#==================================================
# IMPORT: TEST
#==================================================
import asyncio
from ably import AblyRest
def token_ably(apikey, token):
    try:
        ably = AblyRest(apikey)
        ##client_id = request.args.get("clientId")
        ##client_id = request.json.get('clientId')
        client_id = request.headers.get('clientId')
        ##client_id = generate_random_id()
        token_request_obj = asyncio.run(ably.auth.create_token_request({ 
            "client_id": client_id,
            "ttl": 60 * 60 * 1000 * 24,  # ⏱️ 有效 24 小時（60 分 x 60 秒 x 1000 毫秒）
        }))  # 非匿名
        token_request_dict = token_request_obj.to_dict()  # 轉成 dict
        token_request_dict["token"] = token  # 新增一個 key-value
        return jsonify(token_request_dict)
    except Exception as e:
        return jsonify({"status": "fail", "reason": "Invalid link", "detail": str(e)}), 403

        

import stanza  ##safe
nlp = stanza.Pipeline("en")  ##safe

##doc = nlp(text)
##for sent in doc.sentences:

def get_dependencies(doc):
    deps = []

    for sent in doc.sentences:
        for w in sent.words:
            head_word = None if w.head == 0 else sent.words[w.head - 1]

            deps.append({
                "sent_id": sent.id if hasattr(sent, "id") else None,
                "dependent_id": w.id,
                "dependent": w.text,
                "head_id": w.head,
                "head": "ROOT" if w.head == 0 else head_word.text,
                "deprel": w.deprel
            })

    return deps
def get_dependencies_v0(sent):
    deps = []

    for w in sent.words:
        head_word = None if w.head == 0 else sent.words[w.head - 1]

        deps.append({
            "dependent_id": w.id,
            "dependent": w.text,
            "head_id": w.head,
            "head": "ROOT" if w.head == 0 else head_word.text,
            "deprel": w.deprel
        })

    return deps

#==================================================
# IMPORT: ALL
#==================================================
from urllib.parse import urlencode                  #webpush
import requests                                     #deurl  #webpush
from requests.exceptions import RequestException    #deurl
import re                                           #deurl
##import socket

from gradio_client import Client                    #X
from huggingface_hub import HfApi                   #dataset

import os                                           #SECRET
from cryptography.fernet import Fernet              #FUNCTION
import hashlib                                      #FUNCTION
import json                                         #FUNCTION  #webpush
import base64                                       #FUNCTION
import string                                       #FUNCTION
import random                                       #FUNCTION
import ast                                          #FUNCTION
import pandas as pd                                 #FUNCTION

import io                                           #dataset

##import json                                          
from pywebpush import webpush, WebPushException     #webpush
##import requests


from google import genai                            #llms
from google.genai.errors import APIError            #llms

#==================================================
# SECRET: ALL
#==================================================
        
def checkSecret(SECRET):
    if SECRET:
        # 遮罩 Token 顯示前 5 個字符，其餘用 *** 代替
        masked_token = SECRET[:3] + "***" + SECRET[-3:]
        
        # ⚠️ 僅在您自己檢查 Logs 時使用，檢查完畢後應刪除此行或註釋掉！
        print(f"SECRET: {masked_token}") 
        
        # 繼續運行您的應用程式...
        return SECRET
    else:
        print(f"SECRET: The 'SECRET' is not a setting.")
        return False


GS_ID = os.getenv("GOOGLE_SHEET_ID");checkSecret(GS_ID)
HF_READ = os.getenv("HUGGINGFACE_READ_TOKEN");checkSecret(HF_READ)
HF_WRITE = os.getenv("HUGGINGFACE_WRITE_TOKEN");checkSecret(HF_WRITE)
OC_APIKEY = os.getenv("OPENAI_CHATGPT_APIKEY");checkSecret(OC_APIKEY)
AR_APIKEY = os.environ.get("ABLY_REALTIME_APIKEY");checkSecret(AR_APIKEY)
TG_TOKEN = os.environ.get("TELEGRAM_BOT_TOKEN");checkSecret(TG_TOKEN)

