if chat client is available
    if chat_client is None:
        error_count += 1
        logger.error(f"[{request_id}] Chat client not initialized")
        raise HTTPException(status_code=503, detail="Service temporarily unavailable - chat client not initialized")

    try:
        # Comprehensive validation
        if not request.messages:
            raise HTTPException(status_code=400, detail="Messages array cannot be empty")

        if len(request.messages) > 100:  # Reasonable limit
            raise HTTPException(status_code=400, detail="Too many messages (max 100)")

        # Validate model
        if not request.model or not isinstance(request.model, str):
            raise HTTPException(status_code=400, detail="Model must be a non-empty string")

        # Extract and validate prompt
        message_dicts = []
        total_content_length = 0

        for i, msg in enumerate(request.messages):
            # More flexible role validation
            role = getattr(msg, 'role', None)
            if not role:
                raise HTTPException(status_code=400, detail=f"Missing role in message {i}")

            # Convert to string and validate
            role_str = str(role).lower().strip()
            valid_roles = ["system", "user", "assistant", "function", "tool"]

            if role_str not in valid_roles:
                raise HTTPException(status_code=400, detail=f"Invalid role '{role_str}' in message {i}. Valid roles: {valid_roles}")

            msg_dict = {"role": role_str}

            if msg.content is not None:
                # Handle different content types
                if isinstance(msg.content, str):
                    content_length = len(msg.content)
                elif isinstance(msg.content, list):
                    content_length = sum(len(str(item)) for item in msg.content)
                else:
                    content_length = len(str(msg.content))

                total_content_length += content_length
                msg_dict["content"] = msg.content

            message_dicts.append(msg_dict)

        # Check total content length (reasonable limit)
        if total_content_length > 100000:  # 100KB limit
            raise HTTPException(status_code=400, detail="Total message content too large")

        prompt = format_prompt(message_dicts)

        if not prompt.strip():
            raise HTTPException(status_code=400, detail="No valid message content found")

        logger.info(f"[{request_id}] Formatted prompt length: {len(prompt)}")

        # Determine streaming mode
        should_stream = request.stream if request.stream is not None else True

        if should_stream:
            logger.info(f"[{request_id}] Starting streaming response")
            return StreamingResponse(
                generate_stream_response(request, prompt, request_id),
                media_type="text/event-stream",
                headers={
                    "Cache-Control": "no-cache",
                    "Connection": "keep-alive",
                    "X-Accel-Buffering": "no",
                    "Access-Control-Allow-Origin": "*",
                    "Access-Control-Allow-Headers": "*"
                }
            )
        else:
            logger.info(f"[{request_id}] Starting non-streaming response")
            # Get non-streaming response
            result = chat_client.chat(prompt, stream=False)

            if result is None:
                logger.error(f"[{request_id}] Chat client returned None")
                raise HTTPException(status_code=503, detail="GitHub Copilot service unavailable")

            # Ensure result is a string
            response_text = result if isinstance(result, str) else str(result)

            if not response_text.strip():
                logger.warning(f"[{request_id}] Empty response from chat client")
                response_text = "I apologize, but I couldn't generate a response. Please try again."

            logger.info(f"[{request_id}] Non-streaming response length: {len(response_text)}")

            response = ChatCompletionResponse(
                id=f"chatcmpl-{uuid.uuid4().hex}",
                created=int(time.time()),
                model=request.model,
                choices=[ChatCompletionChoice(
                    message=ChatMessage(role="assistant", content=response_text),
                    finish_reason="stop"
                )],
                usage={
                    "prompt_tokens": len(prompt.split()),
                    "completion_tokens": len(response_text.split()),
                    "total_tokens": len(prompt.split()) + len(response_text.split())
                }
            )
            return response

    except HTTPException:
        error_count += 1
        raise
    except ValidationError as e:
        error_count += 1
        logger.error(f"[{request_id}] Validation error: {e}")
        raise HTTPException(status_code=422, detail=f"Request validation failed: {str(e)}")
    except Exception as e:
        error_count += 1
        logger.error(f"[{request_id}] Unexpected error: {e}\n{traceback.format_exc()}")
        raise HTTPException(status_code=500, detail="Internal server error occurred")

