マルチエージェントフィードバックループ
中級
これはContent Creation, Multimodal AI分野の自動化ワークフローで、13個のノードを含みます。主にIf, Set, Code, ManualTrigger, SplitInBatchesなどのノードを使用。 GPT-4マルチエージェントフィードバックシステムに基づく反復コンテンツ最適化
前提条件
- •OpenAI API Key
ワークフロープレビュー
ノード接続関係を可視化、ズームとパンをサポート
ワークフローをエクスポート
以下のJSON設定をn8nにインポートして、このワークフローを使用できます
{
"id": "rdQbhY9lEyDizxvy",
"meta": {
"instanceId": "205780d3749915e5b72849aad8492ed96140f3a3cf0d7f077a27e6df51ded409",
"templateCredsSetupCompleted": true
},
"name": "Multi-Agent Feedback Loops",
"tags": [
{
"id": "o58aq7CrdbXQHYA0",
"name": "template",
"createdAt": "2025-06-14T10:20:41.387Z",
"updatedAt": "2025-06-14T10:20:41.387Z"
}
],
"nodes": [
{
"id": "74337855-c422-4186-9510-ce899d67e39f",
"name": "AIエージェント",
"type": "@n8n/n8n-nodes-langchain.agent",
"position": [
-1180,
240
],
"parameters": {
"text": "=Product: {{$json[\"productDescription\"]}}",
"options": {
"systemMessage": "You are a creative branding agent. Based on this product description, generate 3 distinct name + tagline options.\n\nReturn your output in this format:\n[\n { \"name\": \"Name1\", \"tagline\": \"Tagline1\" },\n { \"name\": \"Name2\", \"tagline\": \"Tagline2\" },\n { \"name\": \"Name3\", \"tagline\": \"Tagline3\" }\n]\n"
},
"promptType": "define"
},
"typeVersion": 2
},
{
"id": "e056b2fa-3e6f-4fa7-b908-dba91d01ec61",
"name": "アイテムをループ処理",
"type": "n8n-nodes-base.splitInBatches",
"position": [
-780,
240
],
"parameters": {
"options": {
"reset": true
}
},
"typeVersion": 3
},
{
"id": "f7c4971f-1ea8-4448-ab73-35d945d76604",
"name": "条件分岐",
"type": "n8n-nodes-base.if",
"position": [
-360,
360
],
"parameters": {
"options": {},
"conditions": {
"options": {
"version": 2,
"leftValue": "",
"caseSensitive": true,
"typeValidation": "strict"
},
"combinator": "or",
"conditions": [
{
"id": "595ebd64-59de-470c-9827-24d204db58f9",
"operator": {
"type": "number",
"operation": "equals"
},
"leftValue": "={{ $json.turn }}",
"rightValue": 5
},
{
"id": "fb2ec186-4676-4e91-85b1-af7ce745b203",
"operator": {
"name": "filter.operator.equals",
"type": "string",
"operation": "equals"
},
"leftValue": "={{ $json.done }}",
"rightValue": "Yes"
}
]
}
},
"typeVersion": 2.2
},
{
"id": "b8c57f43-6f23-4e05-81c0-6c1db0264f18",
"name": "批評エージェント",
"type": "@n8n/n8n-nodes-langchain.agent",
"position": [
-120,
420
],
"parameters": {
"text": "=Ideas:\n{{JSON.stringify($('Edit Fields').item.json.ideas)}}",
"options": {
"systemMessage": "You are a critical branding analyst. For each name + tagline below, provide a short flaw summary and a suggestion for improvement."
