from fastapi import FastAPI, Depends, HTTPException, status, UploadFile, File
from fastapi.middleware.cors import CORSMiddleware
from fastapi.security import HTTPBearer, HTTPAuthorizationCredentials
from sqlalchemy.orm import Session, joinedload
from typing import List
import os
import shutil
from datetime import datetime, timedelta

from app.core.database import engine, get_db, Base
from app.core.security import verify_password, get_password_hash, create_access_token, decode_access_token
from app import models, schemas

# Crear tablas
Base.metadata.create_all(bind=engine)

app = FastAPI(title="Checklist Inteligente API", version="1.0.0")

# CORS
app.add_middleware(
    CORSMiddleware,
    allow_origins=["http://localhost:5173", "http://localhost:3000"],
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

security = HTTPBearer()

# Dependency para obtener usuario actual
def get_current_user(
    credentials: HTTPAuthorizationCredentials = Depends(security),
    db: Session = Depends(get_db)
):
    token = credentials.credentials
    payload = decode_access_token(token)
    print(f"Token payload: {payload}")  # Debug
    if payload is None:
        raise HTTPException(
            status_code=status.HTTP_401_UNAUTHORIZED,
            detail="Token inválido o expirado"
        )
    
    user_id = int(payload.get("sub"))
    print(f"Looking for user ID: {user_id}")  # Debug
    user = db.query(models.User).filter(models.User.id == user_id).first()
    if user is None:
        print(f"User not found with ID: {user_id}")  # Debug
        raise HTTPException(status_code=404, detail="Usuario no encontrado")
    
    return user


# ============= AUTH ENDPOINTS =============
@app.post("/api/auth/register", response_model=schemas.User)
def register(user: schemas.UserCreate, db: Session = Depends(get_db)):
    # Verificar si usuario existe
    db_user = db.query(models.User).filter(models.User.username == user.username).first()
    if db_user:
        raise HTTPException(status_code=400, detail="Usuario ya existe")
    
    # Crear usuario
    hashed_password = get_password_hash(user.password)
    db_user = models.User(
        username=user.username,
        email=user.email,
        full_name=user.full_name,
        role=user.role,
        password_hash=hashed_password
    )
    db.add(db_user)
    db.commit()
    db.refresh(db_user)
    return db_user


@app.post("/api/auth/login", response_model=schemas.Token)
def login(user_login: schemas.UserLogin, db: Session = Depends(get_db)):
    user = db.query(models.User).filter(models.User.username == user_login.username).first()
    if not user or not verify_password(user_login.password, user.password_hash):
        raise HTTPException(
            status_code=status.HTTP_401_UNAUTHORIZED,
            detail="Usuario o contraseña incorrectos"
        )
    
    access_token = create_access_token(data={"sub": str(user.id), "role": user.role})
    return {
        "access_token": access_token,
        "token_type": "bearer",
        "user": user
    }


@app.get("/api/auth/me", response_model=schemas.User)
def get_me(current_user: models.User = Depends(get_current_user)):
    return current_user


# ============= CHECKLIST ENDPOINTS =============
@app.get("/api/checklists", response_model=List[schemas.Checklist])
def get_checklists(
    skip: int = 0,
    limit: int = 100,
    active_only: bool = False,
    db: Session = Depends(get_db),
    current_user: models.User = Depends(get_current_user)
):
    query = db.query(models.Checklist)
    if active_only:
        query = query.filter(models.Checklist.is_active == True)
    return query.offset(skip).limit(limit).all()


