基于Gemini AI和Decodo的Telegram学术论文研究助手
高级
这是一个Document Extraction, AI RAG领域的自动化工作流，包含 25 个节点。主要使用 If, Set, Switch, Telegram, ConvertToFile 等节点。 基于Gemini AI和Decodo的Telegram学术论文研究助手
前置要求
- •Telegram Bot Token
- •Google Gemini API Key
使用的节点 (25)
工作流预览
可视化展示节点连接关系，支持缩放和平移
导出工作流
复制以下 JSON 配置到 n8n 导入，即可使用此工作流
{
"meta": {
"instanceId": "689fa22e68cd4198e4ae37f3cc44f498087edd235a867e22515be823bab694c7",
"templateCredsSetupCompleted": true
},
"nodes": [
{
"id": "9dbc82a7-bc8a-4047-8875-c7f8905e122d",
"name": "Decodo",
"type": "@decodo/n8n-nodes-decodo.decodoTool",
"position": [
832,
-208
],
"parameters": {},
"credentials": {
"decodoApi": {
"id": "kfy3Vs5AgxfDGzpC",
"name": "Johan Decodo"
}
},
"typeVersion": 1
},
{
"id": "fb3b266b-50a5-4bc9-a7bc-33c5a86f7e69",
"name": "简单记忆",
"type": "@n8n/n8n-nodes-langchain.memoryBufferWindow",
"position": [
720,
-208
],
"parameters": {
"sessionKey": "={{ $json.chatId }}",
"sessionIdType": "customKey"
},
"typeVersion": 1.3
},
{
"id": "5c31bc4e-b47b-44bc-b140-c6e5e5273487",
"name": "发送备用文本",
"type": "n8n-nodes-base.telegram",
"position": [
-320,
-96
],
"webhookId": "7f900de4-3fd7-4a1d-920f-a85c89690d34",
"parameters": {
"text": "=Sorry, i can only support text, photo, and voice message",
"chatId": "={{ $json.message.chat.id }}",
"additionalFields": {
"appendAttribution": false
}
},
"credentials": {
"telegramApi": {
"id": "xQbv3KvDfR6stpmr",
"name": "Johan Telegram Research Bot"
}
},
"typeVersion": 1.2
},
{
"id": "cc90e87d-dd74-46dc-a434-290580532858",
"name": "启动 Telegram 机器人",
"type": "n8n-nodes-base.telegramTrigger",
"position": [
-768,
-368
],
"webhookId": "19e6481c-ed29-4f13-890d-1930d24a68eb",
"parameters": {
"updates": [
"message"
],
"additionalFields": {}
},
"credentials": {
"telegramApi": {
"id": "xQbv3KvDfR6stpmr",
"name": "Johan Telegram Research Bot"
}
},
"typeVersion": 1.2
},
{
"id": "e2158b5b-c375-4e23-ba3c-6f07e1f2d423",
"name": "检测消息类型",
"type": "n8n-nodes-base.switch",
"position": [
-544,
-400
],
"parameters": {
"rules": {
"values": [
{
"outputKey": "Photo",
"conditions": {
"options": {
"version": 2,
"leftValue": "",
"caseSensitive": true,
"typeValidation": "strict"
},
"combinator": "and",
"conditions": [
{
"id": "9011a574-9680-451e-9290-4001a103887d",
"operator": {
"type": "array",
"operation": "notEmpty",
"singleValue": true
},
"leftValue": "={{ $json.message.photo }}",
"rightValue": ""
}
]
},
"renameOutput": true
},
{
"outputKey": "Text",
"conditions": {
"options": {
"version": 2,
"leftValue": "",
"caseSensitive": true,
"typeValidation": "strict"
},
"combinator": "and",
"conditions": [
{
"id": "32ffc6b3-5f14-4ad9-922b-76088e2492c3",
"operator": {
"type": "string",
"operation": "notEmpty",
"singleValue": true
},
"leftValue": "={{ $json.message.text }}",
"rightValue": ""
}
]
},
"renameOutput": true
},
{
"outputKey": "Voice",
"conditions": {
"options": {
"version": 2,
"leftValue": "",
"caseSensitive": true,
"typeValidation": "strict"
},
"combinator": "and",
"conditions": [
{
"id": "a1d47f49-6acf-404a-8176-d9ba6b17523e",
"operator": {
"type": "object",
"operation": "notEmpty",
"singleValue": true
},
"leftValue": "={{ $json.message.voice }}",
"rightValue": ""
}
]
},
"renameOutput": true
}
]
},
"options": {
"fallbackOutput": "extra"
}
},
"typeVersion": 3.3
},
{
"id": "d14d10c3-9504-4c36-ad73-d06402e17a3f",
"name": "下载 Telegram 照片",