},
"promptType": "define"
},
"typeVersion": 2
},
{
"id": "cdcab121-8082-4295-aa7f-475bde50414e",
"name": "改良エージェント",
"type": "@n8n/n8n-nodes-langchain.agent",
"position": [
320,
520
],
"parameters": {
"text": "=**Original Ideas:** {{ JSON.stringify($('Edit Fields').item.json.ideas) }}\n\n**Critics Feedback:** {{ $json.output }}",
"options": {
"systemMessage": "You are a refiner. Use the original ideas and the critic’s feedback to create 3 improved name + tagline combinations.\n\nBe concise, clever, and impactful.\n\nReturn your output in this format:\n[\n { \"name\": \"Name1\", \"tagline\": \"Tagline1\" },\n { \"name\": \"Name2\", \"tagline\": \"Tagline2\" },\n { \"name\": \"Name3\", \"tagline\": \"Tagline3\" }\n]\n"
},
"promptType": "define"
},
"typeVersion": 2
},
{
"id": "9a63dab8-269c-4073-bdf5-8661fcbfdc82",
"name": "フィールド編集",
"type": "n8n-nodes-base.set",
"position": [
-560,
360
],
"parameters": {
"options": {},
"assignments": {
"assignments": [
{
"id": "1c6c6fc2-96de-469a-b30f-7ffae52c9847",
"name": "ideas",
"type": "array",
"value": "={{ $json.output }}"
},
{
"id": "1a857aee-ae3f-4e81-a09c-a9e83edd9b98",
"name": "turn",
"type": "number",
"value": "={{ $json.turn || 1}}"
},
{
"id": "d310985b-59a3-44b8-aa5a-e77f2b352ae4",
"name": "done",
"type": "string",
"value": "={{ $json.done || \"No\"}}"
}
]
}
},
"typeVersion": 3.4
},
{
"id": "cdf532b0-6dd9-43e6-a1b2-76edb177e9fa",
"name": "コード",
"type": "n8n-nodes-base.code",
"position": [
1180,
880
],
"parameters": {
"jsCode": "const ideas = $json.output.ideas || [];\nconst currentTurn = $json.output.turn ?? 0;\nconst currentDone = $json.output.done || \"No\";\n\nconst nextTurn = currentTurn + 1;\n\nreturn [\n {\n json: {\n ideas,\n turn: nextTurn,\n done: currentDone\n }\n }\n];\n"
},
"typeVersion": 2
},
{
"id": "6d853089-99d1-4ee9-bd80-34ebcc2fd5a3",
"name": "フィールド編集1",
"type": "n8n-nodes-base.set",
"position": [
1400,
880
],
"parameters": {
"options": {},
"assignments": {
"assignments": [
{
"id": "21a77915-3d73-450d-bc9c-f655471a3b56",
"name": "output",
"type": "array",
"value": "={{ $json.ideas }}"
},
{
"id": "6dee88ae-c3f3-464a-a3cc-179bd28b7c5e",
"name": "turn",
"type": "number",
"value": "={{ $json.turn }}"
},
{
"id": "2f2caa57-797c-4fa3-9126-da46777a2b63",
"name": "done",
"type": "string",
"value": "={{ $json.done }}"
}
]
}
},
"typeVersion": 3.4
},
{
"id": "3702b8fc-61e2-4d01-aa10-0ddb5bff48c7",
"name": "OpenAI チャットモデル",
"type": "@n8n/n8n-nodes-langchain.lmChatOpenAi",
"position": [
-100,
800
],
"parameters": {
"model": {
"__rl": true,
"mode": "list",
"value": "gpt-4.1-mini",
"cachedResultName": "gpt-4.1-mini"
},
"options": {}
},
"credentials": {
"openAiApi": {
"id": "oPvbzpxkrzwKpKeY",
"name": "OpenAi account"
}
},
"typeVersion": 1.2
},
{
"id": "17809998-4a9c-41d9-984e-c17cc24e7769",
"name": "構造化出力パーサー",
"type": "@n8n/n8n-nodes-langchain.outputParserStructured",
"position": [
880,
820
],
"parameters": {
"jsonSchemaExample": "{\n\t\"ideas\": [\n { \"name\": \"EnterpriseMind\", \"tagline\": \"Harness AI to transform your business ideas into reality.\" },\n { \"name\": \"InnoLogix\", \"tagline\": \"Capture and evolve your entrepreneurial ideas with AI-powered journaling.\" },\n { \"name\": \"VentureScript\", \"tagline\": \"Document and accelerate your startup journey with intelligent AI journaling.\" }],\n \"done\": \"Yes\",\n \"turn\": 2\n}"
