Commit e0f3eda

@@ -0,0 +1,46 @@
+---
+name: baseimg
+
+on:
+  workflow_dispatch:
+  push:
+    paths:
+      - docker/base.Dockerfile
+      - .github/workflows/baseimg.yml
+      - uv.lock
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
+permissions: write-all
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@main
+
+      - name: Set up QEMU
+        uses: docker/setup-qemu-action@master
+
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@master
+      - name: Login to GHCR
+        uses: docker/login-action@master
+        with:
+          username: ${{ github.repository_owner }}
+          password: ${{ secrets.GITHUB_TOKEN }}
+          registry: ghcr.io
+
+      - name: Build
+        uses: docker/build-push-action@master
+        with:
+          context: .
+          file: docker/base.Dockerfile
+          cache-from: type=registry,ref=ghcr.io/${{ github.repository }}:build-cache
+          cache-to: type=registry,ref=ghcr.io/${{ github.repository }}:build-cache,mode=max
+          platforms: linux/amd64
+          push: ${{ github.ref == 'refs/heads/main' }}
+          tags: ghcr.io/${{ github.repository }}:base

@@ -0,0 +1,66 @@
+---
+name: docker
+
+on:
+  workflow_dispatch:
+  workflow_run:
+    workflows: [baseimg]
+    types:
+      - completed
+  push:
+    paths:
+      - src/**
+      - .github/workflows/docker.yml
+      - docker/Dockerfile
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
+permissions: write-all
+
+jobs:
+  build:
+    if: ${{ github.event.workflow_run.conclusion != 'failure' }}
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout
+        uses: actions/checkout@main
+
+      - name: Docker meta
+        id: meta
+        uses: docker/metadata-action@master
+        with:
+          images: ghcr.io/${{ github.repository }}
+          tags: |
+            type=ref,event=branch
+            type=ref,event=pr
+            type=sha,prefix=,format=long
+            type=semver,pattern=v{{version}}
+            type=semver,pattern=v{{major}}.{{minor}}
+            type=semver,pattern=v{{major}}
+            type=raw,value=latest,enable={{is_default_branch}}
+
+      - name: Set up QEMU
+        uses: docker/setup-qemu-action@master
+
+      - name: Set up Docker Buildx
+        uses: docker/setup-buildx-action@master
+      - name: Login to GHCR
+        uses: docker/login-action@master
+        with:
+          username: ${{ github.repository_owner }}
+          password: ${{ secrets.GITHUB_TOKEN }}
+          registry: ghcr.io
+
+      - name: Build
+        uses: docker/build-push-action@master
+        with:
+          context: .
+          file: docker/Dockerfile
+          build-args: |
+            IMAGE_NAME=ghcr.io/${{ github.repository }}
+          platforms: linux/amd64
+          push: ${{ github.ref == 'refs/heads/main' }}
+          tags: ${{ steps.meta.outputs.tags }}
+          labels: ${{ steps.meta.outputs.labels }}

@@ -0,0 +1,13 @@
+---
+version: 2
+
+updates:
+  - package-ecosystem: pip
+    directory: /
+    schedule:
+      interval: monthly
+
+  - package-ecosystem: docker
+    directory: /docker
+    schedule:
+      interval: monthly

@@ -0,0 +1,27 @@
+FROM ghcr.io/benny-dou/ffmpeg:latest@sha256:f184d2518e6e5a0986393420a42ef965dd03f51f083635951545acf0b3dcba5d AS ffmpeg
+FROM shinsenter/s6-overlay:latest@sha256:048acd13b6221dd1eea0d6677bca7f463b86474a5257f4154582884098a7a4ea AS s6
+FROM python:3.13-slim@sha256:23a81be7b258c8f516f7a60e80943cace4350deb8204cf107c7993e343610d47 AS python
+
+FROM python AS venv
+RUN --mount=type=bind,source=uv.lock,target=uv.lock \
+    --mount=type=bind,source=pyproject.toml,target=pyproject.toml \
+    --mount=from=ghcr.io/astral-sh/uv:0.5,source=/uv,target=/bin/uv \
+    uv venv --relocatable --no-python-downloads --no-cache && \
+    uv sync --frozen --no-dev --no-python-downloads --no-cache --no-install-project --no-editable
+COPY --from=ffmpeg /ffmpeg /.venv/bin/ffmpeg
+COPY --from=ffmpeg /ffprobe /.venv/bin/ffprobe
+
+FROM python
+COPY --link --from=s6 / /
+
+## add PUID and PGID support
+COPY docker/fix-permission /etc/cont-init.d/10-adduser
+# copy venv
+COPY --from=venv /.venv /venv
+
+RUN groupadd --gid 1000 abc && \
+    useradd -u 1000 -g 1000 --create-home -d /app -s /bin/false abc
+
+ENV PATH=/command:/venv/bin:$PATH
+# important: sets s6-overlay entrypoint
+ENTRYPOINT ["/init"]

@@ -0,0 +1,5 @@
+ARG IMAGE_NAME
+FROM $IMAGE_NAME:base
+COPY --chown=abc src /app
+WORKDIR /app
+CMD ["with-contenv", "s6-setuidgid", "abc", "python", "/app/main.py"]

@@ -0,0 +1,16 @@
+#!/command/with-contenv bash
+
+PUID=${PUID:-1000}
+PGID=${PGID:-1000}
+
+groupmod -o -g "$PGID" abc
+usermod -o -u "$PUID" abc
+
+
+echo "
+User uid:    $(id -u abc)
+User gid:    $(id -g abc)
+-------------------------------------
+"
+
+chown abc:abc /app

@@ -0,0 +1,35 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+"""此脚本用于生成 pyrogram 的 session string.
+
+APP ID 和 APP HASH 请在此链接申请: https://my.telegram.org/apps
+
+如果是个人账号 (用户账号), 只需要提供 APP ID 和 APP HASH
+如果是机器人账号, 还需要提供 BOT TOKEN. (BOT TOKEN可以在 @BotFather 那里获取)
+"""
+
+import argparse
+from pathlib import Path
+
+from pyrogram.client import Client
+
+parser = argparse.ArgumentParser()
+parser.add_argument("--appid", type=str, required=True, help="APP ID")
+parser.add_argument("--apphash", type=str, required=True, help="APP HASH")
+parser.add_argument("--bot-token", type=str, required=False, help="BOT TOKEN (Optional, for bot account only)")
+args = parser.parse_args()
+
+proxy = {
+    "scheme": "socks5",  # "socks4", "socks5" or "http"
+    "hostname": "127.0.0.1",
+    "port": 7890,
+    # "username": "username",
+    # "password": "password",
+}
+session_path = Path(__file__).parent / "account.session_string"
+session_path.unlink(missing_ok=True)
+with Client("account", api_id=int(args.appid), api_hash=args.apphash, bot_token=args.bot_token, in_memory=True, proxy=proxy) as app:
+    session_str = app.export_session_string()  # type: ignore
+    session_path.write_text(session_str)
+
+print(session_path.read_text())

@@ -0,0 +1,169 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+
+# From: https://github.com/TencentCloud/tencentcloud-speech-sdk-python
+import base64
+import hashlib
+import hmac
+import time
+from json import JSONDecodeError
+
+from httpx import AsyncClient, RequestError
+from loguru import logger
+
+from config import PROXY
+
+
+# 录音识别极速版
+class FlashRecognitionRequest:
+    def __init__(self, engine_type):
+        self.engine_type = engine_type
+        self.speaker_diarization = 0
+        self.hotword_id = ""
+        self.hotword_list = ""
+        self.input_sample_rate = 0
+        self.customization_id = ""
+        self.filter_dirty = 0
+        self.filter_modal = 0
+        self.filter_punc = 0
+        self.convert_num_mode = 1
+        self.word_info = 0
+        self.voice_format = ""
+        self.first_channel_only = 1
+        self.reinforce_hotword = 0
+        self.sentence_max_length = 0
+
+    def set_first_channel_only(self, first_channel_only):
+        self.first_channel_only = first_channel_only
+
+    def set_speaker_diarization(self, speaker_diarization):
+        self.speaker_diarization = speaker_diarization
+
+    def set_filter_dirty(self, filter_dirty):
+        self.filter_dirty = filter_dirty
+
+    def set_filter_modal(self, filter_modal):
+        self.filter_modal = filter_modal
+
+    def set_filter_punc(self, filter_punc):
+        self.filter_punc = filter_punc
+
+    def set_convert_num_mode(self, convert_num_mode):
+        self.convert_num_mode = convert_num_mode
+
+    def set_word_info(self, word_info):
+        self.word_info = word_info
+
+    def set_hotword_id(self, hotword_id):
+        self.hotword_id = hotword_id
+
+    def set_hotword_list(self, hotword_list):
+        self.hotword_list = hotword_list
+
+    def set_input_sample_rate(self, input_sample_rate):
+        self.input_sample_rate = input_sample_rate
+
+    def set_customization_id(self, customization_id):
+        self.customization_id = customization_id
+
+    def set_voice_format(self, voice_format):
+        self.voice_format = voice_format
+
+    def set_sentence_max_length(self, sentence_max_length):
+        self.sentence_max_length = sentence_max_length
+
+    def set_reinforce_hotword(self, reinforce_hotword):
+        self.reinforce_hotword = reinforce_hotword
+
+
+class FlashRecognizer:
+    def __init__(self, appid, credential):
+        self.credential = credential
+        self.appid = appid
+
+    def _format_sign_string(self, param):
+        signstr = "POSTasr.cloud.tencent.com/asr/flash/v1/"
+        for t in param:
+            if "appid" in t:
+                signstr += str(t[1])
+                break
+        signstr += "?"
+        for x in param:
+            tmp = x
+            if "appid" in x:
+                continue
+            for t in tmp:
+                signstr += str(t)
+                signstr += "="
+            signstr = signstr[:-1]
+            signstr += "&"
+        return signstr[:-1]
+
+    def _build_header(self):
+        header = {}
+        header["Host"] = "asr.cloud.tencent.com"
+        return header
+
+    def _sign(self, signstr, secret_key):
+        hmacstr = hmac.new(secret_key.encode("utf-8"), signstr.encode("utf-8"), hashlib.sha1).digest()
+        s = base64.b64encode(hmacstr)
+        return s.decode("utf-8")
+
+    def _build_req_with_signature(self, secret_key, params, header):
+        query = sorted(params.items(), key=lambda d: d[0])
+        signstr = self._format_sign_string(query)
+        signature = self._sign(signstr, secret_key)
+        header["authorization"] = signature
+        requrl = "https://"
+        requrl += signstr[4::]
+        return requrl
+
+    def _create_query_arr(self, req):
+        query_arr = {}
+        query_arr["appid"] = self.appid
+        query_arr["secretid"] = self.credential.secret_id
+        query_arr["timestamp"] = str(int(time.time()))
+        query_arr["engine_type"] = req.engine_type
+        query_arr["voice_format"] = req.voice_format
+        query_arr["speaker_diarization"] = req.speaker_diarization
+        if req.hotword_id != "":
+            query_arr["hotword_id"] = req.hotword_id
+        if req.hotword_list != "":
+            query_arr["hotword_list"] = req.hotword_list
+        if req.input_sample_rate != 0:
+            query_arr["input_sample_rate"] = req.input_sample_rate
+        query_arr["customization_id"] = req.customization_id
+        query_arr["filter_dirty"] = req.filter_dirty
+        query_arr["filter_modal"] = req.filter_modal
+        query_arr["filter_punc"] = req.filter_punc
+        query_arr["convert_num_mode"] = req.convert_num_mode
+        query_arr["word_info"] = req.word_info
+        query_arr["first_channel_only"] = req.first_channel_only
+        query_arr["reinforce_hotword"] = req.reinforce_hotword
+        query_arr["sentence_max_length"] = req.sentence_max_length
+        return query_arr
+
+    async def recognize(self, req, data) -> dict:
+        header = self._build_header()
+        query_arr = self._create_query_arr(req)
+        req_url = self._build_req_with_signature(self.credential.secret_key, query_arr, header)
+        async with AsyncClient(http2=True, proxy=PROXY.TENCENT, follow_redirects=True) as hx:
+            try:
+                resp = await hx.post(req_url, headers=header, data=data, timeout=30)
+                resp.raise_for_status()
+                if resp.json().get("code") != 0:
+                    logger.warning(f"ASR failed: {resp.json()}")
+            except JSONDecodeError:
+                logger.error(f"ASR Json decode failed: {resp}")
+            except RequestError as exc:
+                logger.warning(f"ASR failed while requesting {exc.request.url!r}.")
+            except Exception as e:
+                logger.warning(f"ASR failed: {e}")
+        return resp.json()
+
+
+class Credential:
+    def __init__(self, secret_id, secret_key, token=""):
+        self.secret_id = secret_id
+        self.secret_key = secret_key
+        self.token = token

