app.py

from flask import Flask, render_template, request, jsonify
from werkzeug.utils import secure_filename
import os, json, time

APP_ROOT = os.path.dirname(os.path.abspath(__file__))
UPLOAD_DIR = os.path.join(APP_ROOT, 'uploads')
os.makedirs(UPLOAD_DIR, exist_ok=True)

app = Flask(__name__)
app.config['MAX_CONTENT_LENGTH'] = 25 * 1024 * 1024  # 25MB

# === Load system prompt ===
prompt_path = os.path.join(APP_ROOT, 'prompts', 'skin_prompt.txt')
if os.path.exists(prompt_path):
    with open(prompt_path, 'r', encoding='utf-8') as f:
        SYSTEM_PROMPT = f.read()
else:
    SYSTEM_PROMPT = "MISSING_PROMPT: please add prompts/skin_prompt.txt"

# === (Optional) Load JSON schema for basic shape validation ===
schema_path = os.path.join(APP_ROOT, 'schema', 'output_schema.json')
if os.path.exists(schema_path):
    with open(schema_path, 'r', encoding='utf-8') as f:
        OUTPUT_SCHEMA = json.load(f)
else:
    OUTPUT_SCHEMA = {}

def _stub_model_inference(image_paths, metadata: dict):
    """
    Deterministic stub to simulate the model output that conforms to the schema.
    Replace this with your real model call (e.g., OpenAI, local vision model, etc.).
    """
    num_images = len(image_paths)
    quality = 'good' if num_images and metadata.get('lighting', 'even') == 'even' else 'fair'

    result = {
        "image_quality": {
            "overall": quality,
            "issues": [] if quality == 'good' else ["lighting"],
            "retake_advice": "确保均匀光线、正面构图、去除滤镜与妆容。"
        },
        "global_assessment": {
            "skin_type_estimate": metadata.get('self_report_skin_type', 'unknown'),
            "overall_risk_score": 42,
            "notes": "演示结果：请接入真实模型以获得更准确评估。"
        },
        "region_scores": {
            "T_zone_oiliness": 58,
            "U_zone_dryness": 25,
            "cheek_redness": 18,
            "undereye_dark_circle": 40
        },
        "concerns": [
            {"name": "acne", "severity": 35, "evidence": "T区可见散在闭口样阴影。", "confidence": 0.72},
            {"name": "pores", "severity": 55, "evidence": "鼻翼与鼻梁两侧纹理增粗。", "confidence": 0.77}
        ],
        "pigmentation": {
            "uneven_tone": 28,
            "spots": 10,
            "melasma_like": 0,
            "sun_damage_like": 22
        },
        "aging_signs": {
            "fine_lines": 20,
            "wrinkles": 8,
            "loss_of_elasticity": 12,
            "texture_roughness": 30
        },
        "sensitivity_inflammation": {
            "redness": 18,
            "visible_capillaries": 5,
            "irritation": 10
        },
        "top_priorities": [
            {"name": "pores", "reason": "毛孔纹理明显，影响观感。"},
            {"name": "oil_control", "reason": "T区油脂分泌偏旺。"},
            {"name": "uneven_tone", "reason": "肤色轻度不匀影响通透感。"}
        ],
        "daily_plan": {
            "am": [
                "氨基酸洁面",
                "2-5%烟酰胺精华（避眼周）",
                "广谱防晒 SPF50+ PA++++"
            ],
            "pm": [
                "温和洁面",
                "0.5-2%水杨酸/杜鹃花酸按需点涂",
                "清爽型保湿乳"
            ],
            "extras_weekly": [
                "每周1-2次温和去角质（避免与酸叠加）"
            ]
        },
        "ingredient_recommendations": [
            {
                "goal": "oil_control",
                "actives": ["niacinamide 2-5%", "salicylic acid 0.5-2%"],
                "usage": "先低后高，观察耐受，白天注意防晒。",
                "cautions": "与强酸类/维A初期分开使用，避免刺激叠加。"
            }
        ],
        "risk_flags": [],
        "follow_up": {
            "reshoot_tips": ["remove_makeup", "even_lighting", "front_facing", "no_filters"],
            "retake_interval_days": 28
        }
    }

    return result

@app.route('/')
def index():
    return render_template('index.html')

@app.post('/api/analyze')
def analyze():
    files = request.files.getlist('images')
    metadata_json = request.form.get('metadata', '{}')
    try:
        metadata = json.loads(metadata_json)
    except Exception:
        return jsonify({"error": "metadata must be valid JSON"}), 400

    saved = []
    ts = int(time.time())
    for f in files[:5]:  # cap 5 images
        if not f.filename:
            continue
        fname = f"{ts}_" + secure_filename(f.filename)
        path = os.path.join(UPLOAD_DIR, fname)
        f.save(path)
        saved.append(path)

    # Call your model (stubbed here)
    output = _stub_model_inference(saved, metadata)

    # Minimal shape check
    required_top_keys = [
        "image_quality", "global_assessment", "region_scores", "concerns",
        "pigmentation", "aging_signs", "sensitivity_inflammation",
        "top_priorities", "daily_plan", "ingredient_recommendations",
        "risk_flags", "follow_up"
    ]
    missing = [k for k in required_top_keys if k not in output]
    if missing:
        return jsonify({"error": f"missing keys: {missing}"}), 500

    return jsonify(output)

if __name__ == '__main__':
    app.run(host='0.0.0.0', port=5000, debug=True)