Add youtube-transcript-api as tool (#10772)

Co-authored-by: crazywoola <427733928@qq.com>

api/core/tools/provider/builtin/transcript/_assets/icon.svg

@@ -0,0 +1,11 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<svg width="800px" height="800px" viewBox="0 -38 256 256" version="1.1" xmlns="http://www.w3.org/2000/svg"
+     xmlns:xlink="http://www.w3.org/1999/xlink" preserveAspectRatio="xMidYMid">
+    <g>
+        <path d="M250.346231,28.0746923 C247.358133,17.0320558 238.732098,8.40602109 227.689461,5.41792308 C207.823743,0 127.868333,0 127.868333,0 C127.868333,0 47.9129229,0.164179487 28.0472049,5.58210256 C17.0045684,8.57020058 8.37853373,17.1962353 5.39043571,28.2388718 C-0.618533519,63.5374615 -2.94988224,117.322662 5.5546152,151.209308 C8.54271322,162.251944 17.1687479,170.877979 28.2113844,173.866077 C48.0771024,179.284 128.032513,179.284 128.032513,179.284 C128.032513,179.284 207.987923,179.284 227.853641,173.866077 C238.896277,170.877979 247.522312,162.251944 250.51041,151.209308 C256.847738,115.861464 258.801474,62.1091 250.346231,28.0746923 Z"
+              fill="#FF0000">
+        </path>
+        <polygon fill="#FFFFFF" points="102.420513 128.06 168.749025 89.642 102.420513 51.224">
+        </polygon>
+    </g>
+</svg>

api/core/tools/provider/builtin/transcript/tools/transcript.py

@@ -0,0 +1,81 @@
+from typing import Any, Union
+from urllib.parse import parse_qs, urlparse
+
+from youtube_transcript_api import YouTubeTranscriptApi
+
+from core.tools.entities.tool_entities import ToolInvokeMessage
+from core.tools.tool.builtin_tool import BuiltinTool
+
+
+class YouTubeTranscriptTool(BuiltinTool):
+    def _invoke(
+        self, user_id: str, tool_parameters: dict[str, Any]
+    ) -> Union[ToolInvokeMessage, list[ToolInvokeMessage]]:
+        """
+        Invoke the YouTube transcript tool
+        """
+        try:
+            # Extract parameters with defaults
+            video_input = tool_parameters["video_id"]
+            language = tool_parameters.get("language")
+            output_format = tool_parameters.get("format", "text")
+            preserve_formatting = tool_parameters.get("preserve_formatting", False)
+            proxy = tool_parameters.get("proxy")
+            cookies = tool_parameters.get("cookies")
+
+            # Extract video ID from URL if needed
+            video_id = self._extract_video_id(video_input)
+
+            # Common kwargs for API calls
+            kwargs = {"proxies": {"https": proxy} if proxy else None, "cookies": cookies}
+
+            try:
+                if language:
+                    transcript_list = YouTubeTranscriptApi.list_transcripts(video_id, **kwargs)
+                    try:
+                        transcript = transcript_list.find_transcript([language])
+                    except:
+                        # If requested language not found, try translating from English
+                        transcript = transcript_list.find_transcript(["en"]).translate(language)
+                    transcript_data = transcript.fetch()
+                else:
+                    transcript_data = YouTubeTranscriptApi.get_transcript(
+                        video_id, preserve_formatting=preserve_formatting, **kwargs
+                    )
+
+                # Format output
+                formatter_class = {
+                    "json": "JSONFormatter",
+                    "pretty": "PrettyPrintFormatter",
+                    "srt": "SRTFormatter",
+                    "vtt": "WebVTTFormatter",
+                }.get(output_format)
+
+                if formatter_class:
+                    from youtube_transcript_api import formatters
+
+                    formatter = getattr(formatters, formatter_class)()
+                    formatted_transcript = formatter.format_transcript(transcript_data)
+                else:
+                    formatted_transcript = " ".join(entry["text"] for entry in transcript_data)
+
+                return self.create_text_message(text=formatted_transcript)
+
+            except Exception as e:
+                return self.create_text_message(text=f"Error getting transcript: {str(e)}")
+
+        except Exception as e:
+            return self.create_text_message(text=f"Error processing request: {str(e)}")
+
+    def _extract_video_id(self, video_input: str) -> str:
+        """
+        Extract video ID from URL or return as-is if already an ID
+        """
+        if "youtube.com" in video_input or "youtu.be" in video_input:
+            # Parse URL
+            parsed_url = urlparse(video_input)
+            if "youtube.com" in parsed_url.netloc:
+                return parse_qs(parsed_url.query)["v"][0]
+            else:  # youtu.be
+                return parsed_url.path[1:]
+        return video_input  # Assume it's already a video ID