@@ -0,0 +1,216 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+import re
+from pathlib import Path
+
+from loguru import logger
+from pyrogram.client import Client
+from pyrogram.types import Message
+
+from asr.tecent_asr import Credential, FlashRecognitionRequest, FlashRecognizer
+from config import ASR_MAX_DURATION, ENABLE, PREFIX, TOKEN, cache
+from message_utils import modify_progress, send2tg
+from multimedia import convert_to_audio, parse_media_info
+
+# ruff: noqa: RUF001
+
+# https://cloud.tencent.com/document/product/1093/52097
+HELP = f"""🗣**语音转文字**
+使用说明: 以 `{PREFIX.ASR}` 回复包含音频的消息 (如语音, 视频, 音乐)
+默认可以识别普通话、粤语、英语三种语言。
+识别其他语种可在`{PREFIX.ASR}`后加上语种代码, 如:
+以`{PREFIX.ASR} ja`回复音频消息识别日语
+以`{PREFIX.ASR} fr`回复音频消息识别法语
+
+**目前支持以下语种:**
+fy: 多种方言, 上海话、四川话、武汉话、贵阳话、昆明话、西安话、郑州话、太原话、兰州话、银川话、西宁话、南京话、合肥话、南昌话、长沙话、苏州话、杭州话、济南话、天津话、石家庄话、黑龙江话、吉林话、辽宁话
+ja: 日语
+ko: 韩语
+vi: 越南语
+ms: 马来语
+id: 印度尼西亚语
+fil: 菲律宾语
+th: 泰语
+pt: 葡萄牙语
+tr: 土耳其语
+ar: 阿拉伯语
+es: 西班牙语
+hi: 印地语
+fr: 法语
+de: 德语
+"""
+
+ENGINE_MAP = {
+    "16k_zh-PY": "中英粤",
+    "16k_fy": "多种方言, 上海话、四川话、武汉话、贵阳话、昆明话、西安话、郑州话、太原话、兰州话、银川话、西宁话、南京话、合肥话、南昌话、长沙话、苏州话、杭州话、济南话、天津话、石家庄话、黑龙江话、吉林话、辽宁话",
+    "16k_ja": "日语",
+    "16k_ko": "韩语",
+    "16k_vi": "越南语",
+    "16k_ms": "马来语",
+    "16k_id": "印度尼西亚语",
+    "16k_fil": "菲律宾语",
+    "16k_th": "泰语",
+    "16k_pt": "葡萄牙语",
+    "16k_tr": "土耳其语",
+    "16k_ar": "阿拉伯语",
+    "16k_es": "西班牙语",
+    "16k_hi": "印地语",
+    "16k_fr": "法语",
+    "16k_de": "德语",
+}
+
+
+@cache.memoize(ttl=30)
+async def voice_to_text(
+    client: Client,
+    message: Message,
+    *,
+    asr_need_prefix: bool | None = None,
+    asr_skip_voice: bool | None = None,
+    asr_skip_audio: bool | None = None,
+    asr_skip_video: bool | None = None,
+    **kwargs,
+) -> None:
+    """Voice, audio, video message to text.
+
+    By default, "/asr" prefix is needed in in Group & Channel & Bot chats to trigger this function.
+    In private chat, no need to add "/asr" prefix for voice message, but the video & audio message still need it.
+
+    Args:
+        client (Client): The Pyrogram client.
+        message (Message): The trigger message object.
+        asr_need_prefix (bool, optional): If True, must use "/asr" prefix to reply a audio message.
+        asr_skip_voice (bool, optional): If True, skip voice message.
+        asr_skip_audio (bool, optional): If True, skip audio message.
+        asr_skip_video (bool, optional): If True, skip video message.
+    """
+    # send docs if message == "/asr", without reply
+    if str(message.text).lower().strip() == PREFIX.ASR and not message.reply_to_message:
+        await send2tg(client, message, texts=HELP, **kwargs)
+        return
+
+    if not (trigger_message := get_trigger_message(message, asr_need_prefix, asr_skip_voice, asr_skip_audio, asr_skip_video)):
+        return
+
+    asr_engine = "16k_zh-PY"  # default: 中英粤
+    if matched := re.match(r"/asr\s+([^.。,，/\s]+)", str(message.text)):  # /asr yue
+        asr_engine = f"16k_{matched.group(1)}"
+    asr_engine = asr_engine.replace("16k_fy", "16k_zh_dialect")  # fix dialect engine code
+
+    msg = f"Recieved {trigger_message.media.name} message, start recognizing by {ENGINE_MAP.get(asr_engine, 'Unknown')}..."
+    logger.info(msg)
+    if kwargs.get("show_progress"):
+        res = await send2tg(client, message, texts=msg, **kwargs)
+        kwargs["progress"] = res[0]
+    if asr_engine not in ENGINE_MAP:
+        await modify_progress(text=f"Unsupported ASR engine: {asr_engine}", force_update=True, **kwargs)
+        return
+    voice_format = ""
+    path: str | Path = await trigger_message.download()  # type: ignore
+    if trigger_message.media.name == "VOICE":  # audio/ogg
+        voice_format = str(trigger_message.voice.mime_type).split("/")[-1]  # set voice format
+    elif trigger_message.media.name in ["AUDIO", "VIDEO"]:
+        path = convert_to_audio(path, ext="m4a")
+        voice_format = "m4a"
+
+    if not Path(path).expanduser().resolve().is_file():
+        msg = "Failed to download audio, please try again later."
+        logger.error(msg)
+        await modify_progress(text=msg, force_update=True, **kwargs)
+        return
+    path = Path(path).expanduser().resolve()
+
+    if not voice_format:
+        voice_format = path.suffix.removeprefix(".")
+
+    # fix format code
+    if voice_format in ["oga", "ogg", "opus"]:
+        voice_format = "ogg-opus"
+    if voice_format == "mp4":
+        voice_format = "m4a"
+    if voice_format not in ["m4a", "ogg-opus", "wav", "pcm", "speex", "silk", "mp3", "aac", "amr"]:
+        msg = f"Unsupported audio format: {voice_format}"
+        logger.error(msg)
+        await modify_progress(text=msg, force_update=True)
+        return
+
+    # 音频长度
+    duration = parse_media_info(path).get("duration", 0)
+    if duration > ASR_MAX_DURATION:
+        msg = f"无法识别时长超过{ASR_MAX_DURATION}秒的音频, 当前音频时长: {duration}秒"
+        logger.error(msg)
+        await modify_progress(text=msg, force_update=True, **kwargs)
+        return
+
+    logger.debug(f"Recognizing {voice_format} audio by {asr_engine}: {path.as_posix()}")
+    credential_var = Credential(TOKEN.TENCENT_ASR_SECRET_ID, TOKEN.TENCENT_ASR_SECRET_KEY)
+    recognizer = FlashRecognizer(TOKEN.TENCENT_ASR_APPID, credential_var)
+    req = FlashRecognitionRequest(engine_type=asr_engine)
+    req.set_voice_format(voice_format)
+
+    final = ""
+    try:
+        with path.open("rb") as f:
+            resp = await recognizer.recognize(req, f.read())
+            logger.trace(resp)
+        texts = [channel.get("text", "") for channel in resp.get("flash_result", [])]
+        if len(set(texts)) == 1:  # single channel
+            final = texts[0]
+        else:
+            for cid, text in enumerate(texts):
+                final += f"通道{cid + 1}: {text}\n"
+        if final:
+            final = f"🗣语音转文字:\n{final}"
+        logger.success(f"Recognized text: {final}")
+        await send2tg(client, trigger_message, texts=final, **kwargs)
+        await modify_progress(del_status=True, **kwargs)
+    except Exception as e:
+        logger.error(f"Failed to recognize audio: {e}")
+    finally:
+        path.unlink(missing_ok=True)
+
+
+@cache.memoize(ttl=10)
+def get_trigger_message(
+    message: Message,
+    asr_need_prefix: bool | None = None,
+    asr_skip_voice: bool | None = None,
+    asr_skip_audio: bool | None = None,
+    asr_skip_video: bool | None = None,
+) -> Message | None:
+    """Check if the message is triggerable for voice recognition.
+
+    By default, "/asr" prefix is needed in in Group & Channel & Bot chats to trigger this function.
+    In private chat, no need to add "/asr" prefix for voice message, but the video & audio message still need it.
+    """
+    if not ENABLE.ASR:
+        return None
+    if message.chat.type.name in ["GROUP", "SUPERGROUP", "CHANNEL", "BOT"]:
+        asr_need_prefix = asr_need_prefix or True
+        asr_skip_voice = asr_skip_voice or False
+        asr_skip_audio = asr_skip_audio or False
+        asr_skip_audio = asr_skip_audio or False
+    else:  # private chat
+        asr_need_prefix = asr_need_prefix or False
+        asr_skip_voice = asr_skip_voice or False
+        asr_skip_audio = asr_skip_audio or False
+        asr_skip_video = asr_skip_video or False
+
+    # only trigger if msg has "/asr" prefix
+    if asr_need_prefix and not str(message.text).lower().strip().startswith(PREFIX.ASR):
+        return None
+
+    # treat the reply_to_message as the real message need to be recognized
+    trigger_msg = message.reply_to_message if asr_need_prefix or str(message.text).lower().strip().startswith(PREFIX.ASR) else message
+
+    if not trigger_msg:
+        return None
+    if not trigger_msg.media or trigger_msg.media.name not in ["VOICE", "AUDIO", "VIDEO"]:
+        return None
+    if asr_skip_voice and trigger_msg.media.name == "VOICE":
+        return None
+    if asr_skip_audio and trigger_msg.media.name == "AUDIO":
+        return None
+    if asr_skip_video and trigger_msg.media.name == "VIDEO":
+        return None
+    return trigger_msg

