用 n8n 自动审查合同风险并重写条款

Google Drive 上传 PDF 合同 → Groq LLM 提取条款 → 与 Notion 标准库比对 → 生成风险报告与改写建议 → Slack 告警高风险合同 → Notion 存档

给谁

法务、采购或合规团队，需要批量审查供应商合同并与公司标准条款比对

你会得到

合同上传后自动提取关键条款、评估风险等级、生成谈判建议与改写文本，高风险合同触发 Slack 告警并归档至 Notion

你需要

n8n · Google Drive · Notion · Slack

步骤

1. 配置合同上传触发与文本提取

   在 n8n 中配置 Google Drive Trigger 节点，选择 specificFolder 模式并指定监听文件夹。当有新文件上传时，Download File 节点下载文件，Check if PDF 节点判断 MIME 类型；若为 PDF 则经 Extract content from PDF 提取文本，否则经 Extract Content 代码节点将二进制转为文本。输出包含 fileContent 的 JSON 供下游使用。

2. 提取条款并与 Notion 标准库比对

   将合同文本传入 Extract Contract Clauses（Basic LLM Chain）节点，连接 Groq Chat Model3（如 llama-3.3-70b-versatile）和 Structured Output Parser3，强制输出 payment_terms、penalties、sla、termination、renewal、liability、data_protection 七个字段的 JSON。同时 Fetch Standard Clauses 节点从 Notion Standard Clause Library 数据库拉取标准条款。Merge Vendor + Standards 节点合并两路数据后，Compare Vendor vs Standard Clauses 代码节点按 clause type 逐条比对，输出偏差分析。

   运行提示词 · 来源原文。Extract Contract Clauses 节点的 text 参数（n8n Basic LLM Chain）

   复制提示词

   =From the contract below extract:
   
   - Payment Terms
   - Late Payment Penalties
   - SLA / Service Levels
   - Termination Clause
   - Renewal Terms
   - Liability Clause
   - Data Protection Clause
   
   Return JSON format:
   
   {
    "payment_terms":"",
    "penalties":"",
    "sla":"",
    "termination":"",
    "renewal":"",
    "liability":"",
    "data_protection":""
   }
   
   Contract:
   {{ $json.fileContent || $json.text }}

   运行提示词 · 来源原文。Extract Contract Clauses 节点的 messages.messageValues[0].message（system message）

   复制提示词

   You are a legal contract analyst. Extract key procurement clauses from contracts. Return STRICT JSON.

3. 生成法律风险分析报告

   将比对结果传入 AI Risk Analysis & Executive Report 节点，连接 Groq Chat Model4（如 openai/gpt-oss-120b）和 Structured Output Parser4，生成包含 overall_risk_level、executive_summary、key_risks、clause_risk_review、missing_or_weak_protections、negotiation_recommendations、final_recommendation 的结构化报告。随后 Prepare Clause Comparison Dataset 代码节点将风险等级关联到每条条款，为后续改写做准备。

   运行提示词 · 来源原文。AI Risk Analysis & Executive Report 节点的 text 参数（n8n Basic LLM Chain）

   复制提示词

   =Here is the clause comparison data:
   
   {{ JSON.stringify($json.comparison, null, 2) }}
   
   Create a Legal Risk Report with these sections:
   
   1. Executive Summary
   - Overall contract risk level (Low / Medium / High)
   - Short explanation why
   
   2. Key Risk Findings
   Explain the biggest risks and their impact.
   
   3. Clause-by-Clause Risk Review
   For each deviating clause:
   - Why it is risky
   - Possible business impact
   
   4. Missing or Weak Protections
   Identify gaps vs standard protections.
   
   5. Negotiation Recommendations
   Give clear actions procurement/legal should request.
   
   End with a short conclusion suitable for leadership.

   运行提示词 · 来源原文。AI Risk Analysis & Executive Report 节点的 messages.messageValues[0].message（system message）

   复制提示词

   You are a senior legal counsel reviewing a vendor contract.  You receive structured clause comparison data between: - Company standard clauses - Vendor contract clauses  Your task: Produce a professional legal risk report for executives.  Write in clear business English. Be concise but insightful. Do NOT output JSON. Use formatted sections.

4. AI 改写中高风险条款

   Loop Through Clauses 节点逐条遍历条款，Check If Clause Needs Rewrite 节点过滤 risk_level 非 Low 的条款。对需改写的条款，AI Rewrite Risky Clauses 节点连接 LLM（如 meta-llama/llama-4-scout-17b-16e-instruct）和 Parse Rewritten Clause Output 结构化解析器，逐条输出 clause_type 与 revised_clause。Combine All Revised Clauses 代码节点汇总所有改写结果。

   运行提示词 · 来源原文。AI Rewrite Risky Clauses 节点的 text 参数（n8n Basic LLM Chain）

   复制提示词

   =You are a senior procurement and legal expert.
   