"type": "n8n-nodes-base.telegram",
"position": [
-320,
-576
],
"webhookId": "b72f1572-ecdd-4b2a-81c5-dff9474e3bbe",
"parameters": {
"fileId": "={{ $json.message.photo[3].file_id }}",
"resource": "file",
"additionalFields": {}
},
"credentials": {
"telegramApi": {
"id": "xQbv3KvDfR6stpmr",
"name": "Johan Telegram Research Bot"
}
},
"typeVersion": 1.2
},
{
"id": "bce76f9d-c571-4743-9314-2ef5b0041f1d",
"name": "下载 Telegram 语音",
"type": "n8n-nodes-base.telegram",
"position": [
-320,
-288
],
"webhookId": "b72f1572-ecdd-4b2a-81c5-dff9474e3bbe",
"parameters": {
"fileId": "={{ $json.message.voice.file_id }}",
"resource": "file",
"additionalFields": {}
},
"credentials": {
"telegramApi": {
"id": "xQbv3KvDfR6stpmr",
"name": "Johan Telegram Research Bot"
}
},
"typeVersion": 1.2
},
{
"id": "a919f02b-3cd6-4e56-8f4b-c204ddf22bc2",
"name": "分析图像内容",
"type": "@n8n/n8n-nodes-langchain.googleGemini",
"position": [
-96,
-576
],
"parameters": {
"modelId": {
"__rl": true,
"mode": "list",
"value": "models/gemini-2.5-flash",
"cachedResultName": "models/gemini-2.5-flash"
},
"options": {},
"resource": "image",
"inputType": "binary",
"operation": "analyze"
},
"credentials": {
"googlePalmApi": {
"id": "C4mfRdRQfQ524QK8",
"name": "Johan Gemini"
}
},
"typeVersion": 1
},
{
"id": "3a74f55a-135c-4247-93f9-c25eec8977ed",
"name": "转录语音消息",
"type": "@n8n/n8n-nodes-langchain.googleGemini",
"position": [
-96,
-288
],
"parameters": {
"modelId": {
"__rl": true,
"mode": "list",
"value": "models/gemini-2.5-flash",
"cachedResultName": "models/gemini-2.5-flash"
},
"options": {},
"resource": "audio",
"inputType": "binary"
},
"credentials": {
"googlePalmApi": {
"id": "C4mfRdRQfQ524QK8",
"name": "Johan Gemini"
}
},
"typeVersion": 1
},
{
"id": "0a66c1aa-9275-4d71-9cfe-7b4781b8c105",
"name": "格式化图像文本",
"type": "n8n-nodes-base.set",
"position": [
128,
-576
],
"parameters": {
"options": {},
"assignments": {
"assignments": [
{
"id": "aa2b9784-2bcf-41ba-8954-4e59f4182f32",
"name": "message.text",
"type": "string",
"value": "=Caption: {{ $('Start Telegram Bot').item.json.message.caption ?? \"[none]\" }}\n---\nImage: {{ $json.content.parts[0].text }}"
}
]
}
},
"typeVersion": 3.4
},
{
"id": "7c216c18-a379-4197-99ec-53e732f03bcf",
"name": "格式化语音文本",
"type": "n8n-nodes-base.set",
"position": [
128,
-288
],
"parameters": {
"options": {},
"assignments": {
"assignments": [
{
"id": "133940af-39d6-49e3-927c-fedfc5843aac",
"name": "message.text",
"type": "string",
"value": "={{ $json.content.parts[0].text }}"
}
]
}
},
"typeVersion": 3.4
},
{
"id": "3b745aeb-a258-4da1-845e-122621a5e09d",
"name": "准备聊天数据",
"type": "n8n-nodes-base.set",
"position": [
352,
-432
],
"parameters": {
"options": {},
"assignments": {
"assignments": [
{
"id": "7ce0d75f-1ad5-4c6b-a5b0-f108da87b6fc",
"name": "chatInput",
"type": "string",
"value": "={{ $json.message.text }}"
},
{
"id": "7753621e-9d0b-4561-ab9e-f20d0931e523",
"name": "chatId",
"type": "string",
"value": "={{ $('Start Telegram Bot').item.json.message.chat.id }}"
}
]
}
},
"typeVersion": 3.4
},
{
"id": "e5bb021b-56fc-4223-a295-396bc17dde35",
"name": "研究摘要 Agent",
"type": "@n8n/n8n-nodes-langchain.agent",
"position": [
640,
-432
],
"parameters": {
"options": {