NS_SECRET = os.getenv("NINS_CRYPT_SECRET");checkSecret(NS_SECRET)  #webpush  #32bit!
VAPID_PUBLIC = os.getenv("VAPID_PUBLIC_KEY");checkSecret(VAPID_PUBLIC)  #webpush
VAPID_PRIVATE = os.getenv("VAPID_PRIVATE_KEY");checkSecret(VAPID_PRIVATE)  #webpush
VAPID_CLAIMS = {
    "sub": "mailto:duguaynins@gmail.com",
    ##"aud": "https://copyright.nins.cc"
}
key = NS_SECRET 
cipher = Fernet(key)
print(VAPID_CLAIMS)

#==================================================
# FUNCTION: ALL
#==================================================

##import json
def str_to_json(string):
    return json.loads(string)

def json_to_str(object):
    return json.dumps(object)

##import json
##import ast
def inputs_to_dict(inputs):
    # 如果本來是 dict 直接用
    if isinstance(inputs, dict):
        pass
        return inputs
    # 如果是 JSON 字串
    elif isinstance(inputs, str):
        try:
            inputs = str_to_json(inputs)          # 先試 JSON
        except json.JSONDecodeError:
            inputs = ast.literal_eval(inputs)    # 再試 Python dict 格式
        return inputs
    # 如果都不是
    else: raise ValueError(f"Cannot convert to dict: {inputs}")
                           
##import string
##import random
def generate_random_id():
    azAZ = string.ascii_letters  # abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ
    random_id = ''.join(random.choices(azAZ, k=32)) 
    return random_id
    
##import base64
def image_to_base64(path):
    with open(path, "rb") as f:
        return base64.b64encode(f.read()).decode("utf-8")


def dohash_data(data: str) -> str:
    # "hello".encode("utf-8")  # b"hello".decode("utf-8")  # "utf-8", "ascii", "big5", ""
    # hashlib.md5, hashlib.sha1, hashlib.sha256, hashlib.sha512, hashlib.blake2b *** # bytes only!!!
    return hashlib.sha256(data.encode("utf-8")).hexdigest()
    
def encrypt_data(data: str) -> str:
    return cipher.encrypt(data.encode("utf-8")).decode("utf-8")

def decrypt_data(data: str) -> str:
    return cipher.decrypt(data.encode("utf-8")).decode("utf-8")

def encrypt_data_to_bytes(data: str) -> bytes:
    return cipher.encrypt(data.encode("utf-8"))


##search_sheet(gid, index="index", target="webpush", signal="tag", verify="***")
def search_sheet(gid, index, target, signal, verify, value):  ##token_channel  ##index='email', target='webpush', signal='time', verify="*O*"
    url = f'https://docs.google.com/spreadsheets/d/{gid}/export?format=csv'
   
    df = pd.read_csv(url,
                   encoding='utf-8',
                   dtype=str,
                   keep_default_na=False)
   
    # 顯示重複、清理重複
    ##duplicates = df[df[index].duplicated(keep=False)]
    ##print(f"duplicated: \n{duplicates}")
    df = df.drop_duplicates(subset=[index], keep='last')  ##keep='first'  ##只保留最後的重複值
   
    # 篩選並取出 status 為 "正常"
    ##df = df[df["C"] != "HEHE"]  ##只留不等於HEHE
    df = df[df[signal].isin(["***", "*O*", verify])]  ##"TEST"  ##驗證欄位
    values = df[index].tolist()
    values = set(values)
    value = value.lower().strip()
    
    print(f"{len(values)} >>> {values}")
   
    if value in values:
        ##return True
        search = dict(zip(df[index], df[target]))
        webpush = search.get(value)
        return webpush
    else:
        ##return False
        return None

def upload_to_hf(path_or_fileobj, path_in_repo, repo_id, token=None):
    token = token or HF_WRITE
    path_or_fileobj.seek(0)  # 確保從開頭讀
    api = HfApi()
    api.upload_file(
        path_or_fileobj=path_or_fileobj,
        path_in_repo=path_in_repo,
        repo_id=repo_id,
        repo_type="dataset",
        token=token
    )
    return True

