評価指標例：RAGドキュメント関連性
上級
これはEngineering, AI分野の自動化ワークフローで、26個のノードを含みます。主にSet, Evaluation, GoogleSheets, ManualTrigger, Agentなどのノードを使用、AI技術を活用したスマート自動化を実現。 評価指標サンプル：RAGドキュメント関連性
前提条件
- •Google Sheets API認証情報
- •OpenAI API Key
使用ノード (26)
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
{
"meta": {
"instanceId": "bf40384a063e00f3b983f4f9bada22b57a8231a04c0fb48d363e26d7b0f2b7e7",
"templateCredsSetupCompleted": true
},
"nodes": [
{
"id": "86d6a658-f056-4434-9720-dfd3ca514cb3",
"name": "付箋1",
"type": "n8n-nodes-base.stickyNote",
"position": [
420,
20
],
"parameters": {
"color": 7,
"width": 300,
"height": 180,
"content": "Check whether the documents returned a relevant to the question"
},
"typeVersion": 1
},
{
"id": "90aac144-164f-4961-89f8-4b8402177202",
"name": "付箋3",
"type": "n8n-nodes-base.stickyNote",
"position": [
-1220,
-40
],
"parameters": {
"width": 200,
"height": 500,
"content": "## How it works\nThis template shows how to calculate a workflow evaluation metric: **retrieved document relevance** (i.e. whether the information retrieved from the vector store is relevant to the question).\n\nThe workflow takes a question and checks whether the information retrieved to answer it is relevant.\n\nYou can find more information on workflow evaluation [here](https://docs.n8n.io/advanced-ai/evaluations/overview), and other metric examples [here](https://docs.n8n.io/advanced-ai/evaluations/metric-based-evaluations/#2-calculate-metrics)."
},
"typeVersion": 1
},
{
"id": "f3970d8c-b205-4348-b375-d919d2f56f94",
"name": "付箋4",
"type": "n8n-nodes-base.stickyNote",
"position": [
-960,
280
],
"parameters": {
"color": 7,
"width": 220,
"height": 220,
"content": "Read in [this test dataset](https://docs.google.com/spreadsheets/d/1uuPS5cHtSNZ6HNLOi75A2m8nVWZrdBZ_Ivf58osDAS8/edit?gid=630157309#gid=630157309) of questions"
},
"typeVersion": 1
},
{
"id": "677b90c8-aa6f-4cc5-88df-d8e635b0bde2",
"name": "データセット行取得時",
"type": "n8n-nodes-base.evaluationTrigger",
"position": [
-900,
340
],
"parameters": {
"sheetName": {
"__rl": true,
"mode": "url",
"value": "https://docs.google.com/spreadsheets/d/1uuPS5cHtSNZ6HNLOi75A2m8nVWZrdBZ_Ivf58osDAS8/edit?gid=630157309#gid=630157309"
},
"documentId": {
"__rl": true,
"mode": "url",
"value": "https://docs.google.com/spreadsheets/d/1uuPS5cHtSNZ6HNLOi75A2m8nVWZrdBZ_Ivf58osDAS8/edit?gid=630157309#gid=630157309"
}
},
"credentials": {
"googleSheetsOAuth2Api": {
"id": "bpr2LoSELMlxpwnN",
"name": "Google Sheets account David"
}
},
"typeVersion": 4.6
},
{
"id": "c381a3c3-12b6-44fe-a2ed-6adf6f997534",
"name": "評価中？",
"type": "n8n-nodes-base.evaluation",
"position": [
-60,
200
],
"parameters": {
"operation": "checkIfEvaluating"
},
"typeVersion": 4.6
},
{
"id": "e1cf0122-c668-41d2-9316-a173d425d897",
"name": "チャットメッセージ受信時",
"type": "@n8n/n8n-nodes-langchain.chatTrigger",
"position": [
-900,
100
],
"webhookId": "d53bc846-db11-446c-8436-9d25dc265663",
"parameters": {
"options": {}
},
"typeVersion": 1.1
},
{
"id": "b521aa40-c8da-4485-b834-50a129c27d3c",
"name": "チャット形式に一致",
"type": "n8n-nodes-base.set",
"position": [
-680,
340
],
"parameters": {
"options": {},
"assignments": {
"assignments": [
{
"id": "93f89095-7918-45ad-aa74-a0bbcf0d5788",
"name": "chatInput",
"type": "string",
"value": "={{ $json.question }}"
}
]
}
},
"typeVersion": 3.4
},
{
"id": "d0170cbd-ba01-45db-a405-aec943bf1608",
"name": "付箋",
"type": "n8n-nodes-base.stickyNote",
"position": [
-80,
100
],
"parameters": {
"color": 7,
"width": 150,
"height": 260,
"content": "Only calculate metrics if we're evaluating, to reduce costs"
},
"typeVersion": 1
},
{
"id": "a21f98f0-25eb-44cc-a04a-33181d7e5288",
"name": "チャット応答を返す",
"type": "n8n-nodes-base.noOp",
"position": [
220,
340
],
"parameters": {},
"typeVersion": 1
},
{
"id": "fd8628a1-9f31-4f42-88df-b87eda7216ee",
"name": "メトリクスを設定",
"type": "n8n-nodes-base.evaluation",
"position": [
820,
80
],
"parameters": {
"metrics": {
"assignments": [
{
"id": "230589eb-34c8-4d10-9296-4a78d673077a",
"name": "similarity",
"type": "number",
"value": "={{ $json.message.content.score }}"
}
]
},
"operation": "setMetrics"
},
"typeVersion": 4.6
},
{
"id": "eefe3a5a-722a-4cce-b84b-65c0520a225f",
"name": "データセットを取得",
"type": "n8n-nodes-base.googleSheets",
"position": [
-680,
-620
],
"parameters": {
"options": {},
"sheetName": {
"__rl": true,
"mode": "url",
"value": "https://docs.google.com/spreadsheets/d/1uuPS5cHtSNZ6HNLOi75A2m8nVWZrdBZ_Ivf58osDAS8/edit?gid=630157309#gid=630157309"
