Create pipes/hix_ai_pipe.py

pipes/hix_ai_pipe.py

@@ -0,0 +1,351 @@
+"""
+title: DeepSeek R1 From Hix
+author: Jason
+description: Hix AI Chat Pipe with multi-turn conversation, storing chatId->botId in a local file.
+version: 1.0.0
+licence: MIT
+"""
+import datetime
+import os
+import json
+import random
+import uuid
+import hashlib
+import re
+import asyncio
+from typing import AsyncGenerator, Callable, Awaitable, Optional
+
+import cloudscraper
+from pydantic import BaseModel, Field
+
+# 用于从 assistant 消息中提取 chatId
+CHAT_ID_REGEX = re.compile(r"Chat ID:\s*([a-zA-Z0-9_-]+)")
+
+
+class Pipe:
+    class Valves(BaseModel):
+        API_DOMAIN: str = Field(default="https://hix.ai", description="API Domain")
+
+    def __init__(self):
+        # 1) 配置 & 常量
+        self.valves = self.Valves()
+        self.f = "nvKTEonFAll-in-One AI Writing CopilotwNLf2plwvtlcCxam"
+        self.salt = "xJ7fTJBgQ55/9r|"
+        self.sse_prefix = "data: "
+
+        # 2) 会话属性 (同进程多轮)
+        self.scraper = None
+        self.device_id = None
+        self.device_number = None
+        self.csrf_token = None
+        self.logged_in = False  # 是否已匿名登录
+        # 英文/中文词汇示例，可根据需要自行扩充
+        self.ENGLISH_WORDS = [
+            "Alpha", "Beta", "Gamma", "Deep", "Seek", "Magic", "Galaxy",
+            "Hello", "Future", "Project", "Vision", "Global", "Spark",
+            "Harmony", "Creative", "Insight", "Innovate", "Design", "Space",
+        ]
+
+        self.CHINESE_WORDS = [
+            "世界", "探索", "未来", "灵感", "空间", "科技", "思考",
+            "聚合", "演示", "测试", "案例", "交互", "创造", "变革",
+            "创新", "设计", "实验", "方案", "方向", "策略", "策划"
+        ]
+        # 3) 多子模型列表
+        self.sub_pipes = [
+            {"id": "73", "bot_id": 73, "name": "HIX Chat"},
+            {"id": "85426", "bot_id": 85426, "name": "DeepSeek-R1"},
+            # {"id": "85427", "bot_id": 85427, "name": "DeepSeek-V3"},
+            # {"id": "85422", "bot_id": 85422, "name": "Claude 3.5 Haiku"},
+            # {"id": "1181", "bot_id": 1181, "name": "OpenAI o1"},
+            # {"id": "85424", "bot_id": 85424, "name": "OpenAI o3-mini"},
+            # {"id": "1182", "bot_id": 1182, "name": "OpenAI o1-mini"},
+            # {"id": "85428", "bot_id": 85428, "name": "Grok-2"},
+            # {"id": "5", "bot_id": 5, "name": "GPT-4o"},
+            # {"id": "6", "bot_id": 6, "name": "GPT-4o 128K"},
+            # {"id": "86", "bot_id": 86, "name": "GPT-4o mini"},
+            # {"id": "8", "bot_id": 8, "name": "GPT-4 Turbo"},
+            # {"id": "9", "bot_id": 9, "name": "GPT-4 Turbo 128K"},
+            # {"id": "10", "bot_id": 10, "name": "GPT-4"},
+            # {"id": "42", "bot_id": 42, "name": "Claude"},
+            # {"id": "50", "bot_id": 50, "name": "Claude 3.5 Sonnet v2"},
+            # {"id": "52", "bot_id": 52, "name": "Claude 3 Haiku"},
+            # {"id": "53", "bot_id": 53, "name": "Claude 3 Opus"},
+            # {"id": "85423", "bot_id": 85423, "name": "Claude 3.5 Haiku 200K"},
+            # {"id": "54", "bot_id": 54, "name": "Claude 3.5 Sonnet v2 200K"},
+            # {"id": "55", "bot_id": 55, "name": "Claude 3 Sonnet 200K"},
+            # {"id": "56", "bot_id": 56, "name": "Claude 3 Haiku 200K"},
+            # {"id": "57", "bot_id": 57, "name": "Claude 3 Opus 200K"},
+            # {"id": "59", "bot_id": 59, "name": "Gemini 1.5 Flash"},
+            # {"id": "60", "bot_id": 60, "name": "Gemini 1.5 Pro"},