#==================================================
# FUNCTION: MODEL.SYS
#==================================================

from sentence_transformers import SentenceTransformer, util
##import gradio as gr

# 載入多語言模型
##model = SentenceTransformer("sentence-transformers/multi-qa-MiniLM-L6-cos-v1")  # 90.9M
##model = SentenceTransformer("sentence-transformers/all-MiniLM-L6-v2")  # 90.9M
##model = SentenceTransformer('sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2')  # 50+  # 471M  # Microsoft Research / UKP Lab
##model = SentenceTransformer('sentence-transformers/distiluse-base-multilingual-cased-v2')   # 50+  # 539M  # Hugging Face / UKP Lab
##model = SentenceTransformer("sentence-transformers/paraphrase-multilingual-mpnet-base-v2")  # 50+*  # 1.11GB  # Microsoft Research / UKP Lab
model_meta = SentenceTransformer("sentence-transformers/stsb-xlm-r-multilingual")         # 100+*  # 1.11G  # Trained by Sentence-Transformers based on XLM-RoBERTa
model_goog = SentenceTransformer("sentence-transformers/LaBSE")                  # 100+*  # 1.88GB  # GOOGLE Research

"""
# 定義計算相似度的函數
def semantic_similarity(inputs):
    
    print(inputs);print(type(inputs))
    inputs = inputs_to_dict(inputs)
    print(inputs);print(type(inputs))
    
    sent1, sent2 = inputs.values()
    embeddings = model.encode([sent1, sent2], convert_to_tensor=True)
    cos_sim = util.pytorch_cos_sim(embeddings[0], embeddings[1])
    return round(cos_sim.item(), 4)  """


# 定義計算相似度的函數
def semantic_similarity(embeddings):
    ##sent1, sent2 = inputs.values()
    ##embeddings = model.encode([sent1, sent2], convert_to_tensor=True)
    cos_sim = util.pytorch_cos_sim(embeddings[0], embeddings[1])
    return round(cos_sim.item(), 4)  # float, int, str, dict

def fn_main_ds(inputs):
    
    ##print(inputs);print(type(inputs))  ##20260402
    inputs = inputs_to_dict(inputs)
    ##print(inputs);print(type(inputs))  ##20260402
    
    sent1, sent2 = inputs.values()
    sim1 = semantic_similarity(model_goog.encode([sent1, sent2], convert_to_tensor=True))
    sim2 = semantic_similarity(model_meta.encode([sent1, sent2], convert_to_tensor=True))
    return {"goog": sim1, "meta": sim2}  ##20260420
    return str('{' + 'goog:' + str(sim1) + '||' + 'meta:' + str(sim2) + '}')  # float, int, str, dict
    return json.dumps({  ##20260420
        "goog": sim1,
        "meta": sim2
    })
    

#==================================================
# FUNCTION: LLMs.SYS
#==================================================

##import requests
def openai_text_generate(data):
    ##data = request.json
    
    if not data:
        return {"error": "data?"}, 400
    if not data.get("system"):
        return {"error": "system?"}, 400
    if not data.get("user"):
        return {"error": "user?"}, 400
    if not data.get("models"):
        return {"error": "models?"}, 400
    ## if not data.get("apikey"): return {"error": "apikey?"}, 400
        
    models = data.get("models", "gpt-4o-mini")
    system = data.get("system", "sys prompt")
    user = data.get("user", "usr prompt")
    apikey = data.get("apikey", os.getenv("OPENAI_CHATGPT_APIKEY"))  ##不會使用到後者

    headers = {
        ##"Content-Type": "application/json", "Authorization": f"Bearer {os.getenv('OPENAI_CHATGPT_APIKEY')}"
        "Content-Type": "application/json", "Authorization": f"Bearer {apikey}"
    }