@@ -0,0 +1,64 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+
+import re
+
+from loguru import logger
+from pyrogram.client import Client
+from pyrogram.types import Message, ReplyParameters
+
+from config import ENABLE, PREFIX, cache
+from message_utils import send2tg
+from utils import i_am_bot
+
+BOT_NAME = "GLBetabot"
+
+
+@cache.memoize(ttl=10)
+async def ocr_to_miaomiao(client: Client, message: Message, **kwargs):
+    """Send photo to miaomiao bot for OCR.
+
+    See docs in `bridge/README.md` for details.
+    """
+    if not ENABLE.WGET:
+        return
+    # send docs if message == "/ocr", without reply
+    if str(message.text).lower().strip() == PREFIX.OCR and not message.reply_to_message:
+        await send2tg(client, message, texts=f"**图片转文字**: 以`{PREFIX.OCR}`回复图片消息即可提取文字", **kwargs)
+        return
+    msg = message.text or message.caption or ""  # /ocr args
+    if not msg.startswith(PREFIX.OCR):
+        return
+    if await i_am_bot(client):  # bot can't send message to other bots
+        return
+    # get the img file_id
+    if message.photo:
+        img = message.photo.file_id
+    elif message.reply_to_message and message.reply_to_message.photo:
+        img = message.reply_to_message.photo.file_id
+    else:
+        return
+
+    cid = kwargs.get("target_chat", message.chat.id)  # MSG-A's cid
+    mid = kwargs.get("reply_msg_id", message.id)  # MSG-A's mid
+    msg += f" \n#ID=({cid},{mid})".replace("None", "0")
+    logger.warning(f"OCR via 妙妙小工具 (@{BOT_NAME}): {msg!r}")
+    await client.send_photo(chat_id=f"@{BOT_NAME}", photo=img, caption=msg)
+
+
+@cache.memoize(ttl=10)
+async def forward_results_from_miaomiao(client: Client, message: Message):
+    """See docs in `bridge/README.md` for details."""
+    if message.from_user.username != BOT_NAME or not message.reply_to_message:
+        return
+    reply_msg = message.reply_to_message
+    reply_msg_text = reply_msg.text or reply_msg.caption or ""
+
+    # forward ocr (result should be a photo)
+    if message.photo and message.caption and reply_msg_text.startswith("/ocr") and (matched := re.search(r"#ID=\((-?\d+),(-?\d+)\)", str(reply_msg_text))):
+        target_cid = matched.group(1)  # MSG-A's cid
+        target_mid = int(matched.group(2)) if int(matched.group(2)) != 0 else None  # MSG-A's mid
+        cid = message.chat.id  # result's cid
+        mid = message.id  # result's mid
+        logger.info(f"Forwarding chat=@{BOT_NAME}, id={mid} -> chat={target_cid}, id={target_mid}")
+        await client.copy_message(chat_id=target_cid, from_chat_id=cid, message_id=mid, reply_parameters=ReplyParameters(message_id=target_mid))  # type: ignore

@@ -0,0 +1,41 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+
+import re
+
+from loguru import logger
+from pyrogram.client import Client
+from pyrogram.types import Message, ReplyParameters
+
+from config import cache
+from message_utils import parse_msg
+from utils import i_am_bot
+
+BOT_NAME = "ParsehubBot"
+
+
+@cache.memoize(ttl=10)
+async def send_to_parsehub(client: Client, message: Message, text: str, **kwargs):
+    """See docs in `bridge/README.md` for details."""
+    if await i_am_bot(client):  # bot can't send message to other bots
+        return
+    cid = kwargs.get("target_chat", message.chat.id)  # MSG-A's cid
+    mid = kwargs.get("reply_msg_id", message.id)  # MSG-A's mid
+    msg = f"#ID=({cid},{mid})\n{text}".replace("None", "0")
+    logger.warning(f"Trying parsehub (@{BOT_NAME}): {msg!r}")
+    await client.send_message(chat_id=f"@{BOT_NAME}", text=msg)
+
+
+@cache.memoize(ttl=10)
+async def forward_results_from_parsehub(client: Client, message: Message):
+    """See docs in `bridge/README.md` for details."""
+    if message.from_user.username != BOT_NAME or not message.media:
+        return
+    parse_msg(message)
+    if message.reply_to_message and (matched := re.search(r"#ID=\((-?\d+),(-?\d+)\)", str(message.reply_to_message.text))):
+        target_cid = matched.group(1)  # MSG-A's cid
+        target_mid = int(matched.group(2)) if int(matched.group(2)) != 0 else None  # MSG-A's mid
+        cid = message.chat.id  # result's cid
+        mid = message.id  # result's mid
+        logger.info(f"Forwarding chat=@{BOT_NAME}, id={mid} -> chat={target_cid}, id={target_mid}")
+        await client.copy_message(chat_id=target_cid, from_chat_id=cid, message_id=mid, reply_parameters=ReplyParameters(message_id=target_mid))  # type: ignore

@@ -0,0 +1,46 @@
+# BridgeBot
+
+Use third-party bots to finish some tasks.
+
+## Preliminary
+
+cid: chat_id
+mid: message_id
+
+## Scenario
+
+1. A user sends us a message, named MSG-A
+
+2. We first reply to MSG-A with a progress updating message, named MSG-UPDATE. (Like "Processing your request...")
+
+3. Then we construct a special message (Named MSG-B) which can trigger the function of the third-party bot (bridgebot).
+   We log the chat_id (MSG-A's cid) and message_id (MSG-A's mid) to #ID tag in the MSG-B, format: #ID=(MSG-A's cid, MSG-A's mid)
+   And log the MSG-UPDATE tag in this format: #PROGRESS=(MSG-UPDATE's cid, MSG-UPDATE's mid) to MSG-B
+
+4. Send MSG-B to bridgebot to finish the task.
+
+5. After the task is finished, the bridgebot will reply to MSG-B with a new message containing the task results, named MSG-RES.
+So, we can parse the reply message of MSG-RES (i.e. MSG-B) to get the #ID tag and MSG-UPDATE tag, and then forward MSG-RES to MSG-A.
+
+6. Finally, we delete MSG-UPDATE.
+
+For example, a user send us a douyin link which need to previewed. (MSG-A)
+
+We want to use @ParsehubBot to finish this task.
+
+First, we reply to MSG-A with a progress updating message. (MSG-UPDATE)
+
+Then we construct a special message (MSG-B) which can trigger the function of @ParsehubBot:
+
+```txt
+#ID=(-1001234455, 2385)
+#PROGRESS=(-1001234455, 2386)
+https://v.douyin.com/helloworld
+```
+
+Send MSG-B to @ParsehubBot to finish the task.
+
+After the task is finished, @ParsehubBot will reply to MSG-B with a new message containing the video, named MSG-RES
+We parse the reply message of MSG-RES (i.e. MSG-B) to get cid= -1001234455, mid= 2385, and then forward MSG-RES to MSG-A.
+
+Finally, we delete MSG-UPDATE (cid= -1001234455, mid= 2386).

@@ -0,0 +1,72 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+
+
+import re
+from pathlib import Path
+
+import puremagic
+from loguru import logger
+from pyrogram.client import Client
+from pyrogram.types import Message, ReplyParameters
+
+from config import ENABLE, MAX_FILE_BYTES, PREFIX
+from message_utils import modify_progress, send2tg
+from multimedia import is_valid_video, validate_img
+from networking import download_file
+from utils import https_url, readable_size
+
+
+async def download_url_in_message(client: Client, message: Message, **kwargs):
+    """Download the url from the message."""
+    if not ENABLE.WGET:
+        return
+    if not str(message.text).strip().lower().startswith(PREFIX.WGET):
+        return
+    target_chat = kwargs["target_chat"] if kwargs.get("target_chat") else message.chat.id
+    reply_msg_id = kwargs.get("reply_msg_id", 0)
+    if reply_msg_id == 0:
+        reply_to = message.id
+    elif reply_msg_id == -1:
+        reply_to = None
+    else:
+        reply_to = reply_msg_id
+    reply_parameters = ReplyParameters(message_id=reply_to)  # type: ignore
+
+    regex = r"(?i)\b((?:https?://|www\d{0,3}[.]|[a-z0-9.\-]+[.][a-z]{2,4}/)(?:[^\s()<>]+|\(([^\s()<>]+|(\([^\s()<>]+\)))*\))+(?:\(([^\s()<>]+|(\([^\s()<>]+\)))*\)|[^\s`!()\[\]{};:'\".,<>?«»“”‘’]))"  # noqa: RUF001
+    if matched := re.findall(regex, message.text):
+        url = https_url(matched[0][0])
+        logger.debug(f"URL found from message text: {url}")
+
+    msg = f"⏬开始下载:\n{url}"
+    if kwargs.get("show_progress"):
+        res = await send2tg(client, message, texts=msg, **kwargs)
+        kwargs["progress"] = res[0]
+    success = False
+    try:
+        path = await download_file(url, workers_proxy=True, **kwargs)
+        suffix = puremagic.from_file(path)  # guess the file type
+        if Path(path).suffix != suffix:
+            Path(path).rename(Path(path).with_suffix(suffix))
+            path = Path(path).with_suffix(suffix)
+        if img := validate_img(path, delete=False):
+            await modify_progress(text=f"🖼图片下载成功: {readable_size(path=img)}", force_update=True, **kwargs)
+            success = await send2tg(client, message, target_chat, reply_msg_id, texts=url, media=[{"photo": img}])
+        elif Path(path).suffix in [".m4a", ".mp3", ".wav", ".ogg", ".opus", ".flac", ".aac"]:
+            await modify_progress(text=f"🎧音频下载成功: {readable_size(path=path)}", force_update=True, **kwargs)
+            success = await client.send_audio(target_chat, Path(path).as_posix(), caption=url, reply_parameters=reply_parameters)
+        elif is_valid_video(path, delete=False):
+            await modify_progress(text=f"🎬视频下载成功: {readable_size(path=path)}", force_update=True, **kwargs)
+            success = await send2tg(client, message, target_chat, reply_msg_id, texts=url, media=[{"video": path}])
+        elif Path(path).stat().st_size < MAX_FILE_BYTES:
+            await modify_progress(text=f"💾文件下载成功: {readable_size(path=path)}", force_update=True, **kwargs)
+            success = await client.send_document(target_chat, Path(path).as_posix(), caption=url, reply_parameters=reply_parameters)
+        else:
+            await modify_progress(text=f"❌文件大小: {readable_size(path=path)} 超出限制\nTelegram只允许上传小于{round(MAX_FILE_BYTES / 1024 / 1024)}MB的文件", force_update=True, **kwargs)
+    except Exception as e:
+        logger.error(e)
+        await modify_progress(text=f"❌下载失败: {url}", force_update=True, **kwargs)
+
+    if success:
+        await modify_progress(del_status=True, **kwargs)
+        Path(path).unlink(missing_ok=True)