+            # {"id": "62", "bot_id": 62, "name": "Gemini 1.5 Flash 128K"},
+            # {"id": "63", "bot_id": 63, "name": "Gemini 1.5 Pro 128K"},
+            # {"id": "65", "bot_id": 65, "name": "Gemini 1.5 Flash 1M"},
+            # {"id": "67", "bot_id": 67, "name": "Gemini 1.5 Pro 1M"},
+            # {"id": "2", "bot_id": 2, "name": "ChatGPT"},
+            # {"id": "3", "bot_id": 3, "name": "GPT-3.5 Turbo"},
+            # {"id": "4", "bot_id": 4, "name": "GPT-3.5 Turbo 16K"},
+            # {"id": "44", "bot_id": 44, "name": "Claude Instant 100K"},
+            # {"id": "45", "bot_id": 45, "name": "Claude 2"},
+            # {"id": "47", "bot_id": 47, "name": "Claude 2 100K"},
+            # {"id": "49", "bot_id": 49, "name": "Claude 2.1 200K"},
+            # {"id": "51", "bot_id": 51, "name": "Claude 3 Sonnet"},
+            # {"id": "83", "bot_id": 83, "name": "Gemini"},
+            {"id": "58", "bot_id": 58, "name": "Gemini 1.0 Pro"},
+        ]
+
+        # 4) chatId->botId 映射：持久化到文件
+        self.map_file = "/tmp/chat_map.json"
+        self.chat_map = {}
+        if os.path.exists(self.map_file):
+            try:
+                with open(self.map_file, "r", encoding="utf-8") as f:
+                    self.chat_map = json.load(f)  # {chat_id: bot_id}
+            except:
+                self.chat_map = {}
+
+    def pipes(self):
+        """
+        返回子模型列表到 OpenWebUI，如:
+        [
+          {"id": "73", "name": "HIX Chat (botId=73)"},
+          {"id": "85427", "name": "DeepSeek-V3 (botId=85427)"}
+          ...
+        ]
+        """
+        results = []
+        for sp in self.sub_pipes:
+            results.append({"id": sp["id"], "name": f"{sp['name']} (botId={sp['id']})"})
+        return results
+
+    async def pipe(
+        self, body: dict, __event_emitter__: Callable[[dict], Awaitable[None]] = None
+    ) -> AsyncGenerator[str, None]:
+        """
+        主接口:
+          1) 解析 body,获取 model => sub_id
+          2) 从 messages 中提取 user问题 & chatId
+          3) 在后台执行 _exec_chat_flow => yield SSE
+        """
+        chat_id = None
+        question = ""
+        sub_id = None
+
+        # A) 解析 "model": e.g. "hix-r1.85427"
+        if "model" in body:
+            model_id = body["model"]
+            if "." in model_id:
+                _, sub_id = model_id.split(".", 1)  # "85427"
+            else:
+                sub_id = model_id
+        else:
+            # 若没给 => 默认
+            if self.sub_pipes:
+                sub_id = self.sub_pipes[0]["id"]
+            else:
+                sub_id = "85427"  # fallback
+
+        # B) 解析 user问题 + chatId
+        if "messages" in body and isinstance(body["messages"], list):
+            for msg in body["messages"]:
+                if msg["role"] == "assistant":
+                    match = CHAT_ID_REGEX.search(msg.get("content", ""))
+                    if match:
+                        chat_id = match.group(1)
+                elif msg["role"] == "user":
+                    question = msg.get("content", "")
+
+        if not question:
+            yield json.dumps({"error": "No user question found"}, ensure_ascii=False)
+            return
+
+        if __event_emitter__:
+            await __event_emitter__(
+                {
+                    "type": "status",
+                    "data": {"description": f"Asking: {question[:30]}", "done": False},