    payload = {
        "model": f"{models}",
        ##"model": "gpt-4o-mini",
        ##"model": "gpt-3.5-turbo",
        ##"model": "gpt-4o-nano",
        "messages": [
            ##{"role": "system", "content": f"You are a professional translator who can translate any language into grammatically perfect English."},
            {"role": "system", "content": f"{system}"},
            {"role": "user", "content": f"{user}"},
        ],
        "max_tokens": 500, "temperature": 0,  # 翻譯建議使用 0
    }
    
    response = requests.post("https://api.openai.com/v1/chat/completions", headers=headers, json=payload)

    if response.status_code == 200:
        return response.json()['choices'][0]['message']['content'].strip()
        
        result = response.json()['choices'][0]['message']['content'].strip()
        # 直接返回 dict，不用 jsonify
        return {
            "system": data.get('system'),
            "user": data.get('user'),
            "result": result,
        }
    else:
        ##return jsonify({"error": response.text}), response.status_code
        return {"error": response.text}, response.status_code


##from google import genai
##from google.genai.errors import APIError
def google_text_generation(data):
    ##data = request.json
    
    if not data:
        return {"error": "data?"}, 400
    if not data.get("system"):
        return {"error": "system?"}, 400
    if not data.get("user"):
        return {"error": "user?"}, 400
    if not data.get("models"):
        return {"error": "models?"}, 400
    ## if not data.get("apikey"): return {"error": "apikey?"}, 400
        
    models = data.get("models", "gemini-1.5-flash")
    system = data.get("system", "sys prompt")
    user = data.get("user", "usr prompt")
    apikey = data.get("apikey", os.getenv("GOOGLE_GEMINI_APIKEY"))  ##不會使用到後者
    
    ##client = genai.Client(api_key=os.getenv("GOOGLE_GEMINI_APIKEY"))
    client = genai.Client(api_key=apikey)
    
    # --- 步驟 3: 設定和呼叫模型 ---
    '''
    # 呼叫模型生成內容
    response = client.models.generate_content(
        model=models,
        contents=prompt
    )  '''

    # 呼叫模型生成內容
    response = client.models.generate_content(
        model=models,
        config={"system_instruction": system},
        contents=user,
    )
    
    # 輸出模型的結果
    ##print("--- 模型回覆 ---")  ##20260402
    ##print(response.text)  ##20260402
    ##print("---------------")  ##20260402

    return response.text

"""
from transformers import pipeline

# 建立文字生成管線
generator = pipeline("text-generation", model="gpt2")  # 可換成你喜歡的 HF 模型
def pipeline_text_generation(data):
    user = data.get("user", "auto")
    result = generator(user, max_length=100, do_sample=True, top_p=0.9)
    return result[0]['generated_text']  ##data.result.confidences  ##data是JS變數名稱  ##const data = await HuggingFaceAPI(...);
"""

def fn_main_llms(inputs):
    inputs = inputs_to_dict(inputs)##print(inputs)  ##20260402
    ##return pipeline_text_generation(inputs)
    if inputs.get("models").lower().startswith("gemini"): return google_text_generation(inputs)
    elif inputs.get("models").lower().startswith("gpt"): return openai_text_generate(inputs)
    else: return "???"

#==================================================
# FUNCTION: DATASET.SYS
#==================================================

##from cryptography.fernet import Fernet
##import hashlib
##import io
def update_file(id, data, path_in_repo, repo_id, folder="default"):
    try:
            
        encrypted_data = encrypt_data(data)
        decrypted_data = decrypt_data(encrypted_data)
        content = io.BytesIO(encrypted_data.encode("utf-8"))  ##text_content
        ##content = io.BytesIO( encrypt_data_to_bytes(data) )   ##***
        