@app.get("/api/checklists/{checklist_id}", response_model=schemas.ChecklistWithQuestions)
def get_checklist(checklist_id: int, db: Session = Depends(get_db)):
    checklist = db.query(models.Checklist).options(
        joinedload(models.Checklist.questions)
    ).filter(models.Checklist.id == checklist_id).first()
    
    if not checklist:
        raise HTTPException(status_code=404, detail="Checklist no encontrado")
    
    return checklist


@app.post("/api/checklists", response_model=schemas.Checklist)
def create_checklist(
    checklist: schemas.ChecklistCreate,
    db: Session = Depends(get_db),
    current_user: models.User = Depends(get_current_user)
):
    if current_user.role != "admin":
        raise HTTPException(status_code=403, detail="No autorizado")
    
    db_checklist = models.Checklist(**checklist.dict(), created_by=current_user.id)
    db.add(db_checklist)
    db.commit()
    db.refresh(db_checklist)
    return db_checklist


@app.put("/api/checklists/{checklist_id}", response_model=schemas.Checklist)
def update_checklist(
    checklist_id: int,
    checklist: schemas.ChecklistUpdate,
    db: Session = Depends(get_db),
    current_user: models.User = Depends(get_current_user)
):
    if current_user.role != "admin":
        raise HTTPException(status_code=403, detail="No autorizado")
    
    db_checklist = db.query(models.Checklist).filter(models.Checklist.id == checklist_id).first()
    if not db_checklist:
        raise HTTPException(status_code=404, detail="Checklist no encontrado")
    
    for key, value in checklist.dict(exclude_unset=True).items():
        setattr(db_checklist, key, value)
    
    db.commit()
    db.refresh(db_checklist)
    return db_checklist


# ============= QUESTION ENDPOINTS =============
@app.post("/api/questions", response_model=schemas.Question)
def create_question(
    question: schemas.QuestionCreate,
    db: Session = Depends(get_db),
    current_user: models.User = Depends(get_current_user)
):
    if current_user.role != "admin":
        raise HTTPException(status_code=403, detail="No autorizado")
    
    db_question = models.Question(**question.dict())
    db.add(db_question)
    
    # Actualizar max_score del checklist
    checklist = db.query(models.Checklist).filter(
        models.Checklist.id == question.checklist_id
    ).first()
    if checklist:
        checklist.max_score += question.points
    
    db.commit()
    db.refresh(db_question)
    return db_question


@app.put("/api/questions/{question_id}", response_model=schemas.Question)
def update_question(
    question_id: int,
    question: schemas.QuestionUpdate,
    db: Session = Depends(get_db),
    current_user: models.User = Depends(get_current_user)
):
    if current_user.role != "admin":
        raise HTTPException(status_code=403, detail="No autorizado")
    
    db_question = db.query(models.Question).filter(models.Question.id == question_id).first()
    if not db_question:
        raise HTTPException(status_code=404, detail="Pregunta no encontrada")
    
    for key, value in question.dict(exclude_unset=True).items():
        setattr(db_question, key, value)
    
    db.commit()
    db.refresh(db_question)
    return db_question


@app.delete("/api/questions/{question_id}")
def delete_question(
    question_id: int,
    db: Session = Depends(get_db),
    current_user: models.User = Depends(get_current_user)
):
    if current_user.role != "admin":
        raise HTTPException(status_code=403, detail="No autorizado")
    
    db_question = db.query(models.Question).filter(models.Question.id == question_id).first()
    if not db_question:
        raise HTTPException(status_code=404, detail="Pregunta no encontrada")
    
    db.delete(db_question)
    db.commit()
    return {"message": "Pregunta eliminada"}


# ============= INSPECTION ENDPOINTS =============
@app.get("/api/inspections", response_model=List[schemas.Inspection])
def get_inspections(
    skip: int = 0,
    limit: int = 100,
    vehicle_plate: str = None,
    status: str = None,
    db: Session = Depends(get_db),
    current_user: models.User = Depends(get_current_user)
):
    query = db.query(models.Inspection)
    
    # Mecánicos solo ven sus inspecciones
    if current_user.role == "mechanic":
        query = query.filter(models.Inspection.mechanic_id == current_user.id)
    
    if vehicle_plate:
        query = query.filter(models.Inspection.vehicle_plate.contains(vehicle_plate))
    
    if status:
        query = query.filter(models.Inspection.status == status)
    
    return query.order_by(models.Inspection.created_at.desc()).offset(skip).limit(limit).all()