},
"typeVersion": 1.2
},
{
"id": "b47e29f2-bd26-4e9f-813f-2983b1841d33",
"name": "ワークフロー実行時",
"type": "n8n-nodes-base.manualTrigger",
"position": [
-1360,
240
],
"parameters": {},
"typeVersion": 1
},
{
"id": "7d996197-5542-4d89-8c58-ced050c7fffc",
"name": "付箋",
"type": "n8n-nodes-base.stickyNote",
"position": [
-1920,
120
],
"parameters": {
"color": 5,
"width": 3920,
"height": 1160,
"content": "# ⚙️ Setup Guide\n## This is a workflow for user's customization.\n### 🧑💻 Author: [Sebastian/OptiLever](www.linkedin.com/in/sebastian-9ab9b9242)\n\n### 🚀 Prerequisite:\n\n**1. Connect to Openai API account.**\n\n\n---\n\n### 🚀 Steps to Connect:\n\n\n1. **Edit Fields Node Setup (Initial)**\n- Locate the **\"Edit Fields\"** node at the beginning of the workflow.\n- **Edit the following fields:**\n- Ensure the input field **\"ideas\"** is set as an array.\n- Add a field **\"turn\"** to track the number of loop iterations (default value: 1).\n- Add a field **\"done\"** to track if the evaluation is complete (default value: false).\n\n\n\n2. **Critic Agent Setup**\n- Configure the **\"Critic Agent\"** node to analyze the input and provide feedback.\n- Ensure the output is structured to include suggestions for improvements.\n\n\n\n3. **Refiner Agent Setup**\n- Configure the **\"Refiner Agent\"** node to use the original input and critic feedback.\n- Ensure the output is an improved version of the input (e.g., three improved names and taglines).\n\n\n\n4. **Evaluation Agent Setup**\n- Configure the **\"Evaluation Agent\"** node to rank the refined outputs.\n- Ensure the output includes a **\"done\"** field indicating if the result is satisfactory.\n\n\n\n5. **Edit Fields Node Setup (Final)**\n- Ensure the final **\"Edit Fields\"** node matches the structure of the initial input.\n- This ensures the loop can process the data correctly in each iteration.\n\n\n\n6. **If Node Setup**\n- Configure the **\"If\"** node to check if **\"turn\"** exceeds a threshold (e.g., 5) or if **\"done\"** is true.\n- If true, exit the loop and proceed to the next step in the workflow.\n\n\n\n7. **Optional: Additional AI Agents**\n- Add additional AI agents after the loop to further process the final output.\n\n\n---\n\n\n"
},
"typeVersion": 1
},
{
"id": "d64015c8-17f1-4fcc-a12f-f29d2ac0aa75",
"name": "評価エージェント",
"type": "@n8n/n8n-nodes-langchain.agent",
"position": [
700,
640
],
"parameters": {
"text": "=**Results To Examine:** {{ $json.output }}\n\n**Turns Through Loop:** {{ $('Edit Fields').item.json.turn }}",
"options": {
"systemMessage": "You are an evaluator. Rank the 3 improved name ideas and taglines. Recommend if the names and taglines are so awesome that it would make anyone want to purchase from this company, or further improvement is needed (Yes/No). Output a JSON that follows this format:\n{\n\t\"ideas\": [\n { \"name\": \"EnterpriseMind\", \"tagline\": \"Harness AI to transform your business ideas into reality.\" },\n { \"name\": \"InnoLogix\", \"tagline\": \"Capture and evolve your entrepreneurial ideas with AI-powered journaling.\" },\n { \"name\": \"VentureScript\", \"tagline\": \"Document and accelerate your startup journey with intelligent AI journaling.\" }],\n \"done\": \"Yes\",\n \"turn\": 2\n}"
},
"promptType": "define",