        ##user = dohash_data( id.strip().lower().encode("utf-8") )
        user = dohash_data( id.strip().lower() )
        ##path_in_repo = path_in_repo or f"{folder}/{user}/data.txt"
        path_in_repo = f"{folder}/{user}/data.txt"

        repo_id = repo_id  # f"duguaynins/dataset-v20260406"
        if not repo_id: return 404

        res = upload_to_hf(path_or_fileobj=content, 
                     path_in_repo=path_in_repo, 
                     repo_id=repo_id, 
                     token=None
        )
        print("HF result:", res)
        print("DEBUG", flush=True)

        return 200
    except Exception as e: return 500

def fn_main_dataset(inputs):
    id, data, path_in_repo, repo_id, folder = (
        inputs["id"],
        inputs["data"],
        inputs["path_in_repo"],
        inputs["repo_id"],
        inputs["folder"]
    )
    return update_file(id, data, path_in_repo, repo_id, folder)

#==================================================
# FUNCTION: WEBPUSH.SYS
#==================================================

##VAPID_PUBLIC_KEY = VAPID_PUBLIC
##VAPID_PRIVATE_KEY = VAPID_PRIVATE

def send_push_notifications(string):
    """
    批次推播通知，使用全域變數 subscriptions、payload、VAPID_PRIVATE_KEY、VAPID_CLAIMS
    """
    
    payload = str_to_json(string)
    body = payload.get("data", {}).get("body", "")
    
    params = {  ##追加網址的參數  ##20260414
        "userInput": payload["data"]["body"]
    }

    payload["data"]["title"] = "nins/©"
    ##payload["data"]["url"] = "https://copyright.nins.cc/"  ##鎖定網址的推放
    payload["data"]["url"] = (  ##追加網址的參數  ##20260414
        "https://copyright.nins.cc/?" + urlencode(params)
    )

    data = json_to_str( payload.get("data", {}) )
    subscriptions = payload.get("subscription", [])

    email = payload.get("data", {}).get("mail", "")  ##email
    
    for sub in subscriptions:
            
        try:
            ##if type(sub) == str:  ##pass  ##進行解碼
            if isinstance(sub, str):
                sub = decrypt_data(sub)  ##20260406
                sub = str_to_json(sub)  ##20260407
            if isinstance(sub, dict): 
                verify = encrypt_data( json_to_str(sub) )
                
                email_hash = hashlib.sha256(email.strip().lower().encode("utf-8")).hexdigest()
                url = f'https://huggingface.co/datasets/duguaynins/dataset-v20260406/resolve/main/{email_hash}/data.txt'
                url = f'https://huggingface.co/datasets/duguaynins/nins/resolve/main/{email_hash}/data.txt'
                
                if verify==requests.get(url).text: pass
                ##else: continue

            print(type(sub), sub)
            webpush(
                data=data,  ##json_to_str(payload)
                subscription_info=sub,  ##str_to_json(sub)  ##subscription
                vapid_private_key=VAPID_PRIVATE,
                vapid_claims=VAPID_CLAIMS
            )
            print("推播成功:", sub)  ##20260402
            continue ##return "O"
            
        except WebPushException as ex:
            print("推播失敗:", sub)  ##20260402
            print(ex)               ##20260402
            print("\n\n\n")
            continue ##return "X"

    if True:
        return f"{len(subscriptions)} * Subscribed"
        return "end."

# 呼叫函數
##send_push_notifications()
def fn_main_webpush(inputs):
    return send_push_notifications(inputs)



#==================================================
# FUNCTION: DEURL.SYS
#==================================================
import httpx
import urllib.parse
from urllib.parse import urlparse, urljoin  # 👈 加上這一行就解決了
import traceback # 👈 記得導入這個
from typing import Tuple, Dict

def expand_urls(url):
    try:
        r = requests.get(
            url,
            headers={"User-Agent": "Mozilla/5.0"},
            timeout=10
        )
        r.raise_for_status()
    except RequestException as e:
        print(f"[WARN] 無法抓取 {url} : {e}")
        return []

    html = r.text
    base_domain = urlparse(url).netloc
    urls = re.findall(r'(?:href|src)=["\'](.*?)["\']', html, re.IGNORECASE)
    