@app.get("/api/inspections/{inspection_id}", response_model=schemas.InspectionDetail)
def get_inspection(
    inspection_id: int,
    db: Session = Depends(get_db),
    current_user: models.User = Depends(get_current_user)
):
    inspection = db.query(models.Inspection).options(
        joinedload(models.Inspection.checklist).joinedload(models.Checklist.questions),
        joinedload(models.Inspection.mechanic),
        joinedload(models.Inspection.answers).joinedload(models.Answer.question),
        joinedload(models.Inspection.answers).joinedload(models.Answer.media_files)
    ).filter(models.Inspection.id == inspection_id).first()
    
    if not inspection:
        raise HTTPException(status_code=404, detail="Inspección no encontrada")
    
    return inspection


@app.post("/api/inspections", response_model=schemas.Inspection)
def create_inspection(
    inspection: schemas.InspectionCreate,
    db: Session = Depends(get_db),
    current_user: models.User = Depends(get_current_user)
):
    # Obtener max_score del checklist
    checklist = db.query(models.Checklist).filter(
        models.Checklist.id == inspection.checklist_id
    ).first()
    
    if not checklist:
        raise HTTPException(status_code=404, detail="Checklist no encontrado")
    
    db_inspection = models.Inspection(
        **inspection.dict(),
        mechanic_id=current_user.id,
        max_score=checklist.max_score
    )
    db.add(db_inspection)
    db.commit()
    db.refresh(db_inspection)
    return db_inspection


@app.put("/api/inspections/{inspection_id}", response_model=schemas.Inspection)
def update_inspection(
    inspection_id: int,
    inspection: schemas.InspectionUpdate,
    db: Session = Depends(get_db),
    current_user: models.User = Depends(get_current_user)
):
    db_inspection = db.query(models.Inspection).filter(
        models.Inspection.id == inspection_id
    ).first()
    
    if not db_inspection:
        raise HTTPException(status_code=404, detail="Inspección no encontrada")
    
    for key, value in inspection.dict(exclude_unset=True).items():
        setattr(db_inspection, key, value)
    
    db.commit()
    db.refresh(db_inspection)
    return db_inspection


@app.post("/api/inspections/{inspection_id}/complete", response_model=schemas.Inspection)
def complete_inspection(
    inspection_id: int,
    db: Session = Depends(get_db),
    current_user: models.User = Depends(get_current_user)
):
    inspection = db.query(models.Inspection).filter(
        models.Inspection.id == inspection_id
    ).first()
    
    if not inspection:
        raise HTTPException(status_code=404, detail="Inspección no encontrada")
    
    # Calcular score
    answers = db.query(models.Answer).filter(models.Answer.inspection_id == inspection_id).all()
    total_score = sum(a.points_earned for a in answers)
    flagged_count = sum(1 for a in answers if a.is_flagged)
    
    inspection.score = total_score
    inspection.percentage = (total_score / inspection.max_score * 100) if inspection.max_score > 0 else 0
    inspection.flagged_items_count = flagged_count
    inspection.status = "completed"
    inspection.completed_at = datetime.utcnow()
    
    db.commit()
    db.refresh(inspection)
    return inspection


# ============= ANSWER ENDPOINTS =============
@app.post("/api/answers", response_model=schemas.Answer)
def create_answer(
    answer: schemas.AnswerCreate,
    db: Session = Depends(get_db),
    current_user: models.User = Depends(get_current_user)
):
    # Obtener la pregunta para saber los puntos
    question = db.query(models.Question).filter(models.Question.id == answer.question_id).first()
    
    if not question:
        raise HTTPException(status_code=404, detail="Pregunta no encontrada")
    