"systemMessage": "=## Overview\nYou are Research Summary Agent, a research assistant specialized in searching, analyzing, and summarizing academic papers based on user queries. Your role is to interpret the user's research intent, generate meaningful search queries, use provided academic sources and tools effectively, and return structured, concise, and useful summaries of relevant papers. \n\n## Tools\n- Decodo — A scraping utility capable of extracting titles, authors, abstracts, and publication details from academic web pages. \n\n## Objectives\n1. Understand User Intent — Analyze the user's message to determine the true research topic, context, and desired type of papers (e.g., comparative, conceptual, applied, or review). \n2. Extract Core Keywords — Identify the essential and meaningful keywords or phrases that represent the main concept of the user’s request. \n - Remove filler words (e.g., “the way of”, “study about”, “paper similar to”, “research on”) and any quotes or formatting. \n - If a specific paper title is mentioned, infer its *thematic keywords* (e.g., if “The Way of the Artificial Intelligence” is mentioned, infer related keywords such as *artificial intelligence, cognitive systems, AI philosophy, AI methodology*). \n3. Plan Source Usage — Review the available academic source URLs (e.g., Google Scholar, arXiv) and understand their parameters to determine the best way to perform relevant searches. \n4. Search & Retrieve — For each source, construct a clear and semantically rich search query using the extracted keywords, ensuring broad yet relevant coverage rather than literal string matching. \n5. Use Tools Appropriately — Use Decodo and other provided tools to scrape or retrieve paper metadata, abstracts, and links from the specified URLs. \n6. Filter & Rank Results — Prioritize results that are: \n - Topically relevant to the inferred research theme. \n - Recent or influential (if available). \n - Representative of the broader concept, not just identical titles. \n7. Summarize Findings — Summarize each relevant paper in 2–3 lines, including title, year, authors, short abstract, and its relation to the topic. \n8. Respond Clearly — Present all findings in a well-structured, easy-to-read format grouped by source (e.g., Google Scholar, arXiv).\n\n## Rules\n1. Only use the provided URLs and their described parameters when performing searches. Also, always search through all provided sources unless the user request for a specific source. \n2. Never use parameters across different sources — each parameter must strictly apply to its corresponding source. \n3. Do not use the user’s message as a literal search phrase. Always convert it into a meaningful keyword-based query that reflects intent and core concepts. \n4. Never make assumptions about unavailable sources or invent results. \n5. Maintain factual accuracy, clarity, and neutrality in all summaries. \n6. When using Decodo, only extract data relevant to research context (titles, abstracts, authors, publication year, and URLs). \n7. Always cite the source (Google Scholar, arXiv, etc.) in the output. \n8. Each paper summary must be limited to 2–3 lines maximum, focusing on key insights, contribution, and relevance to the topic. \n9. Do not include unrelated metadata, extended commentary, or subjective interpretation. \n10. Prefer generalized or semantically equivalent keywords over verbatim text to ensure broader and more effective search coverage.\n\n## Search Sources\nThe following list defines how each academic source and its parameters should be used for automated querying and analysis:\n\n{{INPUT_SEARCH_URL_INSIGHTS}}"