   Your task is to revise the vendor clause so that it fully aligns with the company’s standard clause and risk requirements.
   
   Instructions:
   - Compare the vendor clause against the company standard clause.
   - Identify gaps or deviations.
   - Rewrite the clause so it aligns with company policy.
   - Preserve clear, enforceable legal language.
   - Do NOT summarize.
   - Do NOT explain.
   - Return only valid JSON.
   
   Return JSON format:
   {
     "clause_type": "",
     "risk_level": "",
     "revised_clause": ""
   }
   
   Clause Type:
   {{ $json.clause_type }}
   
   Risk Level:
   {{ $json.risk_level }}
   
   Vendor Clause:
   {{ $json.vendor_clause }}
   
   Company Standard Clause:
   {{ $json.standard_clause }}

   运行提示词 · 来源原文。AI Rewrite Risky Clauses 节点的 messages.messageValues[0].message（system message；注意：来源此处的 system message 与 AI Risk Analysis 节点相同，实际用途为 rewrite 节点共用模板配置）

   复制提示词

   You are a senior legal counsel reviewing a vendor contract.  You receive structured clause comparison data between: - Company standard clauses - Vendor contract clauses  Your task: Produce a professional legal risk report for executives.  Write in clear business English. Be concise but insightful. Do NOT output JSON. Use formatted sections.

5. 格式化记录并判断风险等级

   Format Final Contract Review Record（Edit Fields）节点从上游汇总 revised_clauses、overall_risk_level、executive_summary、key_risks、negotiation_recommendations、final_recommendation 等字段。Check If Contract Is Risky 节点判断 overall_risk_level 是否等于 High，决定后续是否发送 Slack 告警。

6. Slack 告警与 Notion 归档

   若风险等级为 High，Send Slack Risk Alert 节点向指定频道发送格式化消息，包含风险等级、执行摘要、关键风险、谈判建议及建议改写条款。随后 Create Contract Review Record 节点在 Notion Contract Reviews 数据库创建记录，写入 Risk Level、Executive Summary、Key Risks、Recommendations、Final Recommendation、Alternative Clauses 等属性。若风险非 High，则直接创建 Notion 记录而不发送 Slack。

   运行提示词 · 来源原文。Send Slack Risk Alert 节点的 text 参数（n8n Slack 节点消息内容）

   复制提示词

   =*HIGH RISK CONTRACT DETECTED*
   
   *Overall Risk Level:* {{ $('Format Final Contract Review Record').item.json.overall_risk_level }}
   *Executive Summary:*
   {{ $('Format Final Contract Review Record').item.json.executive_summary }}
   
   *Key Risks:*
   {{ $('AI Risk Analysis & Executive Report').item.json.output.key_risks.join("\n• ") }}
   
   *Top Recommendations:*
   {{ $('AI Risk Analysis & Executive Report').item.json.output.negotiation_recommendations.join("\n• ") }}
   
   *Suggested Clause Rewrites*
   {{ Object.entries($items('Format Final Contract Review Record')[0].json.revised_clauses)
     .map(([key, value]) => 
       "• *" + key.replace(/_/g," ").replace(/\b\w/g,l=>l.toUpperCase()) + "*\n" + value
     ).join("\n\n") }}

你是一个 n8n 工作流搭建助手。用户希望复现「合同风险审查与条款改写」自动化流程。请按以下顺序协助：
1. 确认用户已具备 n8n 实例、Google Drive OAuth、Groq API key、Notion 集成、Slack OAuth。
2. 指导用户在 Notion 中创建 Standard Clause Library 数据库（字段：property_clause_type 为 select，property_standard_text 为 rich_text，property_risk_level_if_deviates 为 select）和 Contract Reviews 数据库（字段：Risk Level、Executive Summary、Key Risks、Recommendations、Final Recommendation、Alternative Clauses）。
3. 建议用户导入 workflow.json 或按步骤手动搭建，重点检查 Groq Chat Model 的 model 参数和 Structured Output Parser 的 JSON schema。
4. 提醒 PDF 需为可搜索文本，扫描件需前置 OCR 节点。
5. 建议先用一份测试合同 PDF 验证端到端流程，检查 Notion 字段名大小写是否完全匹配。