    external_urls = set()
    for u in urls:
        full_url = urljoin(url, u)  # 把相對路徑轉成完整網址
        parsed = urlparse(full_url)
        if parsed.scheme not in ("http", "https"):
            continue
        if parsed.netloc != base_domain:  # 排除同網域
            ##external_urls.add(full_url)
            external_urls.add(parsed.netloc)
            if len(external_urls) >= 10: break
    
    return list(external_urls)
    
def expand_url(url):
    try:
        with httpx.Client(
            ##transport=IPv4Only(),
            follow_redirects=True,
            timeout=10,
            max_redirects=5, # 限制跳轉次數，防止死循環
            verify=True,  # ✅ 正常開啟
            headers={
                "User-Agent": "Mozilla/5.0"
            }
        ) as client:
            r = client.get(url)
            return str(r.url)
            
    except httpx.RequestError as e:
        ##return f"Request error: {e}"
        return None
    except Exception as e:
        ##return f"Error: {e}"
        return None

def expand_url_redirectcheck_io(url: str) -> str:  ##expand_url_redirectcheck
    base_api = "https://redirectcheck.io/api/analyze"
    params = {"url": url}

    try:
        with httpx.Client(timeout=20) as client:
            resp = client.get(base_api, params=params)
            data = resp.json()
            # 取出 chain summary 中的 finalUrl
            return data.get("summary", {}).get("finalUrl", "-Unable to get the final URL-")
    except Exception as e:
        ##return f"錯誤: {e}"
        ##raise e
        return None
    
def expand_url_unshorten_me(target_url):  ##get_final_url_via_provider
    # 使用第三方 API 幫我們跑路
    # 這裡以一個簡單的網址展開服務為例
    api_url = f"https://unshorten.me/s/{target_url}"
    
    try:
        with httpx.Client(timeout=20) as client:
            resp = client.get(api_url)
            # 它會直接回傳最終網址的純文字
            final_url = resp.text.strip()
            return final_url
    except Exception as e:
        ##return f"第三方查詢失敗: {e}"
        ##raise e
        return None

def fn_main_deurl(inputs):
    print("[START]")
    urls = expand_urls(inputs)
    ##if urls: return urls

    urla = expand_url(inputs)
    if urla: return {"url":[urla], "urls":urls}
    urlb = expand_url_redirectcheck_io(inputs)
    if urlb: return {"url":[urlb], "urls":urls}
    urlc = expand_url_unshorten_me(inputs)
    if urlc: return {"url":[urlc], "urls":urls}
    print("[END]")



#==================================================
# ROUTE.SYS
#==================================================
from flask import Flask, render_template, redirect, url_for, request, jsonify, send_file
from flask_cors import CORS

app = Flask(__name__)
CORS(app) # 允許所有來源，這對測試最方便


# 404 發生時，自動導回首頁
@app.errorhandler(404)
def page_not_found(e):
    return redirect(url_for("home"))  ##def home():
    
# 正常頁面
@app.route("/")
def home():
    return render_template("index.html")

@app.route("/about")
def about():
    return render_template("about.html")

@app.route("/contact")
def contact():
    return render_template("contact.html")
    
@app.route("/MultipleEmbeddings/predict", methods=["POST"])  # 改！
@app.route("/embeddings/multiple_v1", methods=["POST"])  # 改！
def MultipleEmbeddings():  # 改！
    client = Client(src="https://duguaynins-MultipleEmbeddings-v260331.hf.space", hf_token=HF_READ)  # 初始化
    
    result = client.predict(
        **request.json,  # 將 JSON dict 展開成 fn 的參數
    	api_name="/fn_main",  # 如果知道 api_name，則 fn_index 可設定為 None
        fn_index=None,
        ##data=input_data,
    )
    
    return jsonify({"result": result})
    
@app.route('/generate/llms_v1', methods=['POST'])
def generate_llms():        
    client = Client(src="https://duguaynins-MultipleLLMs-v260401.hf.space", hf_token=HF_READ)  # 初始化
    
    result = client.predict(
        **request.json,  # 將 JSON dict 展開成 fn 的參數
        api_name="/fn_main",    # 如果知道 api_name，則 fn_index 可設定為 None
        ##fn_index=0,           # 如果知道 api_name，則 fn_index 可設定為 None
        ##data=input_data,
    )
    return jsonify({"result": result})
"""
@app.route('/generate/llms', methods=['POST'])  ##???
def generate_llms():  ##???
    inputs = request.json.get('inputs')
    result = fn_main_llms(inputs)  ##???
    return jsonify({"result": result})
"""