    # Calcular puntos según status
    points_earned = 0
    if answer.status == "ok":
        points_earned = question.points
    elif answer.status == "warning":
        points_earned = int(question.points * 0.5)
    
    db_answer = models.Answer(
        **answer.dict(),
        points_earned=points_earned
    )
    db.add(db_answer)
    db.commit()
    db.refresh(db_answer)
    return db_answer


@app.put("/api/answers/{answer_id}", response_model=schemas.Answer)
def update_answer(
    answer_id: int,
    answer: schemas.AnswerUpdate,
    db: Session = Depends(get_db),
    current_user: models.User = Depends(get_current_user)
):
    db_answer = db.query(models.Answer).filter(models.Answer.id == answer_id).first()
    
    if not db_answer:
        raise HTTPException(status_code=404, detail="Respuesta no encontrada")
    
    # Recalcular puntos si cambió el status
    if answer.status and answer.status != db_answer.status:
        question = db.query(models.Question).filter(
            models.Question.id == db_answer.question_id
        ).first()
        
        if answer.status == "ok":
            db_answer.points_earned = question.points
        elif answer.status == "warning":
            db_answer.points_earned = int(question.points * 0.5)
        else:
            db_answer.points_earned = 0
    
    for key, value in answer.dict(exclude_unset=True).items():
        setattr(db_answer, key, value)
    
    db.commit()
    db.refresh(db_answer)
    return db_answer


# ============= MEDIA FILE ENDPOINTS =============
@app.post("/api/answers/{answer_id}/upload", response_model=schemas.MediaFile)
async def upload_photo(
    answer_id: int,
    file: UploadFile = File(...),
    db: Session = Depends(get_db),
    current_user: models.User = Depends(get_current_user)
):
    # Verificar que la respuesta existe
    answer = db.query(models.Answer).filter(models.Answer.id == answer_id).first()
    if not answer:
        raise HTTPException(status_code=404, detail="Respuesta no encontrada")
    
    # Crear directorio si no existe
    upload_dir = f"uploads/inspection_{answer.inspection_id}"
    os.makedirs(upload_dir, exist_ok=True)
    
    # Guardar archivo
    file_extension = file.filename.split(".")[-1]
    file_name = f"answer_{answer_id}_{datetime.now().timestamp()}.{file_extension}"
    file_path = os.path.join(upload_dir, file_name)
    
    with open(file_path, "wb") as buffer:
        shutil.copyfileobj(file.file, buffer)
    
    # Crear registro en BD
    media_file = models.MediaFile(
        answer_id=answer_id,
        file_path=file_path,
        file_type="image"
    )
    db.add(media_file)
    db.commit()
    db.refresh(media_file)
    
    return media_file


# ============= AI ANALYSIS =============
@app.get("/api/ai/models", response_model=List[schemas.AIModelInfo])
def get_available_ai_models(current_user: models.User = Depends(get_current_user)):
    """Obtener lista de modelos de IA disponibles"""
    if current_user.role != "admin":
        raise HTTPException(status_code=403, detail="Solo administradores pueden ver modelos de IA")
    