@@ -0,0 +1,176 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+import re
+
+# From: https://github.com/Rongronggg9/RSS-to-Telegram-Bot/blob/84ce41fd31819d83e146507d50a95e28a5b58afe/src/parsing/weibo_emojify_map.py
+COMMON = {
+    "[微笑]": "🙂",
+    "[可爱]": "😊",
+    "[太开心]": "😆",
+    "[鼓掌]": "👏",
+    "[嘻嘻]": "😁",
+    "[哈哈]": "😄",
+    "[笑cry]": "😂",
+    "[挤眼]": "😜",
+    "[馋嘴]": "😋",
+    "[黑线]": "😑",
+    "[汗]": "😓",
+    "[哼]": "😠",
+    "[怒]": "😡",
+    "[可怜]": "🥺",
+    "[失望]": "😞",
+    "[悲伤]": "😢",
+    "[泪]": "😭",
+    "[害羞]": "😳",
+    "[爱你]": "🥰",
+    "[亲亲]": "😚",
+    "[色]": "😍",
+    "[阴险]": "😏",
+    "[偷笑]": "🤭",
+    "[酷]": "😎",
+    "[并不简单]": "🧐",
+    "[思考]": "🤔",
+    "[晕]": "😵",
+    "[骷髅]": "💀",
+    "[嘘]": "🤫",
+    "[闭嘴]": "🤐",
+    "[傻眼]": "😮",
+    "[吃惊]": "😲",
+    "[吐]": "🤮",
+    "[感冒]": "😷",
+    "[生病]": "🤒",
+    "[拜拜]": "👋",
+    "[鄙视]": "🖕",
+    "[白眼]": "🙄",
+    "[抓狂]": "😖",
+    "[怒骂]": "🤬",
+    "[钱]": "🤑",
+    "[哈欠]": "🥱",
+    "[困]": "😴",
+    "[睡]": "😪",
+    "[吃瓜]": "🍉",
+    "[酸]": "🍋",
+    "[喵喵]": "🐱",
+    "[抱抱]": "🤗",
+    "[摊手]": "🤷",
+    "[跪了]": "🧎",
+    "[鲜花]": "🌹",
+    "[给你小心心]": "💝",
+    "[心]": "❤",
+    "[伤心]": "💔",
+    "[握手]": "🤝",
+    "[赞]": "👍",
+    "[good]": "👍",
+    "[弱]": "👎",
+    "[NO]": "✋",
+    "[耶]": "✌",
+    "[拳头]": "✊",
+    "[ok]": "👌",
+    "[加油]": "💪",
+    "[haha]": "🤟",
+    "[熊猫]": "🐼",
+    "[兔子]": "🐰",
+    "[猪头]": "🐷",
+    "[太阳]": "🌞",
+    "[月亮]": "🌙",
+    "[浮云]": "☁",
+    "[下雨]": "🌧",
+    "[微风]": "🍃",
+    "[围观]": "👨‍👧‍👦",
+    "[飞机]": "✈",
+    "[照相机]": "📷",
+    "[话筒]": "🎙",
+    "[蜡烛]": "🕯",
+    "[音乐]": "🎵",
+    "[可乐]": "🥤",
+    "[干杯]": "🍻",
+    "[蛋糕]": "🎂",
+    "[礼物]": "🎁",
+    "[钟]": "⏰",
+    "[肥皂]": "🧼",
+    "[绿丝带]": "🎗",
+    "[围脖]": "🧣",
+    "[圣诞老人]": "🎅",
+    "[文明遛狗]": "🐕",
+    "[最右]": " →_→ ",
+    "[五仁月饼]": "🥮",
+    "[弗莱见钱眼开]": "🤑",
+    "[棒棒糖]": "🍭",
+    "[炸鸡腿]": "🍗",
+    "[点亮平安灯]": "🏮",
+    "[点亮橙色]": "🖐",
+    "[看涨]": "📈",
+    "[看跌]": "📉",
+    "[星星]": "⭐",
+    "[空星]": "★",
+    "[全家福]": "👪",
+    "[圆月]": "🌕",
+    # customization
+    "[大笑]": "😆",
+}
+
+XHS = {
+    "[零R]": "0️⃣",
+    "[一R]": "1️⃣",
+    "[二R]": "2️⃣",
+    "[三R]": "3️⃣",
+    "[四R]": "4️⃣",
+    "[五R]": "5️⃣",
+    "[六R]": "6️⃣",
+    "[七R]": "7️⃣",
+    "[八R]": "8️⃣",
+    "[九R]": "9️⃣",
+    "[背包R]": "😏",
+    "[满月R]": "😏",
+    "[生气R]": "😡",
+    "[失望R]": "😞",
+    "[抓狂R]": "😖",
+    "[萌萌哒R]": "😊",
+    "[红色心形R]": "❤",
+    "[爆炸R]": "💥",
+    "[炸弹R]": "💣",
+}
+
+DOUYIN = {
+    "[苦涩]": "😭",
+    "[doge]": "🐶",
+    "[挖鼻]": "👃",
+    "[泪]": "😭",
+    "[眼含热泪]": "🥺",
+    "[呲牙]": "😁",
+    "[玫瑰]": "🌹",
+    "[一起加油]": "💪",
+    "[比心]": "💖",
+    "[看]": "🐶",
+    "[宕机]": "😧",
+    "[捂脸]": "🤦",
+    "[泪奔]": "😭",
+    "[尬笑]": "😅",
+    "[抱一抱]": "🤗",
+    "[打call]": "👏",
+    "[哇]": "🥰",
+}
+BILIBILI = {"[笑哭]": "🤦"}
+
+
+def emojify(text: str, platform: str = "all") -> str:
+    """Replace the text emojis with actual emojis.
+
+    [微笑]Hello[爱你] -> 🙂Hello🥰
+    """
+    if platform == "all":
+        EMOJI_MAP = COMMON | BILIBILI | XHS | DOUYIN
+    elif platform == "xhs":
+        EMOJI_MAP = COMMON | XHS
+    elif platform == "douyin":
+        EMOJI_MAP = COMMON | DOUYIN
+    elif platform == "bilibili":
+        EMOJI_MAP = COMMON | BILIBILI
+    pattern = re.compile("|".join(re.escape(k) for k in EMOJI_MAP))
+
+    return pattern.sub(lambda match: EMOJI_MAP[match.group(0)], text)
+
+
+if __name__ == "__main__":
+    text = "[微笑]Hello[爱你]"
+    print(emojify(text))

@@ -0,0 +1,78 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+from pathlib import Path
+
+from loguru import logger
+from pyrogram.client import Client
+from pyrogram.types import Message, ReplyParameters
+
+from config import PREFIX, cache
+from message_utils import modify_progress, parse_msg, send2tg
+from multimedia import convert_to_audio, parse_media_info
+
+# ruff: noqa: RUF001
+
+HELP = f"""
+🎧**视频转音频**
+使用方式:
+1. 以 `{PREFIX.AUDIO}` 回复视频消息提取出音频
+2. 发送视频时, 添加`{PREFIX.AUDIO}`文本描述会同时提取视频音频
+"""
+
+
+@cache.memoize(ttl=60)
+async def extract_audio_file(client: Client, message: Message, **kwargs) -> None:
+    """Extract audio from video message."""
+    # send docs if message == "/audio", without reply
+    if str(message.text).lower().strip() == PREFIX.AUDIO and not message.reply_to_message:
+        await send2tg(client, message, texts=HELP, **kwargs)
+        return
+
+    msg_text = message.text or message.caption or ""
+    msg_text = msg_text.lower().strip()
+
+    if not msg_text.startswith(PREFIX.AUDIO):
+        return
+
+    # 以/audio命令回复一条消息
+    if message.reply_to_message:
+        message = message.reply_to_message
+
+    if not message.media or message.media.name != "VIDEO":
+        return
+    parse_msg(message, verbose=True)
+
+    msg = "🎬收到视频消息, 开始提取🎧音频..."
+    if kwargs.get("show_progress"):
+        res = await send2tg(client, message, texts=msg, **kwargs)
+        kwargs["progress"] = res[0]
+    video: str = await message.download()  # type: ignore
+    if Path(video).expanduser().resolve().is_file():
+        await modify_progress(text="🎬视频下载完成, 提取🎧音频中...", force_update=True, **kwargs)
+    else:
+        await modify_progress(text="🎬视频下载失败, 请稍后重试...", force_update=True, **kwargs)
+        return
+    path = convert_to_audio(video, ext="m4a")
+    if not Path(path).expanduser().resolve().is_file():
+        logger.trace(f"File not found: {path}")
+        return
+    path = Path(path).expanduser().resolve()
+    if path.name.endswith(".final.m4a"):  # remove final suffix
+        path.rename(path.with_name(path.name.replace(".final.m4a", ".m4a")))
+        path = path.with_name(path.name.replace(".final.m4a", ".m4a"))
+    await modify_progress(text="🎧音频提取已完成, 开始上传...", force_update=True, **kwargs)
+    duration = parse_media_info(path).get("duration", 0)
+
+    target_chat = kwargs["target_chat"] if kwargs.get("target_chat") else message.chat.id
+    reply_msg_id = kwargs.get("reply_msg_id", 0)
+    if reply_msg_id == 0:
+        reply_to = message.id
+    elif reply_msg_id == -1:
+        reply_to = None
+    else:
+        reply_to = reply_msg_id
+    reply_parameters = ReplyParameters(message_id=reply_to)  # type: ignore
+    await client.send_audio(target_chat, Path(path).as_posix(), duration=duration, reply_parameters=reply_parameters)
+    Path(video).unlink(missing_ok=True)
+    path.unlink(missing_ok=True)
+    await modify_progress(del_status=True, **kwargs)