+                }
+            )
+
+        try:
+            # 后台线程 => 同步逻辑
+            for token in await asyncio.to_thread(
+                self._exec_chat_flow, question, chat_id, sub_id
+            ):
+                yield token
+        except Exception as e:
+            yield json.dumps({"error": str(e)}, ensure_ascii=False)
+
+        if __event_emitter__:
+            await __event_emitter__(
+                {
+                    "type": "status",
+                    "data": {"description": "Answer complete", "done": True},
+                }
+            )
+
+    # ========== 核心同步逻辑 ==========
+
+    def _exec_chat_flow(self, question: str, chat_id: Optional[str], sub_id: str):
+        """
+        1) 若 chat_id 在 chat_map => 直接用该 botId, 忽略 sub_id
+        2) 若没有 => sub_id => botId => create_chat => 存映射
+        3) init + login => SSE => yield => 末尾 ***\n Chat ID
+        """
+        # 1) 查找 chatId 对应 botId
+        if chat_id and (chat_id in self.chat_map):
+            # 已记录 => 用 map 里的 botId
+            chosen_bot_id = self.chat_map[chat_id]
+        else:
+            # 未记录 => 第一次 => sub_id => botId
+            chosen_bot_id = 85427  # fallback
+            sub_pipe = next((sp for sp in self.sub_pipes if sp["id"] == sub_id), None)
+            if sub_pipe:
+                chosen_bot_id = sub_pipe["bot_id"]
+
+        # 2) 初始化 scraper if needed
+        if not self.scraper:
+            self.scraper = cloudscraper.create_scraper(
+                browser={"browser": "chrome", "platform": "windows", "mobile": False}
+            )
+        if not self.device_id or not self.device_number:
+            self.device_id, self.device_number = self._gen_device_info()
+
+        # 若还没登录 => anonymous_login
+        if not self.logged_in:
+            self.csrf_token = self._get_csrf_token(self.scraper)
+            self._anonymous_login(
+                self.scraper, self.device_id, self.device_number, self.csrf_token
+            )
+            self.logged_in = True
+
+        # 3) 如果没有 chat_id => create 并记录
+        if not chat_id:
+            chat_id = self._create_chat(chosen_bot_id)
+            self.chat_map[chat_id] = chosen_bot_id
+            self._save_map()
+        else:
+            # 若 chat_id 不在 map => 说明是进程重启or首次 => 也要记录
+            if chat_id not in self.chat_map:
+                self.chat_map[chat_id] = chosen_bot_id
+                self._save_map()
+
+        # 4) SSE
+        yield from self._sse_chat(chat_id, question)
+
+        # 5) 末尾
+        yield "\n\n***\n"
+        yield f"Chat ID: {chat_id}\n"
+
+    # ============ 具体函数 ============
+    def _gen_device_info(self):
+        random_uuid = uuid.uuid4().hex.encode("utf-8")
+        device_id = hashlib.md5(random_uuid).hexdigest()
+
+        f_md5 = hashlib.md5(self.f.encode("utf-8")).hexdigest()
+        e_str = f_md5 + device_id + self.salt
+        device_number = hashlib.sha256(e_str.encode("utf-8")).hexdigest()
+        return device_id, device_number
+
+    def _get_csrf_token(self, scraper):
+        url = f"{self.valves.API_DOMAIN}/api/auth/csrf"
+        resp = scraper.get(url, allow_redirects=True)
+        resp.raise_for_status()
+        return resp.json().get("csrfToken")
+
+    def _anonymous_login(self, scraper, device_id, device_number, csrf_token):
+        url = f"{self.valves.API_DOMAIN}/api/auth/callback/anonymous-user"
+        form_data = {