    models_list = [
        # OpenAI Models
        {
            "id": "gpt-4o",
            "name": "GPT-4o (Recomendado)",
            "provider": "openai",
            "description": "Modelo multimodal más avanzado de OpenAI, rápido y preciso para análisis de imágenes"
        },
        {
            "id": "gpt-4o-mini",
            "name": "GPT-4o Mini",
            "provider": "openai",
            "description": "Versión compacta y económica de GPT-4o, ideal para análisis rápidos"
        },
        {
            "id": "gpt-4-turbo",
            "name": "GPT-4 Turbo",
            "provider": "openai",
            "description": "Modelo potente con capacidades de visión y contexto amplio"
        },
        {
            "id": "gpt-4-vision-preview",
            "name": "GPT-4 Vision (Preview)",
            "provider": "openai",
            "description": "Modelo específico para análisis de imágenes (versión previa)"
        },
        # Gemini Models - Actualizados a versiones 2.0, 2.5 y 3.0
        {
            "id": "gemini-3-pro-preview",
            "name": "Gemini 3 Pro Preview (Último)",
            "provider": "gemini",
            "description": "Modelo de próxima generación en preview, máxima capacidad de análisis"
        },
        {
            "id": "gemini-2.5-pro",
            "name": "Gemini 2.5 Pro (Recomendado)",
            "provider": "gemini",
            "description": "Último modelo estable con excelente análisis visual y razonamiento avanzado"
        },
        {
            "id": "gemini-2.5-flash",
            "name": "Gemini 2.5 Flash",
            "provider": "gemini",
            "description": "Versión rápida del 2.5, ideal para inspecciones en tiempo real"
        },
        {
            "id": "gemini-2.0-flash",
            "name": "Gemini 2.0 Flash",
            "provider": "gemini",
            "description": "Modelo rápido y eficiente de la generación 2.0"
        },
        {
            "id": "gemini-1.5-pro-latest",
            "name": "Gemini 1.5 Pro Latest",
            "provider": "gemini",
            "description": "Versión estable 1.5 con contexto de 2M tokens"
        },
        {
            "id": "gemini-1.5-flash-latest",
            "name": "Gemini 1.5 Flash Latest",
            "provider": "gemini",
            "description": "Modelo 1.5 rápido para análisis básicos"
        }
    ]
    
    return models_list


@app.get("/api/ai/configuration", response_model=schemas.AIConfiguration)
def get_ai_configuration(
    db: Session = Depends(get_db),
    current_user: models.User = Depends(get_current_user)
):
    """Obtener configuración de IA actual"""
    if current_user.role != "admin":
        raise HTTPException(status_code=403, detail="Solo administradores pueden ver configuración de IA")
    
    config = db.query(models.AIConfiguration).filter(
        models.AIConfiguration.is_active == True
    ).first()
    
    if not config:
        raise HTTPException(status_code=404, detail="No hay configuración de IA activa")
    
    return config


@app.post("/api/ai/configuration", response_model=schemas.AIConfiguration)
def create_ai_configuration(
    config: schemas.AIConfigurationCreate,
    db: Session = Depends(get_db),
    current_user: models.User = Depends(get_current_user)
):
    """Crear o actualizar configuración de IA"""
    if current_user.role != "admin":
        raise HTTPException(status_code=403, detail="Solo administradores pueden configurar IA")
    
    # Desactivar configuraciones anteriores
    db.query(models.AIConfiguration).update({"is_active": False})
    
    # Crear nueva configuración
    new_config = models.AIConfiguration(
        provider=config.provider,
        api_key=config.api_key,
        model_name=config.model_name,
        is_active=True
    )
    
    db.add(new_config)
    db.commit()
    db.refresh(new_config)
    
    return new_config


@app.put("/api/ai/configuration/{config_id}", response_model=schemas.AIConfiguration)
def update_ai_configuration(
    config_id: int,
    config_update: schemas.AIConfigurationUpdate,
    db: Session = Depends(get_db),
    current_user: models.User = Depends(get_current_user)
):
    """Actualizar configuración de IA existente"""
    if current_user.role != "admin":
        raise HTTPException(status_code=403, detail="Solo administradores pueden actualizar configuración de IA")
    
    config = db.query(models.AIConfiguration).filter(
        models.AIConfiguration.id == config_id
    ).first()
    
    if not config:
        raise HTTPException(status_code=404, detail="Configuración no encontrada")
    