@@ -0,0 +1,400 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+
+
+import asyncio
+import base64
+import copy
+import re
+from pathlib import Path
+from typing import TYPE_CHECKING
+
+from loguru import logger
+from pyrogram.client import Client
+from pyrogram.types import Message
+
+from config import DOWNLOAD_DIR, ENABLE, GPT, PREFIX, PROXY, cache
+from message_utils import modify_progress, send2tg
+from multimedia import convert_to_audio
+from networking import hx_req
+
+if TYPE_CHECKING:
+    from io import BytesIO
+
+clean_name = lambda x: x.removesuffix("-mini").removesuffix("-latest").removesuffix("-turbo").removesuffix("-exp").removesuffix("-chat")
+HELP = f"""🤖**GPT对话**
+当前模型:
+- 文本模型: **{clean_name(GPT.TEXT_MODEL)}**
+- 图片模型: **{clean_name(GPT.VISION_MODEL)}**
+- 视频模型: **{clean_name(GPT.VIDEO_MODEL)}**
+- 音频模型: **{clean_name(GPT.AUDIO_MODEL)}** (30秒以内)
+
+使用说明:
+1. 在 `{PREFIX.GPT}` 后接提示词即可与GPT对话
+2. 以 `{PREFIX.GPT}` 回复消息可将其加入上下文
+3. 音频模型仅支持30秒以内的音频文件, 如需更长可以先用 `{PREFIX.ASR}` 命令转为文字后再使用 `{PREFIX.GPT}`
+"""
+BOT_TIPS = "回复此消息以继续对话"
+
+
+@cache.memoize(ttl=60)
+async def gpt_response(client: Client, message: Message, **kwargs):
+    """Get GPT response from Various API.
+
+    Args:
+        client (Client): The Pyrogram client.
+        message (Message): The trigger message object.
+    """
+    if not ENABLE.GPT:
+        return
+
+    # send docs if message == "/ai", without reply
+    texts = message.text or message.caption or ""
+    if str(texts).lower().strip() == PREFIX.GPT and not message.reply_to_message:
+        await send2tg(client, message, texts=HELP, **kwargs)
+        return
+
+    if not is_valid_conversation(message):
+        return
+
+    # cache media_group message, only process once
+    if media_group_id := message.media_group_id:
+        if cache.get(f"gpt-{message.chat.id}-{media_group_id}"):
+            return
+        cache.set(f"gpt-{message.chat.id}-{media_group_id}", "1", ttl=120)
+
+    contexts = await generate_contexts(client, message)
+    model_conf = get_model_config(contexts)
+    if model_conf["friendly_name"] == "豆包":
+        contexts = fix_doubao(contexts)
+    msg = f"🤖{model_conf['friendly_name']}: 思考中..."
+    if kwargs.get("show_progress"):
+        res = await send2tg(client, message, texts=msg, **kwargs)
+        kwargs["progress"] = res[0]
+    headers = {"authorization": f"Bearer {model_conf['key']}"}
+    if model_conf["model_type"] == "audio":
+        resp = await hx_req(model_conf["url"], "POST", headers=headers, post_json=model_conf["payload"], proxy=PROXY.GPT, check_has_kv=["output.choices"], timeout=model_conf["timeout"])
+        choices = resp.json()["output"]["choices"]
+        ai_response = choices[0].get("message", {}).get("content", [{}])[0].get("text", "")
+    else:
+        resp = await hx_req(model_conf["url"], "POST", headers=headers, post_json=model_conf["payload"], proxy=PROXY.GPT, check_has_kv=["choices"], timeout=model_conf["timeout"])
+        choices = resp.json()["choices"]
+        ai_response = choices[0].get("message", {}).get("content")
+    cleanup(contexts)
+    if not ai_response:
+        await modify_progress(text=f"🤖{model_conf['friendly_name']}未响应, 请稍后重试...", force_update=True, **kwargs)
+        await asyncio.sleep(3)
+        await modify_progress(del_status=True, **kwargs)
+        return
+    texts = f"{model_conf['bot_msg_prefix']}\n\n{ai_response}"
+    logger.debug(texts)
+    await send2tg(client, message, texts=texts, **kwargs)
+    await modify_progress(del_status=True, **kwargs)
+
+
+def get_model_config(contexts: list[dict]):
+    """Clean downloaded files.
+
+    [
+      {
+          "role": "user",
+          "content": [
+              {"type": "text", "text": "text"},
+              {"type": "image_url", "image_url": {"url": "https://server.com/dir/image.jpg"}},
+              {"type": "video_url", "video_url": {"url": "https://server.com/dir/video.mp4"}},  # 智谱
+              {"type": "video", "video": ["https://server.com/dir/1.jpg","https://server.com/dir/2.jpg","https://server.com/dir/3.jpg"]},  # 千问
+              {"audio": "https://server.com/dir/audio.mp3"},
+            ]
+      }
+    ]
+    """
+    model = GPT.TEXT_MODEL
+    model_type = "text"
+    timeout = GPT.TEXT_TIMEOUT
+    for item in contexts:
+        content = item.get("content", [])
+        for x in content:
+            if x.get("image_url"):
+                model = GPT.VISION_MODEL
+                model_type = "vision"
+                timeout = GPT.VISION_TIMEOUT
+                break
+            if x.get("video_url") or x.get("video"):
+                model = GPT.VIDEO_MODEL
+                model_type = "video"
+                timeout = GPT.VIDEO_TIMEOUT
+                break
+            if x.get("audio"):
+                model = GPT.AUDIO_MODEL
+                model_type = "audio"
+                timeout = GPT.AUDIO_TIMEOUT
+                break
+
+    common = {"model_type": model_type, "timeout": int(timeout)}
+    openai = {
+        "url": GPT.OPENAI_BASE_URL + "/chat/completions",
+        "key": GPT.OPENAI_API_KEY,
+    }
+    gemini = {
+        "url": GPT.GEMINI_BASE_URL + "/chat/completions",
+        "key": GPT.GEMINI_API_KEY,
+    }
+    hunyuan = {
+        "url": GPT.HUNYUAN_BASE_URL + "/chat/completions",
+        "key": GPT.HUNYUAN_API_KEY,
+    }
+    qwen = {
+        "url": GPT.DASHSCOPE_BASE_URL + "/chat/completions",
+        "key": GPT.DASHSCOPE_API_KEY,
+    }
+    openrouter = {
+        "url": GPT.OPENROUTER_BASE_URL + "/chat/completions",
+        "key": GPT.OPENROUTER_API_KEY,
+    }
+    glm = {
+        "url": GPT.GLM_BASE_URL + "/chat/completions",
+        "key": GPT.GLM_API_KEY,
+    }
+    doubao = {
+        "url": GPT.ARK_BASE_URL + "/chat/completions",
+        "key": GPT.ARK_API_KEY,
+    }
+    if model.startswith("gpt"):
+        openai["friendly_name"] = clean_name(model.replace("gpt", "GPT"))
+        config = common | openai
+
+    if model.startswith("gemini"):
+        gemini["friendly_name"] = clean_name(model.capitalize())
+        config = common | gemini
+
+    if model.startswith("hunyuan"):
+        hunyuan["friendly_name"] = clean_name(model.replace("hunyuan", "混元"))
+        config = common | hunyuan
+
+    if model.startswith("qwen"):
+        qwen["friendly_name"] = clean_name(model.replace("qwen", "通义千问"))
+        if model_type == "audio":  # qwen-audio
+            qwen["url"] = "https://dashscope.aliyuncs.com/api/v1/services/aigc/multimodal-generation/generation"
+        config = common | qwen
+
+    if model.startswith("deepseek"):  # via openrouter: deepseek/deepseek-chat
+        openrouter["friendly_name"] = clean_name(model.split("/")[-1].capitalize())
+        config = common | openrouter
+
+    if model.startswith("glm"):
+        glm["friendly_name"] = model.upper().removesuffix("-FLASH")
+        config = common | glm
+
+    if model.startswith("doubao"):
+        doubao["friendly_name"] = "豆包"
+        if model_type == "text":
+            model = GPT.DOUBAO_TEXT_ENTRYPOINT
+        elif model_type == "vision":
+            model = GPT.DOUBAO_VISION_ENTRYPOINT
+        config = common | doubao
+
+    config["bot_msg_prefix"] = f"🤖**{config['friendly_name']}**: ({BOT_TIPS})"
+    payload = {"model": model, "input": {"messages": contexts}} if model_type == "audio" else {"model": model, "messages": contexts, "temperature": float(GPT.TEMPERATURE)}
+    config["payload"] = payload
+    logger.trace(config)
+    return config
+
+
+async def generate_contexts(client: Client, message: Message) -> list[dict]:
+    """Generate contexts for GPT conversation.
+
+    Returns:
+        list[dict]: [
+            {
+                "role": "user or assistant",
+                "content": [
+                    {'type': 'text', 'text': 'caption this img'},
+                    {'type': 'image_url', 'image_url': {'url': 'data:image/jpeg;base64,base64_image'}},
+                    {'type': 'image_url', 'image_url': {'url': 'https://server.com/dir/image.jpg'}},
+                ]
+            }
+        ]
+    """
+    # 按时间顺序如下
+    reply_msg = copy.deepcopy(message.reply_to_message)
+    contexts = []
+    if context := await generate_single_msg_context(client, message):  # this message
+        contexts.append(context)
+    while reply_msg:
+        if context := await generate_single_msg_context(client, reply_msg):  # this message
+            contexts.append(context)
+        reply_msg = reply_msg.reply_to_message
+    contexts.reverse()
+    contexts = combine_consecutive_role_contexts(contexts)
+    return contexts[: int(GPT.HISTORY_CONTEXT)]
+
+
+def combine_consecutive_role_contexts(contexts: list[dict]) -> list[dict]:
+    """Combine consecutive user and assistant contexts into one message.
+
+    Some GPT models don't support consecutive user and assistant contexts. (e.g. Hunyuan)
+
+    Args:
+        contexts (list[dict]): [
+            {
+                "role": "user or assistant",
+                "content": [
+                    {'type': 'text', 'text': 'caption this img'},
+                    {'type': 'image_url', 'image_url': {'url': 'data:image/jpeg;base64,base64_image'}},
+                    {'type': 'image_url', 'image_url': {'url': 'https://server.com/dir/image.jpg'}},
+                ]
+            }
+        ]
+    """
+    combined_contexts = []
+    for i, msg in enumerate(contexts):
+        if i == 0:
+            combined_contexts.append(msg)
+            continue
+        if msg["role"] == combined_contexts[-1]["role"]:
+            combined_contexts[-1]["content"].extend(msg["content"])
+        else:
+            combined_contexts.append(msg)
+    return combined_contexts
+
+
+def fix_doubao(contexts: list[dict]) -> list[dict]:
+    """Fix doubao context format.
+
+    Doubao do not support this content for:
+        [{'text': 'hi', 'type': 'text'}], 'role': 'user'}]
+    It support:
+        [{'content': 'hi', 'role': 'user'}]
+
+    Args:
+        contexts (list[dict]): [
+            {
+                "role": "user or assistant",
+                "content": [
+                    {'type': 'text', 'text': 'caption this img'},
+                    {'type': 'image_url', 'image_url': {'url': 'data:image/jpeg;base64,base64_image'}},
+                    {'type': 'image_url', 'image_url': {'url': 'https://server.com/dir/image.jpg'}},
+                ]
+            }
+        ]
+    """
+    fixed_contexts = []
+    for msg in contexts:
+        if (lst := msg.get("content", [])) and all(x.get("type") == "text" for x in lst):
+            msg["content"] = "\n".join([x.get("text") for x in lst])
+            fixed_contexts.append(msg)
+    return fixed_contexts
+
+
+def is_valid_conversation(message: Message) -> bool:
+    # match commands: /ai
+    if str(message.text).strip()[:3].lower() == PREFIX.GPT:
+        return True
+    if str(message.caption).strip()[:3].lower() == PREFIX.GPT:
+        return True
+    # is replying to gpt-bot response message?
+    if not message.reply_to_message:
+        return False
+
+    reply_msg = message.reply_to_message
+    reply_text = reply_msg.text or reply_msg.caption or ""
+    return reply_text.startswith("🤖")
+
+
+async def generate_single_msg_context(client: Client, message: Message) -> dict:
+    """Generate GPT contexts for a single message (Without consider reply message).
+
+    Returns:
+    {
+        "role": "user or assistant",
+        "content": [
+            {"type": "image_url", "image_url": {"url": "https://server.com/dir/image.jpg"}},
+            {"type": "video_url", "video_url": {"url": "https://server.com/dir/video.mp4"}},  # 智谱
+            {"type": "video", "video": ["https://server.com/dir/1.jpg","https://server.com/dir/2.jpg","https://server.com/dir/3.jpg"]},  # 千问
+            {"audio": "https://server.com/dir/audio.mp3"},
+        ],
+    }
+    """
+
+    def clean_text(text: str) -> str:
+        if not text:
+            return ""
+        return re.sub(rf"(.*?){BOT_TIPS}\)", "", text.removeprefix(PREFIX.GPT)).strip()
+
+    role = "assistant" if any(BOT_TIPS in texts for texts in [str(message.text), str(message.caption)]) else "user"
+    # only text
+    if text := clean_text(message.text):
+        return {"role": role, "content": [{"type": "text", "text": text}]}
+
+    if not message.media or message.media.name not in ["PHOTO", "VOICE", "AUDIO", "VIDEO", "DOCUMENT"]:
+        return {}
+
+    # has media
+    messages = await client.get_media_group(message.chat.id, message.id) if message.media_group_id else [message]
+    media = []
+    for msg in messages:
+        try:
+            if GPT.MEDIA_FORMAT == "base64":
+                res: BytesIO = await client.download_media(msg, in_memory=True)  # type: ignore
+                logger.debug(f"Downloaded GPT media: {res.name}")
+                ext = Path(res.name).suffix.removeprefix(".").replace("jpg", "jpeg")
+                b64 = base64.b64encode(res.getvalue()).decode("utf-8")
+                if message.media.name == "PHOTO":
+                    media.append({"type": "image_url", "image_url": {"url": f"data:image/{ext};base64,{b64}"}})
+                elif message.media.name == "VIDEO":
+                    media.append({"type": "video_url", "video_url": {"url": b64}})
+                elif message.media.name == "DOCUMENT" and message.document.mime_type == "text/plain":
+                    media.append({"type": "text", "text": res.getvalue().decode("utf-8")})
+                else:
+                    logger.warning("Audio do not support base64, please use http")
+            else:
+                path: str = await client.download_media(msg)  # type: ignore
+                logger.debug(f"Downloaded GPT media: {path}")
+                if message.media.name == "PHOTO":
+                    media.append({"type": "image_url", "image_url": {"url": f"{GPT.MEDIA_SERVER}/{Path(path).name}"}})
+                elif message.media.name == "VIDEO":
+                    media.append({"type": "video_url", "video_url": {"url": f"{GPT.MEDIA_SERVER}/{Path(path).name}"}})
+                elif message.media.name in ["AUDIO", "VOICE"]:
+                    mp3 = convert_to_audio(path, ext="mp3", codec="libmp3lame")
+                    media.append({"audio": f"{GPT.MEDIA_SERVER}/{mp3.name}"})
+                elif message.media.name == "DOCUMENT" and message.document.mime_type == "text/plain":
+                    media.append({"type": "text", "text": Path(path).read_text()})
+                    Path(path).unlink(missing_ok=True)
+            if caption := msg.caption:
+                media.append({"type": "text", "text": caption})
+        except Exception as e:
+            logger.warning(f"Download image from message failed: {e}")
+            continue
+    return {"role": role, "content": media}
+
+
+def cleanup(messages: list[dict]):
+    """Clean downloaded files.
+
+    [
+      {
+          'role': 'user',
+          'content': [
+              {'type': 'text', 'text': 'text'},
+              {'type': 'image_url', 'image_url': {'url': 'https://server.com/dir/image.jpg'}},
+              {"type": "video_url", "video_url": {"url": "https://server.com/dir/video.mp4"}},  # 智谱
+              {"type": "video", "video": ["https://server.com/dir/1.jpg","https://server.com/dir/2.jpg","https://server.com/dir/3.jpg"]},  # 千问
+              {"audio": "https://server.com/dir/audio.mp3"},
+            ]
+      }
+    ]
+    """
+    for item in messages:
+        content = item.get("content", [])
+        if not isinstance(content, list):
+            continue
+        for x in content:
+            if url := x.get("image_url", {}).get("url"):
+                (Path(DOWNLOAD_DIR) / Path(url).name).unlink(missing_ok=True)
+            if url := x.get("video_url", {}).get("url"):
+                (Path(DOWNLOAD_DIR) / Path(url).name).unlink(missing_ok=True)
+            if urls := x.get("video", []):
+                for url in urls:
+                    (Path(DOWNLOAD_DIR) / Path(url).name).unlink(missing_ok=True)
+            if url := x.get("audio"):
+                (Path(DOWNLOAD_DIR) / Path(url).name).unlink(missing_ok=True)