+            "redirect": "false",
+            "version": "v1",
+            "deviceId": device_id,
+            "deviceNumber": device_number,
+            "csrfToken": csrf_token,
+            "callbackUrl": "https://hix.ai",
+            "json": True,
+        }
+        headers = {
+            "Content-Type": "application/x-www-form-urlencoded",
+            "Referer": "https://hix.ai/",
+        }
+        resp = scraper.post(url, data=form_data, headers=headers)
+        resp.raise_for_status()
+
+
+
+    def generate_random_title(self):
+        """生成一个随机的中英文混合标题，并附带当前时间（YYYY-MM-DD HH:MM）。"""
+
+        # 1) 随机选取英文单词 (1~2个，示例随机)
+        english_part_count = random.randint(1, 2)
+        english_part = " ".join(random.choice(self.ENGLISH_WORDS) for _ in range(english_part_count))
+
+        # 2) 随机选取中文词汇 (1~2个，拼接成一句话)
+        chinese_part_count = random.randint(1, 2)
+        chinese_part = "".join(random.choice(self.CHINESE_WORDS) for _ in range(chinese_part_count))
+
+        # 3) 当前时间
+        now = datetime.datetime.now()
+        time_str = now.strftime("%Y-%m-%d %H:%M")
+
+        # 4) 组合标题
+        # 例如: "Alpha Hello 世界未来 - 2023-08-08 14:25"
+        # 你也可以调整连接符、空格等
+        title = f"{english_part} {chinese_part} - {time_str}"
+
+        return title
+
+    def _create_chat(self, bot_id: int):
+        url = f"{self.valves.API_DOMAIN}/api/trpc/hixChat.createChat?batch=1"
+        new_title = self.generate_random_title()
+        payload = {"0": {"json": {"title": f"{new_title} for hixAI TestTeam.", "botId": bot_id}}}
+        r = self.scraper.post(url, json=payload)
+        r.raise_for_status()
+        data = r.json()
+        chat_id = data[0]["result"]["data"]["json"]["id"]
+        return chat_id
+
+    def _sse_chat(self, chat_id: str, question: str):
+        url = f"{self.valves.API_DOMAIN}/api/hix/chat"
+        payload = {"chatId": chat_id, "question": question, "fileUrl": ""}
+        resp = self.scraper.post(url, json=payload, stream=True)
+        resp.raise_for_status()
+
+        for line_bytes in resp.iter_lines():
+            if not line_bytes:
+                continue
+            line = line_bytes.decode("utf-8", errors="ignore").strip()
+            if not line.startswith(self.sse_prefix):
+                continue
+
+            json_str = line[len(self.sse_prefix) :].strip()
+            if not json_str:
+                continue
+
+            try:
+                data = json.loads(json_str)
+            except json.JSONDecodeError:
+                continue
+
+            if isinstance(data, list):
+                for item in data:
+                    if isinstance(item, str):
+                        yield item
+                    elif isinstance(item, dict) and "content" in item:
+                        yield item["content"]
+                    else:
+                        yield str(item)
+            elif isinstance(data, dict):
+                content = data.get("content")
+                if content:
+                    yield content
+            elif isinstance(data, str):
+                yield data
+            else:
+                yield str(data)
+
+    # ---------- 持久化 chat_map 到文件 ----------
+    def _save_map(self):
+        map_file = "/tmp/chat_map.json"  # 同 self.map_file
+        with open(map_file, "w", encoding="utf-8") as f:
+            json.dump(self.chat_map, f, ensure_ascii=False)