async def generate_stream_response(request: ChatCompletionRequest, prompt: str, request_id: str = None):
    """Enhanced streaming response with comprehensive error handling"""
    completion_id = f"chatcmpl-{uuid.uuid4().hex}"
    created_time = int(time.time())
    request_id = request_id or f"req-{uuid.uuid4().hex[:8]}"

    logger.info(f"[{request_id}] Starting stream generation")

    try:
        # Send initial chunk with role
        initial_chunk = {
            "id": completion_id,
            "object": "chat.completion.chunk",
            "created": created_time,
            "model": request.model,
            "choices": [{
                "index": 0,
                "delta": {"role": "assistant"},
                "finish_reason": None
            }]
        }
        yield f"data: {json.dumps(initial_chunk)}\n\n"

        # Stream content chunks with enhanced error handling
        chunk_count = 0
        total_content = ""

        try:
            chat_stream = chat_client.chat(prompt, stream=True)
            if chat_stream is None:
                raise Exception("Chat client returned None for streaming")

            for chunk in chat_stream:
                if chunk and isinstance(chunk, str):  # Only process non-empty string chunks
                    chunk_count += 1
                    total_content += chunk

                    stream_response = ChatCompletionStreamResponse(
                        id=completion_id,
                        created=created_time,
                        model=request.model,
                        choices=[ChatCompletionStreamChoice(
                            delta={"content": chunk},
                            finish_reason=None
                        )]
                    )

                    try:
                        chunk_json = json.dumps(stream_response.model_dump())
                        yield f"data: {chunk_json}\n\n"
                    except Exception as json_error:
                        logger.error(f"[{request_id}] JSON serialization error: {json_error}")
                        continue

        except Exception as stream_error:
            logger.error(f"[{request_id}] Streaming error after {chunk_count} chunks: {stream_error}")

            # If we got some content, continue gracefully
            if chunk_count > 0:
                logger.info(f"[{request_id}] Partial stream completed with {chunk_count} chunks")
            else:
                # Send error content if no chunks were received
                error_content = "I apologize, but I encountered an error while generating the response. Please try again."
                error_response = ChatCompletionStreamResponse(
                    id=completion_id,
                    created=created_time,
                    model=request.model,
                    choices=[ChatCompletionStreamChoice(
                        delta={"content": error_content},
                        finish_reason=None
                    )]
                )
                yield f"data: {json.dumps(error_response.model_dump())}\n\n"

        logger.info(f"[{request_id}] Stream completed: {chunk_count} chunks, {len(total_content)} characters")

    except Exception as e:
        logger.error(f"[{request_id}] Critical streaming error: {e}")
        # Send error chunk
        error_chunk = {
            "id": completion_id,
            "object": "chat.completion.chunk",
            "created": created_time,
            "model": request.model,
            "choices": [{
                "index": 0,
                "delta": {"content": "Error occurred while streaming response."},
                "finish_reason": "stop"
            }]
        }
        yield f"data: {json.dumps(error_chunk)}\n\n"

    finally:
        # Always send final chunk
        try:
            final_chunk = {
                "id": completion_id,
                "object": "chat.completion.chunk",
                "created": created_time,
                "model": request.model,
                "choices": [{
                    "index": 0,
                    "delta": {},
                    "finish_reason": "stop"
                }]
            }
            yield f"data: {json.dumps(final_chunk)}\n\n"
            yield "data: [DONE]\n\n"
            logger.info(f"[{request_id}] Stream finalized")
        except Exception as final_error:
            logger.error(f"[{request_id}] Error sending final chunk: {final_error}")
            yield "data: [DONE]\n\n"

if __name__ == "__main__":
    try:
        import uvicorn
        port = int(os.getenv("PORT", 8000))
        host = os.getenv("HOST", "0.0.0.0")

        logger.info(f"Starting server on {host}:{port}")
        uvicorn.run(
            "server:app",
            host=host,
            port=port,
            reload=False,  # Disable reload in production
            log_level="info",
            access_log=True
        )
    except ImportError:
        logger.error("uvicorn not installed. Install with: pip install uvicorn")
    except Exception as e:
        logger.error(f"Failed to start server: {e}")