##{"inputs": {"system":_,"user":_,} }
##...{"user": {"source":_,"target":_,"en":_,} }

@app.route('/generate/mqm', methods=['POST'])
def generate_mqm():  ##???  ##data
    buffer = request.json.get('inputs')  
    doc = nlp(buffer.get('intermediate', 'The admin is duguaynins.'))
    dep = get_dependencies(doc)

    #data.pop(list(data.items())[1][0], None)
    #data.pop(list(data.items())[3][0], None)
    #data.pop(list(data.items())[5][0], None)
    inputs = {
        "system": f"{os.getenv('NINS_PROMPT_MQM')}\n{buffer.get('system')}",  ##source  ##list(data.items())[0][0]
        "user": f"{dep}\n{buffer.get('user')}",
        "models": "gpt-4o-mini",  ##gpt-4o-mini
    }
    result = fn_main_llms(inputs)  ##???
    
    result = str_to_json(result)  ##ing
    merge = {}
    merge['MQM'] = result
    merge['dependencies'] = dep
    merge = json_to_str(merge)  ##OK
    result = json_to_str(result)  ##ing
    return jsonify({"result": merge})  ##result


@app.route('/generate/rewrite', methods=['POST'])
def generate_rewrite():  ##???
    buffer = request.json.get('inputs')  
    inputs = {
        "system": f"{os.getenv('NINS_PROMPT_REWRITE')}\n{buffer.get('system')}",
        "user": buffer.get('user'),
        "models": buffer.get('models'),  ##gpt-4o-mini
    }
    result = fn_main_llms(inputs)  ##???

    result = str_to_json(result)  ##ing
    ##for k, v in result.items(): print(k, v)
    sl = list(result.items())[0][1]
    sl_pro = list(result.items())[1][1]
    il = list(result.items())[2][1]
    il_pro = list(result.items())[3][1]
    ##tl = list(result.items())[4][1]  ##可能目標都是英文的時候
    ##tl_pro = list(result.items())[5][1]  ##可能目標都是英文的時候

    embeddings = fn_main_ds({'a': il, 'b': sl})  ##embeddings OR labse
    result['embeddings'] = embeddings  ##embeddings OR labse

    ##dep, mqm = generate_mqm(result)  ##???
    ##result['dep'] = dep
    ##result['mqm'] = mqm

    result = json_to_str(result)  ##ing
    return jsonify({"result": result})

@app.route('/generate/translate', methods=['POST'])
def generate_translate():  ##???
    """
    buffer = request.json.get('inputs')  
    inputs = {
        "system": f"{os.getenv('NINS_PROMPT_TRANSLATE')}\n{buffer.get('system')}",
        "user": buffer.get('user'),
        "models": buffer.get('models'),  ##gpt-4o-mini
    }
    result = fn_main_llms(inputs)  ##???

    result = str_to_json(result)  ##ing
    ##for k, v in result.items(): print(k, v)
    sl = list(result.items())[0][1]
    sl_pro = list(result.items())[1][1]
    il = list(result.items())[2][1]
    il_pro = list(result.items())[3][1]
    ##tl = list(result.items())[4][1]  ##可能目標都是英文的時候
    ##tl_pro = list(result.items())[5][1]  ##可能目標都是英文的時候

    embeddings = fn_main_ds({'a': il, 'b': sl})  ##embeddings OR labse
    result['embeddings'] = embeddings  ##embeddings OR labse
    
    result = json_to_str(result)  ##ing
    return jsonify({"result": result})
    """

@app.route('/generate/llms', methods=['POST'])
@app.route('/generate/llms-vt', methods=['POST'])  ##???
def generate_llms_vt():  ##???
    inputs = request.json.get('inputs')
    result = fn_main_llms(inputs)  ##???
    return jsonify({"result": result})