    # Actualizar campos
    for key, value in config_update.dict(exclude_unset=True).items():
        setattr(config, key, value)
    
    db.commit()
    db.refresh(config)
    
    return config


@app.delete("/api/ai/configuration/{config_id}")
def delete_ai_configuration(
    config_id: int,
    db: Session = Depends(get_db),
    current_user: models.User = Depends(get_current_user)
):
    """Eliminar configuración de IA"""
    if current_user.role != "admin":
        raise HTTPException(status_code=403, detail="Solo administradores pueden eliminar configuración de IA")
    
    config = db.query(models.AIConfiguration).filter(
        models.AIConfiguration.id == config_id
    ).first()
    
    if not config:
        raise HTTPException(status_code=404, detail="Configuración no encontrada")
    
    db.delete(config)
    db.commit()
    
    return {"message": "Configuración eliminada correctamente"}


@app.post("/api/analyze-image")
async def analyze_image(
    file: UploadFile = File(...),
    question_id: int = None,
    db: Session = Depends(get_db),
    current_user: models.User = Depends(get_current_user)
):
    """
    Analiza una imagen usando IA para sugerir respuestas
    Usa la configuración de IA activa (OpenAI o Gemini)
    """
    # Obtener configuración de IA activa
    ai_config = db.query(models.AIConfiguration).filter(
        models.AIConfiguration.is_active == True
    ).first()
    
    if not ai_config:
        return {
            "status": "disabled",
            "message": "No hay configuración de IA activa. Configure en Settings."
        }
    
    # Guardar imagen temporalmente
    import base64
    
    contents = await file.read()
    image_b64 = base64.b64encode(contents).decode('utf-8')
    
    # Obtener contexto de la pregunta si se proporciona
    question_obj = None
    if question_id:
        question_obj = db.query(models.Question).filter(models.Question.id == question_id).first()
    
    try:
        # Construir prompt dinámico basado en la pregunta específica
        if question_obj:
            # Prompt altamente específico para la pregunta
            question_text = question_obj.text
            question_type = question_obj.type
            section = question_obj.section
            
            system_prompt = f"""Eres un mecánico experto realizando una inspección vehicular.

PREGUNTA ESPECÍFICA A RESPONDER: "{question_text}"
Sección: {section}

Analiza la imagen ÚNICAMENTE para responder esta pregunta específica. 
Sé directo y enfócate solo en lo que la pregunta solicita.

Responde en formato JSON:
{{
  "status": "ok|minor|critical",
  "observations": "Respuesta específica a: {question_text}",
  "recommendation": "Acción si aplica",
  "confidence": 0.0-1.0
}}

IMPORTANTE: 
- Responde SOLO lo que la pregunta pide
- No des información genérica del vehículo
- Sé específico y técnico
- Si la pregunta es pass/fail, indica claramente si pasa o falla
- Si la pregunta es bueno/regular/malo, indica el estado específico del componente"""

            user_message = f"Inspecciona la imagen y responde específicamente: {question_text}"
        else:
            # Fallback para análisis general
            system_prompt = """Eres un experto mecánico automotriz. Analiza la imagen y proporciona:
1. Estado del componente (bueno/regular/malo)
2. Nivel de criticidad (ok/minor/critical)
3. Observaciones técnicas breves
4. Recomendación de acción

Responde en formato JSON:
{
  "status": "ok|minor|critical",
  "observations": "descripción técnica",
  "recommendation": "acción sugerida",
  "confidence": 0.0-1.0
}"""
            user_message = "Analiza este componente del vehículo para la inspección general."

        if ai_config.provider == "openai":
            import openai
            openai.api_key = ai_config.api_key
            
            response = openai.ChatCompletion.create(
                model=ai_config.model_name,
                messages=[
                    {"role": "system", "content": system_prompt},
                    {
                        "role": "user",
                        "content": [
                            {
                                "type": "text",
                                "text": user_message
                            },
                            {
                                "type": "image_url",
                                "image_url": {"url": f"data:image/jpeg;base64,{image_b64}"}
                            }
                        ]
                    }
                ],
                max_tokens=500
            )
            
            ai_response = response.choices[0].message.content
            
        elif ai_config.provider == "gemini":
            import google.generativeai as genai
            from PIL import Image
            from io import BytesIO
            
            genai.configure(api_key=ai_config.api_key)
            model = genai.GenerativeModel(ai_config.model_name)
            
