n8n 多智能体 SEO 博客与邮件通讯自动生成器

表单输入主题 → 四智能体流水线（研究→大纲→写作→编辑）→ 按内容类型路由 → 保存至 Airtable 或 Google Sheets

给谁

有 n8n 基础的内容创作者、数字营销人员、博客作者和中小企业团队，需要批量产出带 SEO 优化的博客文章或邮件通讯

你会得到

提交表单后自动完成研究、大纲、写作、编辑，最终生成博客（含 DALL-E 封面图）或邮件通讯，并保存到 Airtable 或 Google Sheets

你需要

n8n · Google Gemini · Google Sheets · Airtable

步骤

1. 触发：表单提交

   在 n8n 中导入工作流后，工作流默认以「On Form Submission」节点触发。填写表单字段：Content Type 选择 Blog 或 Newsletter；Topic 输入要撰写的主题。提交后自动进入后续智能体流水线。也可将此触发器替换为 Schedule、Webhook、Slack 等其他触发方式。【首次测试安全】首次测试前，先将 Save Blog to Airtable 和 Save Newsletter to Google Sheets 节点断开，或创建测试 Airtable base/测试 Google Sheet，确认输出正常后再接入生产环境。同时临时断开 Generate Featured Image（DALL-E）节点或跳过 Blog 图像生成分支，避免首次测试产生 OpenAI 图像费用；确认文案与写入无误后再接入。

2. 配置 AI 模型与运行智能体流水线

   1) 在 OpenRouter Chat Model 节点配置凭证 openRouterApi，模型选择 agentica-org/deepcoder-14b-preview:free，参数 maxTokens=4000、temperature=0.7。该模型为四个顺序 Agent（Research → Outline → Writer → Editor & SEO）提供推理能力。2) 在 Google Gemini Chat Model 节点配置凭证 googlePalmApi，供 Blog Publisher Agent 和 Newsletter Publisher Agent 使用。3) 四个 Agent 顺序执行：Research Agent 收集事实与标题；Outline Agent 生成带字数分配的 JSON 大纲；Writer Agent 扩展为 Medium 风格草稿；Editor & SEO Agent 润色并优化 SEO。每个 Agent 的提示词已内嵌于对应节点，见下方可复制提示词。

   运行提示词 · 来源原文。Research Agent 节点的 text 字段（promptType=define）

   复制提示词

   =Content Type: {{ $json['Content Type'] }}
   Topic: {{ $json.Topic }}
   
   You are the Research Agent, an expert in collecting and summarizing accurate, reliable, and up-to-date information for blog and newsletter writing.
   
   Your tasks:
   1. Take the topic from the user input
   2. Search for the most relevant articles, studies, and competitor content
   3. Extract key facts, statistics, definitions, and perspectives
   4. Identify 5–10 useful subtopics or angles worth covering
   5. Provide 3–5 strong title ideas and keyword phrases (natural, not stuffed)
   6. Cite sources with URLs and rate their reliability from 0–1.0
   
   Output format (JSON):
   {
     "title_ideas": ["Title 1", "Title 2", "Title 3"],
     "keywords": ["keyword1", "keyword2", "keyword3"],
     "facts": [
       {"fact": "Specific fact with numbers", "source": "https://source.com", "reliability": 0.9},
       {"fact": "Another fact", "source": "https://source2.com", "reliability": 0.85}
     ],
     "subtopics": ["Subtopic 1", "Subtopic 2", "Subtopic 3"]
   }

   运行提示词 · 来源原文。Outline Agent 节点的 text 字段（promptType=define）

   复制提示词

   ={{ $json.output }}
   
   You are the Outline Agent, responsible for designing a clear and engaging structure for a blog article or newsletter.
   
   Your tasks:
   1. Read the Research Agent's output (facts, subtopics, title ideas)
   2. Create a content outline with H1 (title), H2/H3 headings, and bullet points
   3. Allocate approximate word counts per section (aim for 1500-2500 words total)
   4. Suggest where visuals, stories, or examples should be added
   5. Ensure logical flow: hook → context → main points → conclusion → call-to-action
   
   Output format (JSON):
   {
     "chosen_title": "The best title from research",
     "outline": [
       {"heading": "Introduction", "level": "H2", "word_count": 200, "notes": "Include hook + brief context"},
       {"heading": "Main Section 1", "level": "H2", "word_count": 400, "notes": "Cover subtopic 1"},
       {"heading": "Subsection 1.1", "level": "H3", "word_count": 200, "notes": "Details"},
       {"heading": "Conclusion", "level": "H2", "word_count": 150, "notes": "Summarize + CTA"}
     ],
     "total_word_count": 1800
   }

   运行提示词 · 来源原文。Writer Agent 节点的 text 字段（promptType=define）

   复制提示词

   ={{ $json.output }}
   
   You are the Writer Agent, a professional Medium-style blog writer and newsletter creator.
   
   Your tasks:
   1. Expand the Outline Agent's structure into a full draft
   2. Use facts and sources from the Research Agent. Insert [citation] markers where needed
   3. Write in a natural, engaging style suitable for Medium:
      - Use short paragraphs (2–4 sentences)
      - Include examples, analogies, or short stories
      - Maintain clarity and flow
      - Avoid jargon unless explained
      - Use active voice
   4. Add a TL;DR at the top (2–3 sentences)
   5. End with a clear conclusion and call-to-action
   6. Target reading level: Grade 8–10
   
   Output: Complete markdown-formatted draft, ready for publication.