@@ -0,0 +1,33 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+
+
+from loguru import logger
+from pyrogram.client import Client
+from pyrogram.types import Message
+
+from config import ENABLE, PREFIX
+from message_utils import parse_msg, send2tg
+from multimedia import split_long_img
+
+
+async def convert_raw_img_file(client: Client, message: Message, **kwargs):
+    if not ENABLE.RAW_IMG_CONVERT:
+        return
+    mime_type = message.document.mime_type if message.document else ""
+    if mime_type not in ["image/png", "image/jpeg", "image/heic"]:
+        return
+    try:
+        info = parse_msg(message)
+        logger.info(f"Splitting raw image: {info['file_name']}")
+        path: str = await message.download()  # type: ignore
+        logger.trace(f"Image {{info['file_name']}} downloaded to {path}")
+        photos = split_long_img(path)
+        media = [{"photo": photo.as_posix()} for photo in photos if photo.is_file()]
+        # send splits
+        if len(media) == 1:
+            await send2tg(client, message, media=media, texts=f"🔁【原图转换】已完成\n发送 **{PREFIX.MAIN}** 命令查看更多功能", **kwargs)
+        else:
+            await send2tg(client, message, media=media, texts=f"✂️【自动切原图】已完成\n发送 **{PREFIX.MAIN}** 命令查看更多功能", **kwargs)
+    except Exception as e:
+        logger.exception(e)