            # Convertir base64 a imagen PIL
            image = Image.open(BytesIO(contents))
            
            prompt = f"{system_prompt}\n\n{user_message}"
            response = model.generate_content([prompt, image])
            ai_response = response.text
            
        else:
            return {
                "success": False,
                "error": f"Provider {ai_config.provider} no soportado"
            }
        
        # Intentar parsear como JSON, si falla, usar texto plano
        try:
            import json
            import re
            
            # Limpiar markdown code blocks si existen
            cleaned_response = ai_response.strip()
            
            # Remover ```json ... ``` si existe
            if cleaned_response.startswith('```'):
                # Extraer contenido entre ``` markers
                match = re.search(r'```(?:json)?\s*\n?(.*?)\n?```', cleaned_response, re.DOTALL)
                if match:
                    cleaned_response = match.group(1).strip()
            
            analysis = json.loads(cleaned_response)
        except:
            # Si no es JSON válido, crear estructura básica
            analysis = {
                "status": "ok",
                "observations": ai_response,
                "recommendation": "Revisar manualmente",
                "confidence": 0.7
            }
        
        return {
            "success": True,
            "analysis": analysis,
            "raw_response": ai_response,
            "model": ai_config.model_name,
            "provider": ai_config.provider
        }
        
    except Exception as e:
        print(f"Error en análisis AI: {e}")
        import traceback
        traceback.print_exc()
        return {
            "success": False,
            "error": str(e),
            "message": "Error analyzing image with AI. Please check AI configuration in Settings."
        }
    
    try:
        import openai
        openai.api_key = settings.OPENAI_API_KEY
        
        # Prompt especializado para inspección vehicular
        system_prompt = """Eres un experto mecánico automotriz. Analiza la imagen y proporciona:
1. Estado del componente (bueno/regular/malo)
2. Nivel de criticidad (ok/minor/critical)
3. Observaciones técnicas breves
4. Recomendación de acción

Responde en formato JSON:
{
  "status": "ok|minor|critical",
  "observations": "descripción técnica",
  "recommendation": "acción sugerida",
  "confidence": 0.0-1.0
}"""

        response = openai.ChatCompletion.create(
            model="gpt-4-vision-preview" if "gpt-4" in str(settings.OPENAI_API_KEY) else "gpt-4o",
            messages=[
                {
                    "role": "system",
                    "content": system_prompt
                },
                {
                    "role": "user",
                    "content": [
                        {
                            "type": "text",
                            "text": f"Analiza este componente del vehículo.\n{question_context}"
                        },
                        {
                            "type": "image_url",
                            "image_url": {
                                "url": f"data:image/jpeg;base64,{image_b64}"
                            }
                        }
                    ]
                }
            ],
            max_tokens=500
        )
        
        ai_response = response.choices[0].message.content
        
        # Intentar parsear como JSON, si falla, usar texto plano
        try:
            import json
            analysis = json.loads(ai_response)
        except:
            # Si no es JSON válido, crear estructura básica
            analysis = {
                "status": "ok",
                "observations": ai_response,
                "recommendation": "Revisar manualmente",
                "confidence": 0.7
            }
        
        return {
            "success": True,
            "analysis": analysis,
            "raw_response": ai_response,
            "model": "gpt-4-vision"
        }
        
    except Exception as e:
        print(f"Error en análisis AI: {e}")
        return {
            "success": False,
            "error": str(e),
            "message": "Error analyzing image with AI"
        }


# ============= HEALTH CHECK =============
@app.get("/")
def root():
    return {"message": "Checklist Inteligente API", "version": "1.0.0", "status": "running"}


@app.get("/health")
def health_check():
    return {"status": "healthy"}