@app.route("/embeddings/multiple", methods=["POST"])  # 改！
@app.route("/embeddings/multiple-vt", methods=["POST"])  # 改！
def embeddings_multiple_vt():    
    inputs = request.json.get('inputs')
    result = fn_main_ds(inputs)  ##???
    return jsonify({"result": result})    

@app.route('/webpush/ninscc', methods=['POST'])  ##???
def webpush_ninscc():  ##???
    inputs = request.json.get('inputs')
    result = fn_main_webpush(inputs)  ##???
    return jsonify({"result": result})

@app.route('/decode/url', methods=['POST'])  ##???
def decode_url():  ##???
    inputs = request.json.get('inputs')
    result = fn_main_deurl(inputs)  ##???
    return jsonify({"result": result})

@app.route('/dataset/update', methods=['POST'])  ##???
def dataset_update():  ##???
    inputs = request.json.get('inputs')
    result = fn_main_dataset(inputs)  ##???
    return jsonify({"result": result})

@app.route("/channels/ably", methods=["POST"])  # 改！  ##20260420
def channels_get():
    data = request.get_json() # VS request.json()
    email = request.args.get("email") or data.get("email")
    ##email = request.args.get("email")
    ##result = token_channel(email)  ##search_sheet
    result = search_sheet(gid=GS_ID, index="index", target="webpush", signal="tag", verify="***", value=email)
    return jsonify({"result": result})

@app.route("/realtime/ably", methods=["POST"])  # 改！
def realtime_ably():  # 改！
    return token_ably(AR_APIKEY, None)


"""
@app.route("/telegram/sendMessage", methods=["GET", "POST"])
def telegram_sendMessage():
    pass
    token = TG_TOKEN
    
    chat_id = request.values.get("chatid")
    text = request.values.get("text")
    keyboard = {}
    url = f"https://api.telegram.org/bot{token}/sendMessage?chat_id={chat_id}&text={text}"

    params = {
        "chat_id": chat_id,
        "text": text,  ##Hello from Python
        "parse_mode": "HTML",  ##Markdown, MarkdownV2, HTML
        ##"reply_markup": json.dumps(keyboard)##keyboard   # ✔ 直接丟 dict
    }

    data = None
    with httpx.Client(timeout=20) as client:
        resp = client.get(url, params=params)
        data = resp.json()
    return data

    r = requests.get(url, params=params, timeout=20)
    ##r = requests.post(url, data=params, timeout=(3, 20))
    
    print(r.json())
    return r.json()

@app.route("/telegram/sendMessage_vt", methods=["GET", "POST"])
async def send_message_vt():
    # 完美！同時支援 GET 和 POST 傳參
    token = TG_TOKEN
    chat_id = request.values.get("chatid")
    text = request.values.get("text")

    if not chat_id or not text:
        return {"ok": False, "description": "Bad Request: chatid or text is missing"}, 400

    url = f"https://api.telegram.org/bot{token}/sendMessage"
    payload = {
        "chat_id": chat_id,
        "text": text
    }
    
    try:
        # 設定較長的 timeout 應對 Hugging Face 的網路波動
        async with httpx.AsyncClient(timeout=30.0) as client:
            response = await client.post(url, data=payload)
            # 直接回傳 Telegram API 的結果，方便除錯
            return response.json()
            
    except httpx.ConnectTimeout:
        return {"ok": False, "error_code": 504, "description": "Gateway Timeout: Telegram connection timed out"}, 504
    except Exception as e:
        return {"ok": False, "description": f"Internal Error: {str(e)}"}, 500
"""


import platform
if __name__ == "__main__":
    sys = platform.system()
    print(sys)
    if sys == "Windows":
        app.run(host="0.0.0.0", port=5800)
    else:
        app.run(host="0.0.0.0", port=7860)
"""
if __name__ == "__main__":
    app.run(host="0.0.0.0", port=7860)  """