}
},
"typeVersion": 2.2
},
{
"id": "4d9c8f9f-d29b-4e59-bbf5-f8225bbd4c6e",
"name": "Gemini URL 解释器",
"type": "@n8n/n8n-nodes-langchain.lmChatGoogleGemini",
"position": [
384,
-784
],
"parameters": {
"options": {}
},
"credentials": {
"googlePalmApi": {
"id": "C4mfRdRQfQ524QK8",
"name": "Johan Gemini"
}
},
"typeVersion": 1
},
{
"id": "32994b1e-cd72-4fbd-82ac-4accd1e23cd0",
"name": "Gemini 研究模型",
"type": "@n8n/n8n-nodes-langchain.lmChatGoogleGemini",
"position": [
576,
-208
],
"parameters": {
"options": {}
},
"credentials": {
"googlePalmApi": {
"id": "C4mfRdRQfQ524QK8",
"name": "Johan Gemini"
}
},
"typeVersion": 1
},
{
"id": "0c4a3956-952a-477f-a38c-924cab806096",
"name": "检查 Telegram 消息长度",
"type": "n8n-nodes-base.if",
"position": [
1040,
-432
],
"parameters": {
"options": {},
"conditions": {
"options": {
"version": 2,
"leftValue": "",
"caseSensitive": true,
"typeValidation": "strict"
},
"combinator": "and",
"conditions": [
{
"id": "1ef96c76-7b6c-4ba4-9a4e-3e7271785b86",
"operator": {
"type": "number",
"operation": "gt"
},
"leftValue": "={{ $json.output.length }}",
"rightValue": 4000
}
]
}
},
"typeVersion": 2.2
},
{
"id": "6628f302-5c75-4092-a336-1d806607188e",
"name": "将输出转换为文本文件",
"type": "n8n-nodes-base.convertToFile",
"position": [
1264,
-528
],
"parameters": {
"options": {
"fileName": "research_summary"
},
"operation": "toText",
"sourceProperty": "output"
},
"typeVersion": 1.1
},
{
"id": "3344c22e-9320-4094-80fd-2a3d2a212692",
"name": "发送研究摘要文件",
"type": "n8n-nodes-base.telegram",
"position": [
1488,
-528
],
"webhookId": "f5c54b0c-5e05-4f48-b36c-051b912ed219",
"parameters": {
"chatId": "={{ $('Prepare Chat Data').item.json.chatId }}",
"operation": "sendDocument",
"binaryData": true,
"additionalFields": {
"caption": "=I've compiled the search result into a txt file because the content is too long for a single telegram message"
}
},
"credentials": {
"telegramApi": {
"id": "xQbv3KvDfR6stpmr",
"name": "Johan Telegram Research Bot"
}
},
"typeVersion": 1.2
},
{
"id": "4ea381ad-7f23-4f4b-932b-3ffc5b630035",
"name": "发送研究摘要消息",
"type": "n8n-nodes-base.telegram",
"position": [
1264,
-336
],
"webhookId": "7f900de4-3fd7-4a1d-920f-a85c89690d34",
"parameters": {
"text": "={{ $json.output }}",
"chatId": "={{ $('Prepare Chat Data').item.json.chatId }}",
"additionalFields": {
"parse_mode": "HTML",
"appendAttribution": false
}
},
"credentials": {
"telegramApi": {
"id": "xQbv3KvDfR6stpmr",
"name": "Johan Telegram Research Bot"
}
},
"typeVersion": 1.2
},
{
"id": "3fdba744-325a-42ec-b3d9-709da173f69f",
"name": "便签",
"type": "n8n-nodes-base.stickyNote",
"position": [
-848,
-1440
],
"parameters": {
"width": 816,
"height": 816,
"content": ""
},
"typeVersion": 1
},
{
"id": "8ea52c9c-2b19-4f22-9d5e-ded9a974ab06",
"name": "生成搜索 URL 洞察",
"type": "@n8n/n8n-nodes-langchain.chainLlm",
"position": [
320,
-1008
],
"parameters": {
"text": "=",
"batching": {},
"messages": {
"messageValues": [
{
"message": "=You are an AI URL Interpreter that specializes in analyzing academic search URLs. Your goal is to extract and explain all query parameters in a clear Markdown format so that an AI Research Agent can understand how to use them for research retrieval."