"hasOutputParser": true
},
"typeVersion": 2
}
],
"active": false,
"pinData": {},
"settings": {
"executionOrder": "v1"
},
"versionId": "40c2c451-49fd-4266-94e1-3b66cabf6ebc",
"connections": {
"f7c4971f-1ea8-4448-ab73-35d945d76604": {
"main": [
[],
[
{
"node": "b8c57f43-6f23-4e05-81c0-6c1db0264f18",
"type": "main",
"index": 0
}
]
]
},
"cdf532b0-6dd9-43e6-a1b2-76edb177e9fa": {
"main": [
[
{
"node": "6d853089-99d1-4ee9-bd80-34ebcc2fd5a3",
"type": "main",
"index": 0
}
]
]
},
"74337855-c422-4186-9510-ce899d67e39f": {
"main": [
[
{
"node": "e056b2fa-3e6f-4fa7-b908-dba91d01ec61",
"type": "main",
"index": 0
}
]
]
},
"9a63dab8-269c-4073-bdf5-8661fcbfdc82": {
"main": [
[
{
"node": "f7c4971f-1ea8-4448-ab73-35d945d76604",
"type": "main",
"index": 0
}
]
]
},
"b8c57f43-6f23-4e05-81c0-6c1db0264f18": {
"main": [
[
{
"node": "cdcab121-8082-4295-aa7f-475bde50414e",
"type": "main",
"index": 0
}
]
]
},
"6d853089-99d1-4ee9-bd80-34ebcc2fd5a3": {
"main": [
[
{
"node": "e056b2fa-3e6f-4fa7-b908-dba91d01ec61",
"type": "main",
"index": 0
}
]
]
},
"cdcab121-8082-4295-aa7f-475bde50414e": {
"main": [
[
{
"node": "d64015c8-17f1-4fcc-a12f-f29d2ac0aa75",
"type": "main",
"index": 0
}
]
]
},
"e056b2fa-3e6f-4fa7-b908-dba91d01ec61": {
"main": [
[],
[
{
"node": "9a63dab8-269c-4073-bdf5-8661fcbfdc82",
"type": "main",
"index": 0
}
]
]
},
"d64015c8-17f1-4fcc-a12f-f29d2ac0aa75": {
"main": [
[
{
"node": "cdf532b0-6dd9-43e6-a1b2-76edb177e9fa",
"type": "main",
"index": 0
}
]
]
},
"3702b8fc-61e2-4d01-aa10-0ddb5bff48c7": {
"ai_languageModel": [
[
{
"node": "74337855-c422-4186-9510-ce899d67e39f",
"type": "ai_languageModel",
"index": 0
},
{
"node": "b8c57f43-6f23-4e05-81c0-6c1db0264f18",
"type": "ai_languageModel",
"index": 0
},
{
"node": "cdcab121-8082-4295-aa7f-475bde50414e",
"type": "ai_languageModel",
"index": 0
},
{
"node": "d64015c8-17f1-4fcc-a12f-f29d2ac0aa75",
"type": "ai_languageModel",
"index": 0
}
]
]
},
"17809998-4a9c-41d9-984e-c17cc24e7769": {
"ai_outputParser": [
[
{
"node": "d64015c8-17f1-4fcc-a12f-f29d2ac0aa75",
"type": "ai_outputParser",
"index": 0
}
]
]
},
"b47e29f2-bd26-4e9f-813f-2983b1841d33": {
"main": [
[
{
"node": "74337855-c422-4186-9510-ce899d67e39f",
"type": "main",
"index": 0
}
]
]
}
}
}よくある質問
このワークフローの使い方は?
上記のJSON設定コードをコピーし、n8nインスタンスで新しいワークフローを作成して「JSONからインポート」を選択、設定を貼り付けて認証情報を必要に応じて変更してください。
このワークフローはどんな場面に適していますか?
中級 - コンテンツ作成, マルチモーダルAI
有料ですか?
このワークフローは完全無料です。ただし、ワークフローで使用するサードパーティサービス(OpenAI APIなど)は別途料金が発生する場合があります。
関連ワークフロー
コンテンツジェネレーター v3
AI驱动ブログ自動化:使用GPT-4生成并公開SEO記事至WordPressとTwitter
If
Set
Code
+
If
Set
Code
144 ノードJay Emp0
コンテンツ作成
YouTube 動画に基づく自律ブログ公開
YouTube 動画から ChatGPT、Sheets、Apify、Pexels、WordPress を使用してブログの自主公開
If
Set
Code
+
If
Set
Code
80 ノードOriol Seguí
コンテンツ作成
Gemini_NanoBanana_テンプレート
Google SheetsからFal.aiのモデル(nano-banana、WAN2.2、Veo3)でUGC広告生成
If
Set
Code
+
If
Set
Code
36 ノードJaruphat J.
コンテンツ作成
AI製品コピーライター(Shopify、Google Merchant Center向けのSEOタイトルと説明文)
AI を使って Shopify および Google Shopping に SEO 最適化された製品説明文を生成
If
Edit Image
Http Request
+
If
Edit Image
Http Request
17 ノードSebastian/OptiLever
コンテンツ作成
Apollo データスクレイピングとタッチアウトフロー 1 ✅
Apollo、AI による解析と計画されたメール.follow-up によるリード生成の自動化
If
Code
Wait
+
If
Code
Wait
39 ノードDeniz
コンテンツ作成
WordPressブログの自動化プロフェッショナル版(先端研究)v2.1マーケットプラグイン
GPT-4o、Perplexity AI、そして多言語対応を使ったSEO最適化ブログ作成の自動化
If
Set
Xml
+
If
Set
Xml
125 ノードDaniel Ng
コンテンツ作成