api/core/tools/provider/builtin/transcript/tools/transcript.yaml

@@ -0,0 +1,101 @@
+identity:
+  name: free_youtube_transcript
+  author: Tao Wang
+  label:
+    en_US: Free YouTube Transcript API
+    zh_Hans: 免费获取 YouTube 转录
+description:
+  human:
+    en_US: Get transcript from a YouTube video for free.
+    zh_Hans: 免费获取 YouTube 视频的转录文案。
+  llm: A tool for retrieving transcript from YouTube videos.
+parameters:
+  - name: video_id
+    type: string
+    required: true
+    label:
+      en_US: Video ID/URL
+      zh_Hans: 视频ID
+    human_description:
+      en_US: Used to define the video from which the transcript will be fetched. You can find the id in the video url. For example - https://www.youtube.com/watch?v=video_id.
+      zh_Hans: 您要哪条视频的转录文案？您可以在视频链接中找到id。例如 - https://www.youtube.com/watch?v=video_id。
+    llm_description: Used to define the video from which the transcript will be fetched. For example - https://www.youtube.com/watch?v=video_id.
+    form: llm
+  - name: language
+    type: string
+    required: false
+    label:
+      en_US: Language Code
+      zh_Hans: 语言
+    human_description:
+      en_US: Language code (e.g. 'en', 'zh') for the transcript.
+      zh_Hans: 字幕语言代码（如'en'、'zh'）。留空则自动选择。
+    llm_description: Used to set the language for transcripts.
+    form: form
+  - name: format
+    type: select
+    required: false
+    default: text
+    options:
+      - value: text
+        label:
+          en_US: Plain Text
+          zh_Hans: 纯文本
+      - value: json
+        label:
+          en_US: JSON Format
+          zh_Hans: JSON 格式
+      - value: pretty
+        label:
+          en_US: Pretty Print Format
+          zh_Hans: 美化格式
+      - value: srt
+        label:
+          en_US: SRT Format
+          zh_Hans: SRT 格式
+      - value: vtt
+        label:
+          en_US: WebVTT Format
+          zh_Hans: WebVTT 格式
+    label:
+      en_US: Output Format
+      zh_Hans: 输出格式
+    human_description:
+      en_US: Format of the transcript output
+      zh_Hans: 字幕输出格式
+    llm_description: The format to output the transcript in. Options are text (plain text), json (raw transcript data), srt (SubRip format), or vtt (WebVTT format)
+    form: form
+  - name: preserve_formatting
+    type: boolean
+    required: false
+    default: false
+    label:
+      en_US: Preserve Formatting
+      zh_Hans: 保留格式
+    human_description:
+      en_US: Keep HTML formatting elements like <i> (italics) and <b> (bold)
+      zh_Hans: 保留HTML格式元素，如<i>（斜体）和<b>（粗体）
+    llm_description: Whether to preserve HTML formatting elements in the transcript text
+    form: form
+  - name: proxy
+    type: string
+    required: false
+    label:
+      en_US: HTTPS Proxy
+      zh_Hans: HTTPS 代理
+    human_description:
+      en_US: HTTPS proxy URL (e.g. https://user:pass@domain:port)
+      zh_Hans: HTTPS 代理地址（如 https://user:pass@domain:port）
+    llm_description: HTTPS proxy to use for the request. Format should be https://user:pass@domain:port
+    form: form
+  - name: cookies
+    type: string
+    required: false
+    label:
+      en_US: Cookies File Path
+      zh_Hans: Cookies 文件路径
+    human_description:
+      en_US: Path to cookies.txt file for accessing age-restricted videos
+      zh_Hans: 用于访问年龄限制视频的 cookies.txt 文件路径
+    llm_description: Path to a cookies.txt file containing YouTube cookies, needed for accessing age-restricted videos
+    form: form

api/core/tools/provider/builtin/transcript/transcript.py

@@ -0,0 +1,11 @@
+from typing import Any
+
+from core.tools.provider.builtin_tool_provider import BuiltinToolProviderController
+
+
+class YouTubeTranscriptProvider(BuiltinToolProviderController):
+    def _validate_credentials(self, credentials: dict[str, Any]) -> None:
+        """
+        No credentials needed for YouTube Transcript API
+        """
+        pass

api/core/tools/provider/builtin/transcript/transcript.yaml

@@ -0,0 +1,13 @@
+identity:
+  author: Tao Wang
+  name: transcript
+  label:
+    en_US: Transcript
+    zh_Hans: Transcript
+  description:
+    en_US: Get transcripts from YouTube videos
+    zh_Hans: 获取 YouTube 视频的字幕/转录文本
+  icon: icon.svg
+  tags:
+    - videos
+credentials_for_provider:

api/pyproject.toml

@@ -187,6 +187,7 @@ websocket-client = "~1.7.0"
 werkzeug = "~3.0.1"
 xinference-client = "0.15.2"
 yarl = "~1.9.4"
+youtube-transcript-api = "~0.6.2"
 zhipuai = "~2.1.5"
 # Before adding new dependency, consider place it in alphabet order (a-z) and suitable group.