},
{
"type": "HumanMessagePromptTemplate",
"message": "=Your task is to take a list of academic search URLs and analyze each one.\n\nFor each URL, you must:\n1. Detect its source (e.g., Google Scholar, arXiv, Semantic Scholar, PubMed).\n2. Combine all URLs from the same source into one section.\n3. Extract and merge all unique parameters from those URLs (e.g., query terms, filters, search types, languages).\n3. Ignore UI-related or tracking parameters.\n5. Write one general note explaining what that platform’s search does and how parameters can modify behavior.\n\nThe output must be written in Markdown format, following this template for each URL:\n### [Platform Name]\n1. URL Examples:\n - [url_1]\n - [url_2]\n2. Note: [summarized explanation of the platform and how parameters affect the search]\n3. Parameters:\n - [parameter_name]: [meaning or function]\n - ..."
},
{
"type": "HumanMessagePromptTemplate",
"message": "=[\n \"https://scholar.google.com/scholar?q=artificial+intelligence&hl=en&as_sdt=0,5\",\n \"https://scholar.google.com/scholar?as_ylo=2025&q=artificial+intelligence&hl=en&as_sdt=0,5\",\n \"https://arxiv.org/search/?query=artificial+intelligence&searchtype=all&abstracts=show&order=-announced_date_first\",\n \"https://arxiv.org/search/?query=artificial+intelligence&searchtype=all&source=header\"\n]"
},
{
"type": "AIMessagePromptTemplate",
"message": "=### Google Scholar\n1. URL Examples:\n - https://scholar.google.com/scholar?q=artificial+intelligence&hl=en&as_sdt=0,5\n - https://scholar.google.com/scholar?as_ylo=2025&q=artificial+intelligence&hl=en&as_sdt=0,5\n2. Note: Google Scholar is a search engine for scholarly articles, theses, and academic literature. These URLs perform searches for “artificial intelligence” in English, including all scholarly articles and citations. Additional parameters like `as_ylo` allow filtering by publication year to find the most recent works.\n3. Parameters:\n - q: The main search query (e.g., \"artificial+intelligence\").\n - hl: Interface language (e.g., \"en\" for English).\n - as_sdt: Search scope control. \"0,5\" includes all articles and citations.\n - as_ylo: Minimum publication year filter (e.g., \"2025\" to show recent works only).\n\n### arXiv\n1. URL Examples:\n - https://arxiv.org/search/?query=artificial+intelligence&searchtype=all&abstracts=show&order=-announced_date_first\n - https://arxiv.org/search/?query=artificial+intelligence&searchtype=all&source=header\n2. Note: arXiv is an open-access repository for research preprints in fields such as computer science, physics, and mathematics. These URLs search for “artificial intelligence” papers across all fields, optionally displaying abstracts and sorting by date. Parameters like `order` and `size` control sorting and number of results shown per page.\n3. Parameters:\n - query: The main keyword being searched (e.g., \"artificial+intelligence\").\n - searchtype: Defines search scope (\"all\" searches across title, abstract, and authors).\n - abstracts: Controls whether abstracts are displayed (\"show\" = visible).\n - order: Sorting order (\"-announced_date_first\" = newest first).\n - size: Number of results displayed per page."