   运行提示词 · 来源原文。Editor & SEO Agent 节点的 text 字段（promptType=define）

   复制提示词

   ={{ $json.output }}
   
   You are the Editor & SEO Agent, responsible for polishing and optimizing the Writer Agent's draft.
   
   Your tasks:
   1. Improve grammar, clarity, and readability (target Grade 8–10 reading level)
   2. Adjust tone to match the target audience (casual, professional, or inspirational)
   3. Ensure smooth flow between sections, fix awkward sentences
   4. Optimize for discoverability on Medium and search engines:
      - Natural keyword integration (no stuffing)
      - Compelling meta description (<160 characters)
      - Suggest 5–7 relevant tags for Medium
      - Optimize headings for featured snippets
   5. Verify all citations are present and properly formatted
   6. Check for plagiarism risk and flag any concerns
   7. Ensure the draft follows Medium's formatting best practices:
      - H2/H3 headings
      - Bullet points and numbered lists
      - Bold/italic for emphasis
      - Code blocks if needed
   
   Output: Final polished markdown content + metadata (meta description, tags).

3. 配置内容路由规则（Switch 节点）

   导入模板后，「Route by Content Type」Switch 节点的判断条件在 JSON 中 leftValue 与 rightValue 均为空字符串，必须手动配置：添加规则使 Content Type equals Blog 时走 Blog Publisher Agent 分支，Content Type equals Newsletter 时走 Newsletter Publisher Agent 分支。若未配置，工作流无法正确分流。

4. 博客分支：生成封面图并保存至 Airtable

   Blog 分支流程：1) Blog Publisher Agent 接收 Editor 输出，提取标题、正文、关键词、meta description，并生成 DALL-E 图像提示词（JSON 输出）。2)「Generate Featured Image (DALL-E)」节点通过 HTTP Request 调用 OpenAI images/generations，模型 dall-e-3，尺寸 1792x1024，质量 standard，返回 b64_json。3)「Convert Image to File」节点将 b64_json 转为 PNG 文件，文件名格式 blog-featured-{yyyy-MM-dd-HHmmss}.png。4)「Save Blog to Airtable」节点创建记录，列映射如下：Title ← Blog Publisher Agent 的 title_column；Status = Draft；Content ← blog_column；Keywords ← keywords；Created Date ← $now.toISO()；Meta Description ← meta_description。需提前将 base/table 的占位符 YOUR_AIRTABLE_BASE_ID / YOUR_AIRTABLE_TABLE_ID 替换为实际 ID。

   运行提示词 · 来源原文。Blog Publisher Agent 节点的 text 字段（promptType=define）

   复制提示词

   ={{ $json.output }}
   
   You are the Blog Publisher Agent. Format the edited content for Airtable storage.
   
   Your tasks:
   1. Extract the final blog title (without markdown formatting)
   2. Extract the full blog content in clean markdown format
   3. Extract keywords and meta description
   4. Generate a concise image prompt for DALL-E (describe the visual that would best represent this blog)
   
   Output format (JSON):
   {
     "title_column": "Blog Title Here",
     "blog_column": "Full markdown content here...",
     "keywords": "keyword1, keyword2, keyword3",
     "meta_description": "Short description under 160 chars",
     "image_prompt": "A professional, high-quality image showing [describe scene]"
   }

5. 邮件通讯分支：格式化并保存至 Google Sheets

   Newsletter 分支流程：Newsletter Publisher Agent 接收 Editor 输出，提取标题、格式化为 HTML 与纯文本版本、生成预览文本（JSON 输出）。随后「Save Newsletter to Google Sheets」节点追加到指定工作表，列映射如下：Title ← title_column；Status = Ready to Send；Content HTML ← content_html；Created Date ← $now.toFormat('yyyy-MM-dd HH:mm:ss')；Preview Text ← preview_text；Content Plain ← content_plain。需提前将 documentId 占位符 YOUR_GOOGLE_SHEET_ID 替换为实际 ID，并确认 sheetName。

   运行提示词 · 来源原文。Newsletter Publisher Agent 节点的 text 字段（promptType=define）

   复制提示词

   ={{ $json.output }}
   
   You are the Newsletter Publisher Agent. Format the edited content for email newsletter distribution.
   
   Your tasks:
   1. Extract the newsletter title
   2. Format content sections for email (simpler formatting than web)
   3. Add email-friendly formatting:
      - Clear section breaks
      - Shorter paragraphs
      - Prominent CTAs
   4. Create plain text version
   
   Output format (JSON):
   {
     "title_column": "Newsletter Title",
     "content_html": "HTML formatted content",
     "content_plain": "Plain text version",
     "preview_text": "Email preview text (50 chars)"
   }

你是一个 n8n 工作流搭建助手。用户希望复现「多智能体 SEO 博客与邮件通讯生成器」模板。请先确认用户已具备：可访问的 n8n 实例、OpenRouter API 密钥、Google Gemini API 密钥、OpenAI API 密钥、Airtable 账户、Google Sheets 访问权限。然后指导用户按以下顺序操作：1) 从 n8n 模板库导入工作流 ID 10293 或上传 workflow.json；2) 在 OpenRouter Chat Model 节点确认模型为 agentica-org/deepcoder-14b-preview:free（或按需求替换为其他模型）并配置 maxTokens/temperature；3) 在 Google Gemini Chat Model 节点配置凭证；4) 为四个 Agent 节点确认已连接对应语言模型；5) 配置 Switch 节点的 Content Type 判断规则；6) 替换 Airtable 和 Google Sheets 的占位符 ID；7) 测试表单提交。若用户想替换模型（如改用 Claude 或 GPT-4），建议逐个替换语言模型节点并调整 prompt 中的风格指令。