@@ -0,0 +1,161 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+
+
+import asyncio
+import io
+from datetime import timedelta
+
+from loguru import logger
+from pyrogram.client import Client
+from pyrogram.enums import MessageEntityType
+from pyrogram.types import Message
+from youtube_transcript_api import YouTubeTranscriptApi
+
+from config import API, ENABLE, PREFIX, PROXY, TOKEN
+from database import cache
+from message_utils import modify_progress, send2tg
+from networking import hx_req, match_social_media_link
+
+HELP = f"""📃**提取字幕**
+使用说明:
+1. `{PREFIX.SUBTITLE} URL` 下载该链接的字幕
+2. 以 `{PREFIX.SUBTITLE}` 回复消息可下载消息中链接的字幕
+
+当前只支持YouTube
+"""
+
+
+@cache.memoize(ttl=30)
+async def get_subtitle(client: Client, message: Message, **kwargs):
+    """Get YouTube Subtitle."""
+    if not ENABLE.SUBTITLE:
+        return
+    target_chat = kwargs["target_chat"] if kwargs.get("target_chat") else message.chat.id
+    # send docs if message == "/ai", without reply
+    texts = message.text or message.caption or ""
+    if str(texts).lower().strip() == PREFIX.SUBTITLE and not message.reply_to_message:
+        await send2tg(client, message, texts=HELP, **kwargs)
+        return
+
+    if not (vid := await find_yt_vid(client, message)):
+        return
+    yt_url = f"https://www.youtube.com/watch?v={vid}"
+    msg = f"🔍**正在获取字幕**\n{yt_url}"
+    if kwargs.get("show_progress"):
+        res = await send2tg(client, message, texts=msg, **kwargs)
+        kwargs["progress"] = res[0]
+    # cache media_group message
+    if media_group_id := message.media_group_id:
+        if cache.get(f"subtitle-{message.chat.id}-{media_group_id}"):
+            return
+        cache.set(f"subtitle-{message.chat.id}-{media_group_id}", "1", ttl=120)
+
+    if res := await fetch_subtitle(vid):
+        logger.success(res)
+        if subtitles := res.get("subtitle", ""):
+            length = len(subtitles)
+            with io.BytesIO(subtitles.encode("utf-8")) as f:
+                await client.send_document(target_chat, f, file_name=f"{vid}.vtt-{length}字符.txt", caption=yt_url)
+        elif error := res.get("error", ""):
+            await modify_progress(text=error, force_update=True, **kwargs)
+            await asyncio.sleep(3)
+    else:
+        await modify_progress(text="❌获取字幕失败", force_update=True, **kwargs)
+        await asyncio.sleep(3)
+    await modify_progress(del_status=True, **kwargs)
+
+
+async def find_yt_vid(client: Client, message: Message) -> str:
+    msg_text = message.text or message.caption or ""
+    if not msg_text.strip().lower().startswith(PREFIX.SUBTITLE):
+        return ""
+    url = find_url_in_message(message)
+    # /subtitle "link"
+    info = await match_social_media_link(url, flatten_first=True)
+    if info["platform"] == "youtube":
+        return info["vid"]
+
+    # is replying to message?
+    if not message.reply_to_message:
+        return ""
+    reply_message = message.reply_to_message
+    # if reply to a media_group, fetch all messages in the group
+    reply_messages = await client.get_media_group(message.chat.id, message.id) if message.media_group_id else [reply_message]
+    for msg in reply_messages:
+        url = find_url_in_message(msg)
+        info = await match_social_media_link(url, flatten_first=True)
+        if info["platform"] == "youtube":
+            return info["vid"]
+    return ""
+
+
+def find_url_in_message(message: Message) -> str:
+    # check first url in entities
+    if message.entities:
+        for entity in message.entities:
+            if entity.type == MessageEntityType.TEXT_LINK:
+                return entity.url
+    if message.caption_entities:
+        for entity in message.caption_entities:
+            if entity.type == MessageEntityType.TEXT_LINK:
+                return entity.url
+    return str(message.text).strip() if message.text else str(message.caption).strip()
+
+
+async def fetch_subtitle(video_id: str) -> dict:
+    proxy = {"http": PROXY.SUBTITLE, "https": PROXY.SUBTITLE} if PROXY.SUBTITLE else None
+    logger.info(f"Fetch Subtitle for {video_id=}, {proxy=}")
+    res = {}
+    try:
+        subtitles: list[dict] = YouTubeTranscriptApi.get_transcript(video_id=video_id, languages=["zh-CN", "zh-Hans", "zh", "zh-HK", "zh-TW", "zh-Hant", "en"], proxies=proxy)
+        res["subtitle"] = to_webvtt(subtitles)
+    except Exception as e:
+        logger.error(f"Failed to get subtitle: {e}")
+        return await fetch_subtitle_tikhub(video_id)
+    return res
+
+
+async def fetch_subtitle_tikhub(video_id: str) -> dict:
+    logger.info(f"Fetch Subtitle for {video_id=}")
+    api_url = f"{API.TIKHUB}/api/v1/youtube/web/get_video_subtitles?video_id={video_id}"
+    headers = {"authorization": f"Bearer {TOKEN.TIKHUB}", "accept": "application/json"}
+    resp = await hx_req(api_url, headers=headers, check_has_kv=["data"], check_kv={"code": 200})
+    if resp.status_code != 200:
+        logger.warning(f"Subtitle API failed: {resp}")
+        return {}
+    if subtitles := resp.json()["data"].get("subtitles", []):
+        return {"subtitle": to_webvtt(subtitles)}
+    if error := resp.json()["data"].get("detail", []):
+        return {"error": error}
+    return {}
+
+
+def to_webvtt(subtitles: list[dict]) -> str:
+    """Converts subtitles to WebVTT format.
+
+    sample subtitles = [
+        {'text': 'hello', 'start': 0.056, 'duration': 2.88},
+        {'text': 'world!', 'start': 2.983, 'duration': 3.244},
+    ]
+    """
+
+    def format_timestamp(seconds: str | float) -> str:
+        """Converts seconds to WebVTT timestamp format (hh:mm:ss.mmm)."""
+        ms = int((float(seconds) % 1) * 1000)
+        time = timedelta(seconds=int(seconds))
+        total_seconds = int(time.total_seconds())
+        hours, remainder = divmod(total_seconds, 3600)
+        minutes, seconds = divmod(remainder, 60)
+        return f"{hours:02}:{minutes:02}:{seconds:02}.{ms:03}"
+
+    vtt_output = ["WEBVTT", ""]  # WebVTT header
+    for subtitle in subtitles:
+        start = format_timestamp(subtitle["start"])
+        end = format_timestamp(subtitle["start"] + subtitle["duration"])
+        text = subtitle["text"]
+        vtt_output.append(f"{start} --> {end}")
+        vtt_output.append(text)
+        vtt_output.append("")  # Add blank line between subtitles
+
+    return "\n".join(vtt_output)

@@ -0,0 +1,161 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+from datetime import datetime
+from pathlib import Path
+from zoneinfo import ZoneInfo
+
+from loguru import logger
+from pyrogram.client import Client
+from pyrogram.types import Message
+
+from bridge.parsehub import send_to_parsehub
+from config import API, DB, TOKEN, TZ, cache
+from database import get_db
+from message_utils import copy_messages_from_db, modify_progress, save_messages, send2tg, summay_media
+from networking import download_first_success_urls, download_media, hx_req
+from others.emoji import emojify
+
+
+@cache.memoize(ttl=30)
+async def preview_douyin(
+    client: Client,
+    message: Message,
+    url: str = "",
+    platform: str = "douyin",
+    douyin_extractor: str | None = None,
+    douyin_comments_extractor: str | None = None,
+    *,
+    fetch_douyin_comments: bool = True,
+    fallback: bool = True,
+    **kwargs,
+):
+    """Preview douyin or tiktok link in the message.
+
+    Args:
+        client (Client): The Pyrogram client.
+        message (Message): The trigger message object.
+        url (str, optional): The douyin or tiktok link.
+        platform(str, optional): The platform name. Defaults to "douyin".
+        douyin_extractor (str, optional): The douyin extractor: "free" or "tikhub". Defaults to "free".
+        douyin_comments_extractor (str, optional): The douyin comments extractor: "free" or "tikhub". Defaults to "free".
+        fetch_douyin_comments (bool, optional): Fetch douyin comments. Defaults to True.
+        fallback (bool, optional): Fallback to other bots. Defaults to True.
+    """
+    if kwargs.get("show_progress") and "progress" not in kwargs:
+        res = await send2tg(client, message, texts=f"🔗正在解析抖音链接\n{url}", **kwargs)
+        kwargs["progress"] = res[0]
+    db_key = url
+    if kv := await get_db(db_key):
+        logger.debug(f"{platform} preview {DB.ENGINE} cache hit for key={db_key}")
+        if await copy_messages_from_db(client, message, key=db_key, kv=kv, **kwargs):
+            return
+        await modify_progress(text=f"❌从{DB.ENGINE}缓存中转发失败, 尝试重新解析...", **kwargs)
+
+    logger.info(f"{platform} link preview for {url}")
+    succ = False
+    if douyin_extractor is None or douyin_extractor == "free":  # try free first
+        api_url = f"{API.TIKHUB_FREE}/api/hybrid/video_data?url={url}"
+        headers = {"accept": "application/json"} if douyin_extractor == "tikhub" else {}
+        try:
+            resp = await hx_req(api_url, headers=headers, check_has_kv=["data"], check_kv={"code": 200})
+            data = resp.json()["data"]
+            succ = True
+        except Exception:
+            logger.warning(f"{platform} API [free] failed: {resp}")
+    if not succ:  # try tikhub
+        api_url = f"{API.TIKHUB}/api/v1/hybrid/video_data?url={url}"
+        headers = {"authorization": f"Bearer {TOKEN.TIKHUB}", "accept": "application/json"}
+        try:
+            resp = await hx_req(api_url, headers=headers, check_has_kv=["data"], check_kv={"code": 200})
+            data = resp.json()["data"]
+        except Exception:
+            logger.warning(f"{platform} API [tikhub] failed: {resp}")
+            if fallback:
+                await modify_progress(text="❌抖音解析失败, 尝试第三方Bot...", **kwargs)
+                await send_to_parsehub(client, message, text=url, **kwargs)
+            await modify_progress(del_status=True, **kwargs)
+            return
+
+    aweme_id = data.get("aweme_id", Path(url).stem)
+    if int(data.get("media_type", 4)) == 2:  # image post
+        media = [{"photo": download_first_success_urls(x.get("url_list", []), workers_proxy=True, **kwargs)} for x in data.get("images", [])]
+    else:  # video post
+        video = data.get("video", {})
+        video_urls = []
+        for key in ["play_addr_h264", "play_addr_265", "play_addr"]:
+            video_urls.extend(video.get(key, {}).get("url_list", []))
+        media = [{"video": download_first_success_urls(video_urls, suffix=".mp4", workers_proxy=True, **kwargs)}]
+    await modify_progress(text=f"⏬正在下载:\n{summay_media(media)}", force_update=True, **kwargs)
+    media = await download_media(media, **kwargs)
+    texts = ""
+    if author := data.get("author", {}).get("nickname", ""):
+        texts += f"\n🎶**[{author}]({url})**"
+    if ts := data.get("create_time"):
+        dt = datetime.fromtimestamp(ts).astimezone(ZoneInfo(TZ))
+        texts += f"\n🕒{dt:%Y-%m-%d %H:%M:%S}"
+    if decs := data.get("desc"):
+        texts += f"\n{decs}"
+
+    comments = []
+    if fetch_douyin_comments and (comments_list := await get_comments(aweme_id, platform, douyin_comments_extractor)):
+        comments.append("\n**> 💬**点此展开评论区**:")
+        for idx, cmt in enumerate(comments_list):
+            cmt_text = cmt["text"].replace("\n", "\n> ")
+            if idx == len(comments_list) - 1:  # last cmt
+                comments.append(f"\n> 💬**{cmt['name']}**{cmt['region']}: {cmt_text}||")
+            else:
+                comments.append(f"\n> 💬**{cmt['name']}**{cmt['region']}: {cmt_text}")
+
+    sent_messages = await send2tg(client, message, texts=emojify(texts), media=media, comments=comments, **kwargs)
+    await modify_progress(del_status=True, **kwargs)
+    await save_messages(messages=sent_messages, key=db_key)
+
+
+async def get_comments(aweme_id: str = "", platform: str = "douyin", douyin_comments_extractor: str | None = None) -> list[dict]:
+    """Fetch douyin or tiktok comments.
+
+    Args:
+        aweme_id (str, optional): post id.
+        platform (str, optional): douyin or tiktok. Defaults to "douyin".
+        douyin_comments_extractor (str | None, optional): The douyin comments extractor: "free" or "tikhub". Defaults to "free".
+
+    Returns:
+        list[dict]: comments list.
+    """
+    comments = []
+    api_urls = {
+        "douyin_tikhub": f"{API.TIKHUB}/api/v1/douyin/web/fetch_video_comments?aweme_id={aweme_id}",
+        "douyin_free": f"{API.TIKHUB_FREE}/api/douyin/web/fetch_video_comments?aweme_id={aweme_id}",
+        "tiktok_tikhub": f"{API.TIKHUB}/api/v1/tiktok/app/v3/fetch_video_comments?aweme_id={aweme_id}",
+        "tiktok_free": f"{API.TIKHUB_FREE}/api/tiktok/web/fetch_post_comment?aweme_id={aweme_id}",
+    }
+    succ = False
+    if douyin_comments_extractor is None or douyin_comments_extractor == "free":  # try free first
+        api_url = api_urls.get(f"{platform}_free")
+        headers = {"accept": "application/json"}
+        try:
+            resp = await hx_req(api_url, headers=headers, check_has_kv=["data.comments"], check_kv={"code": 200})
+            data = resp.json()["data"]["comments"]
+            succ = True
+        except Exception:
+            logger.warning(f"{platform} comments API [free] failed: {resp}")
+    if not succ:  # try tikhub
+        api_url = api_urls.get(f"{platform}_tikhub")
+        headers = {"authorization": f"Bearer {TOKEN.TIKHUB}", "accept": "application/json"}
+        try:
+            resp = await hx_req(api_url, headers=headers, check_has_kv=["data.comments"], check_kv={"code": 200})
+            data = resp.json()["data"]["comments"]
+        except Exception:
+            logger.warning(f"{platform} comments API [tikhub] failed: {resp}")
+            return []
+    try:
+        for node in data:
+            name = node.get("user", {}).get("nickname", "")
+            text = node.get("text", "")
+            region = f"({node['ip_label']})" if node.get("ip_label") else ""
+            if name and text:
+                comments.append({"name": name, "text": emojify(text.strip()), "region": region})
+    except Exception as e:
+        logger.error(e)
+        return []
+    return comments