},
{
"type": "HumanMessagePromptTemplate",
"message": "=={{ $json.urls }}"
}
]
},
"promptType": "define"
},
"typeVersion": 1.7
},
{
"id": "7629b120-ac5b-4343-a1b7-52e50b1a1566",
"name": "定义搜索 URL",
"type": "n8n-nodes-base.set",
"position": [
96,
-896
],
"parameters": {
"options": {},
"assignments": {
"assignments": [
{
"id": "83b1c557-4853-4c63-b720-e5f2a7647ed3",
"name": "urls",
"type": "string",
"value": "={{[\n \"https://scholar.google.com/scholar?q=artificial+intelligence&hl=en&as_sdt=0,5\",\n \"https://scholar.google.com/scholar?as_ylo=2025&q=artificial+intelligence&hl=en&as_sdt=0,5\",\n \"https://arxiv.org/search/?query=artificial+intelligence&searchtype=all&abstracts=show&order=-announced_date_first\",\n \"https://arxiv.org/search/?query=artificial+intelligence&searchtype=all&source=header\",\n]}}"
}
]
}
},
"typeVersion": 3.4
},
{
"id": "220f1646-fba6-4af7-bc24-f04b0c46a567",
"name": "便签2",
"type": "n8n-nodes-base.stickyNote",
"position": [
16,
-1328
],
"parameters": {
"color": 3,
"width": 624,
"height": 704,
"content": "## 搜索 URL 洞察"
},
"typeVersion": 1
},
{
"id": "bf072114-c39e-43f4-b02e-ae235e837e04",
"name": "便签7",
"type": "n8n-nodes-base.stickyNote",
"position": [
-848,
-448
],
"parameters": {
"color": 5,
"width": 256,
"height": 256,
"content": "### 使用 BotFather 创建 Telegram 机器人并将其 API 密钥粘贴到 n8n 凭据中"
},
"typeVersion": 1
},
{
"id": "b75a52f9-7e68-4c70-9497-6934ff1495cb",
"name": "便签8",
"type": "n8n-nodes-base.stickyNote",
"position": [
544,
-528
],
"parameters": {
"color": 5,
"width": 384,
"height": 240,
"content": "### 将系统消息中的 _{{INPUT_SEARCH_URL_INSIGHTS}}_ 替换为您的搜索 URL 洞察（参见红色便签）"
},
"typeVersion": 1
}
],
"pinData": {
"Generate Search URL Insights": [
{
"text": "### Google Scholar\n1. URL Examples:\n - https://scholar.google.com/scholar?q=artificial+intelligence&hl=en&as_sdt=0,5\n - https://scholar.google.com/scholar?as_ylo=2025&q=artificial+intelligence&hl=en&as_sdt=0,5\n2. Note: Google Scholar is a specialized search engine for scholarly literature, including peer-reviewed papers, theses, books, preprints, abstracts, and technical reports from various academic publishers, professional societies, online repositories, and universities. Parameters allow specifying the main search query, filtering by publication year, and controlling the scope of the search to find relevant academic resources.\n3. Parameters:\n - q: The main search query (e.g., \"artificial+intelligence\").\n - hl: Interface language (e.g., \"en\" for English). While this primarily affects the user interface, it can sometimes influence the relevance of results in multi-lingual contexts.\n - as_sdt: Search scope control. \"0,5\" typically means \"articles and citations,\" including all scholarly documents.\n - as_ylo: Minimum publication year filter (e.g., \"2025\" to show works published in or after that year).\n\n### arXiv\n1. URL Examples:\n - https://arxiv.org/search/?query=artificial+intelligence&searchtype=all&abstracts=show&order=-announced_date_first\n - https://arxiv.org/search/?query=artificial+intelligence&searchtype=all&source=header\n2. Note: arXiv is an open-access repository of electronic preprints (e-prints) approved for posting after moderation, but not peer-reviewed. It covers fields such as physics, mathematics, computer science, quantitative biology, quantitative finance, statistics, electrical engineering and systems science, and economics. Parameters allow users to specify search terms, define the search scope (e.g., searching all fields, or specific fields like author or title), control display options like showing abstracts, and set the sorting order for results.\n3. Parameters:\n - query: The main keyword(s) or phrase being searched (e.g., \"artificial+intelligence\").\n - searchtype: Defines the fields to search within (e.g., \"all\" searches across title, abstract, and authors).\n - abstracts: Controls whether abstracts are displayed in the search results (\"show\" makes them visible).\n - order: Specifies the sorting order of the results (e.g., \"-announced_date_first\" sorts by the newest papers first)."