@@ -0,0 +1,151 @@
+#!/usr/bin/env python
+# -*- coding: utf-8 -*-
+
+import asyncio
+from datetime import datetime
+from zoneinfo import ZoneInfo
+
+from bs4 import BeautifulSoup
+from loguru import logger
+from pyrogram.client import Client
+from pyrogram.types import Message
+
+from bridge.parsehub import send_to_parsehub
+from config import API, DB, DOWNLOAD_DIR, TOKEN, TZ, UA, cache
+from database import get_db
+from message_utils import copy_messages_from_db, modify_progress, save_messages, send2tg, summay_media
+from multimedia import is_valid_video, validate_img
+from networking import download_file, download_media, hx_req
+
+
+@cache.memoize(ttl=30)
+async def preview_instagram(client: Client, message: Message, url: str = "", *, fallback: bool = True, **kwargs):
+    """Preview instagram link in the message.
+
+    Args:
+        client (Client): The Pyrogram client.
+        message (Message): The trigger message object.
+        url (str, optional): Tnstagram link.
+        fallback (bool, optional): Fallback to other bots. Defaults to True.
+    """
+    if kwargs.get("show_progress") and "progress" not in kwargs:
+        res = await send2tg(client, message, texts=f"🔗正在解析Instagram链接\n{url}", **kwargs)
+        kwargs["progress"] = res[0]
+    db_key = url
+    if kv := await get_db(db_key):
+        logger.debug(f"Instagram preview {DB.ENGINE} cache hit for key={db_key}")
+        if await copy_messages_from_db(client, message, key=db_key, kv=kv, **kwargs):
+            return
+        await modify_progress(text=f"❌从{DB.ENGINE}缓存中转发失败, 尝试重新解析...", **kwargs)
+
+    api_url = API.TIKHUB_INSTAGRAM + url
+    logger.info(f"Preview Instagram TikHub for {api_url}")
+    headers = {"authorization": f"Bearer {TOKEN.TIKHUB}", "accept": "application/json"}
+    resp = await hx_req(api_url, headers=headers, check_has_kv=["data"], check_kv={"code": 200})
+    if resp.status_code != 200:
+        await modify_progress(text="❌Instagram解析失败, 使用DDInstagram预览", **kwargs)
+        await preview_ddinstagram(client, message, fallback=fallback, **kwargs)
+        await asyncio.sleep(2)
+        await modify_progress(del_status=True, **kwargs)
+        return
+
+    data = resp.json()["data"]
+
+    # parse media
+    media = []
+    if data.get("video_url"):  # reel
+        media.append({"video": download_file(data.get("video_url", ""), **kwargs)})
+    elif media_nodes := data.get("edge_sidecar_to_children", {}).get("edges", []):
+        for node in media_nodes:
+            ftype = "photo" if not node.get("node", {}).get("is_video") else "video"
+            url = node.get("node", {}).get("display_url", "") if ftype == "photo" else node.get("node", {}).get("video_url", "")
+            media.append({ftype: download_file(url, **kwargs)})
+    elif data.get("display_url"):
+        media.append({"photo": download_file(data.get("display_url"), **kwargs)})
+
+    texts = ""
+    if fullname := data.get("owner", {}).get("full_name"):
+        texts += f"🏞**[{fullname}]({url})**\n"
+
+    if metadata_nodes := data.get("edge_media_to_caption", {}).get("edges"):
+        if ts := metadata_nodes[0].get("node", {}).get("created_at"):
+            dt = datetime.fromtimestamp(float(ts)).astimezone(ZoneInfo(TZ))
+            create_time = f"{dt:%Y-%m-%d %H:%M:%S}"
+            texts += f"🕒{create_time}\n"
+        if description := metadata_nodes[0].get("node", {}).get("text", ""):
+            texts += f"{description}\n"
+
+    # parse comments
+    comment_nodes = data.get("edge_media_to_parent_comment", {}).get("edges", [])
+    comment_nodes = sorted(comment_nodes, key=lambda x: x.get("node", {}).get("created_at", 0))
+    comment_list = [{"author": node.get("node", {}).get("owner", {}).get("username", "user"), "text": node.get("node", {}).get("text", "")} for node in comment_nodes]
+    comment_list = [x for x in comment_list if x["text"]]
+    comments: list[str] = []
+    for idx, cmt in enumerate(comment_list):
+        cmt_text = cmt["text"].replace("\n", "\n> ")
+        if idx == 0:
+            comments.append("\n**> 💬**点此展开评论区**:")
+        if idx == len(comment_list) - 1:  # last cmt
+            comments.append(f"\n> 💬**{cmt['author']}**: {cmt_text}||")
+        else:
+            comments.append(f"\n> 💬**{cmt['author']}**: {cmt_text}")
+
+    await modify_progress(text=f"⏬正在下载:\n{summay_media(media)}", force_update=True, **kwargs)
+    media = await download_media(media, **kwargs)
+    sent_messages = await send2tg(client, message, texts=texts.strip(), media=media, comments=comments, **kwargs)
+    await modify_progress(del_status=True, **kwargs)
+    await save_messages(messages=sent_messages, key=db_key)
+
+
+@cache.memoize(ttl=30)
+async def preview_ddinstagram(client: Client, message: Message, url: str, post_type: str, post_id: str, *, fallback: bool = True, **kwargs):
+    """Preview instagram link in the message via DDInstagram.
+
+    https://ddinstagram.com/
+
+    Args:
+        client (Client): The Pyrogram client.
+        message (Message): The trigger message object.
+        url (str, optional): Tnstagram link.
+        post_type (str): post type: "p" or "reel"
+        post_id (str): post id.
+        fallback (bool, optional): Fallback to other bots. Defaults to True.
+    """
+    api_url = f"{API.DDINSTAGRAM}/{post_type}/{post_id}"
+    logger.info(f"Instagram link preview for {api_url}")
+    headers = {"user-agent": UA.TELEGRAM}
+    resp = await hx_req(api_url, headers=headers)
+    if not resp.text:
+        if fallback:
+            await send_to_parsehub(client, message, text=url, **kwargs)
+        return
+
+    soup = BeautifulSoup(resp.text, "html.parser")
+    logger.trace(soup.prettify())
+
+    texts = ""
+    media = {}
+    if tag := soup.find("meta", attrs={"name": "twitter:title"}):
+        author = tag.get("content", "Unknown")  # type: ignore
+        texts += f"🏞**[{author}]({url})\n"
+    if tag := soup.find("meta", attrs={"property": "og:description"}):
+        texts: str = tag.get("content", "")  # type: ignore
+    if tag := soup.find("meta", attrs={"name": "twitter:image"}):
+        img_url = tag.get("content", "")  # type: ignore
+        if img_url:
+            raw_url = f"{API.DDINSTAGRAM}{img_url}"
+            media["photo"] = await download_file(raw_url, path=f"{DOWNLOAD_DIR}/{post_id}.jpg", workers_proxy=True, **kwargs)
+            if not bool(validate_img(media["photo"])):
+                await send_to_parsehub(client, message, text=url, **kwargs)
+                return
+
+    if tag := soup.find("meta", attrs={"property": "og:video"}):
+        video_url = tag.get("content", "")  # type: ignore
+        if video_url:
+            raw_url = f"{API.DDINSTAGRAM}{video_url}"
+            media["video"] = await download_file(raw_url, path=f"{DOWNLOAD_DIR}/{post_id}.mp4", workers_proxy=True, **kwargs)
+            if not is_valid_video(media["video"]):
+                await send_to_parsehub(client, message, text=url, **kwargs)
+                return
+
+    await send2tg(client, message, texts=texts, media=[media], **kwargs)