}
]
},
"connections": {
"Decodo": {
"ai_tool": [
[
{
"node": "Research Summary Agent",
"type": "ai_tool",
"index": 0
}
]
]
},
"Simple Memory": {
"ai_memory": [
[
{
"node": "Research Summary Agent",
"type": "ai_memory",
"index": 0
}
]
]
},
"Format Image Text": {
"main": [
[
{
"node": "Prepare Chat Data",
"type": "main",
"index": 0
}
]
]
},
"Format Voice Text": {
"main": [
[
{
"node": "Prepare Chat Data",
"type": "main",
"index": 0
}
]
]
},
"Prepare Chat Data": {
"main": [
[
{
"node": "Research Summary Agent",
"type": "main",
"index": 0
}
]
]
},
"Define Search URLs": {
"main": [
[
{
"node": "Generate Search URL Insights",
"type": "main",
"index": 0
}
]
]
},
"Start Telegram Bot": {
"main": [
[
{
"node": "Detect Message Type",
"type": "main",
"index": 0
}
]
]
},
"Detect Message Type": {
"main": [
[
{
"node": "Download Telegram Photo",
"type": "main",
"index": 0
}
],
[
{
"node": "Prepare Chat Data",
"type": "main",
"index": 0
}
],
[
{
"node": "Download Telegram Voice",
"type": "main",
"index": 0
}
],
[
{
"node": "Send Fallback Text",
"type": "main",
"index": 0
}
]
]
},
"Analyze Image Content": {
"main": [
[
{
"node": "Format Image Text",
"type": "main",
"index": 0
}
]
]
},
"Gemini Research Model": {
"ai_languageModel": [
[
{
"node": "Research Summary Agent",
"type": "ai_languageModel",
"index": 0
}
]
]
},
"Gemini URL Interpreter": {
"ai_languageModel": [
[
{
"node": "Generate Search URL Insights",
"type": "ai_languageModel",
"index": 0
}
]
]
},
"Research Summary Agent": {
"main": [
[
{
"node": "Check Telegram Message Length",
"type": "main",
"index": 0
}
]
]
},
"Download Telegram Photo": {
"main": [
[
{
"node": "Analyze Image Content",
"type": "main",
"index": 0
}
]
]
},
"Download Telegram Voice": {
"main": [
[
{
"node": "Transcribe Voice Message",
"type": "main",
"index": 0
}
]
]
},
"Transcribe Voice Message": {
"main": [
[
{
"node": "Format Voice Text",
"type": "main",
"index": 0
}
]
]
},
"Convert Output to Text File": {
"main": [
[
{
"node": "Send Research Summary File",
"type": "main",
"index": 0
}
]
]
},
"Check Telegram Message Length": {
"main": [
[
{
"node": "Convert Output to Text File",
"type": "main",
"index": 0
}
],
[
{
"node": "Send Research Summary Message",
"type": "main",
"index": 0
}
]
]
}
}
}常见问题
如何使用这个工作流？
复制上方的 JSON 配置代码，在您的 n8n 实例中创建新工作流并选择「从 JSON 导入」，粘贴配置后根据需要修改凭证设置即可。
这个工作流适合什么场景？
高级 - 文档提取, AI RAG 检索增强
需要付费吗？
本工作流完全免费，您可以直接导入使用。但请注意，工作流中使用的第三方服务（如 OpenAI API）可能需要您自行付费。
相关工作流推荐
LinkedIn和X病毒内容自动引擎
使用AI生成和发布自动创建LinkedIn和X的病毒内容
If
Set
Wait
+26
156 节点Diptamoy Barman
内容创作
使用Gemini、语音和图像生成构建多模态Telegram AI助手
使用Gemini、语音和图像生成构建多模态Telegram AI助手
If
Set
Code
+19
95 节点Iniyavan JC
个人效率
构建带记忆、Google套件和多AI研究成像的WhatsApp助手
构建带记忆、Google套件和多AI研究成像的WhatsApp助手
If
Set
Code
+27
71 节点Iniyavan JC
AI 聊天机器人
我的工作流
卡路里追踪与饮食记录（Telegram、Gemini AI和数据表）
If
Set
Code
+13
73 节点Gerald Denor
内容创作
用于捕获CRM联系人的Telegram机器人
使用Telegram和Gemini AI捕获和存储CRM联系人
If
Set
Code
+12
27 节点Vadim
营养追踪与饮食记录（Telegram、Gemini AI和Google Sheets）
营养追踪与饮食记录（Telegram、Gemini AI和Google Sheets）
If
Set
Code
+13
55 节点John Alejandro SIlva
AI 聊天机器人