},
"documentId": {
"__rl": true,
"mode": "url",
"value": "https://docs.google.com/spreadsheets/d/1uuPS5cHtSNZ6HNLOi75A2m8nVWZrdBZ_Ivf58osDAS8/edit?gid=630157309#gid=630157309"
}
},
"credentials": {
"googleSheetsOAuth2Api": {
"id": "bpr2LoSELMlxpwnN",
"name": "Google Sheets account David"
}
},
"typeVersion": 4.5
},
{
"id": "ea83d68c-f499-4864-918c-5ccaccf555a5",
"name": "重複を削除",
"type": "n8n-nodes-base.removeDuplicates",
"position": [
-460,
-620
],
"parameters": {
"compare": "selectedFields",
"options": {},
"fieldsToCompare": "document_id"
},
"typeVersion": 2
},
{
"id": "6bf59dc8-fe59-49cf-971f-e22dac5d0500",
"name": "Simple Vector Store",
"type": "@n8n/n8n-nodes-langchain.vectorStoreInMemory",
"position": [
-240,
-620
],
"parameters": {
"mode": "insert",
"memoryKey": "rag_evaluation_demo",
"clearStore": true
},
"typeVersion": 1.1
},
{
"id": "66dd59b0-b2e4-43c9-b1f5-7f5a65568f5e",
"name": "Embeddings OpenAI",
"type": "@n8n/n8n-nodes-langchain.embeddingsOpenAi",
"position": [
-300,
-400
],
"parameters": {
"options": {}
},
"credentials": {
"openAiApi": {
"id": "Ag9qPAsY7lpIGkvC",
"name": "JPs n8n openAI key"
}
},
"typeVersion": 1.2
},
{
"id": "213b230d-0f40-4e71-b0bf-4783274e2cdb",
"name": "Default Data Loader",
"type": "@n8n/n8n-nodes-langchain.documentDefaultDataLoader",
"position": [
-100,
-380
],
"parameters": {
"options": {},
"jsonData": "={{ $json.document_text }}",
"jsonMode": "expressionData"
},
"typeVersion": 1
},
{
"id": "64ec7e8c-63f9-4198-9e8d-61345c564c13",
"name": "Recursive Character Text Splitter",
"type": "@n8n/n8n-nodes-langchain.textSplitterRecursiveCharacterTextSplitter",
"position": [
-20,
-220
],
"parameters": {
"options": {}
},
"typeVersion": 1
},
{
"id": "c561f2ce-ef05-4f42-9b6e-6fabd12d6d23",
"name": "付箋2",
"type": "n8n-nodes-base.stickyNote",
"position": [
-980,
-700
],
"parameters": {
"color": 7,
"width": 1180,
"height": 620,
"content": "### Setup: Populate vector DB\nRun this once before running the main workflow. It inserts the documents from the dataset in the vector store, so they can be queried by the agent below"
},
"typeVersion": 1
},
{
"id": "4a7de6bd-17ac-40df-9b27-2d8f6fdbeb07",
"name": "「ワークフロー実行」クリック時",
"type": "n8n-nodes-base.manualTrigger",
"position": [
-900,
-620
],
"parameters": {},
"typeVersion": 1
},
{
"id": "a0f6211f-6202-462c-8fa3-52e10ad89b5d",
"name": "文書関連性指標を計算",
"type": "@n8n/n8n-nodes-langchain.openAi",
"position": [
440,
80
],
"parameters": {
"modelId": {
"__rl": true,
"mode": "list",
"value": "gpt-4o-mini",
"cachedResultName": "GPT-4O-MINI"
},
"options": {},
"messages": {
"values": [
{
"role": "system",
"content": "=You are a teacher evaluating the relevance of student responses to questions.\n\nYou will be given a QUESTION and one or more FACTS provided by a student. Your task is to determine whether these FACTS are relevant to answering the QUESTION.\n\nGrading criteria:\n(1) Evaluate whether the provided FACTS contain information that is relevant to answering the QUESTION\n(2) A FACT is relevant if it contains keywords or concepts semantically related to the QUESTION\n(3) A FACT is relevant if it provides context, examples, or explanations that address any aspect of the QUESTION\n(4) A FACT may contain some unrelated information and still be considered relevant if it also contains information that satisfies criteria (2) or (3)\n\nScoring:\n- Score 1: At least one of the provided FACTS is relevant to the QUESTION\n- Score 0: All provided FACTS are completely unrelated to the QUESTION\n\nEvaluation process:\n1. Identify the key concepts and requirements in the QUESTION\n2. Examine each provided FACT for keywords or concepts related to these key elements\n3. Consider semantic relationships, not just exact keyword matches\n4. Determine if any FACT contributes useful information toward answering the QUESTION\n5. Provide your reasoning in a detailed, step-by-step manner first\n6. Then create a one-sentence summary of your reasoning\n7. Avoid stating your conclusion at the beginning of your extended reasoning\n\nOutput format:\n{\n \"extended_reasoning\": \"<detailed step-by-step analysis of how you evaluated the relevance>\",\n \"reasoning_summary\": \"<one sentence summary of your conclusion>\",\n \"score\": <number: either 0 or 1>\n}\n\nExamples:\n\nQUESTION: \"What factors contributed to the fall of the Roman Empire?\"\nFACTS: \"The Roman Empire fell due to economic problems, military failures, and political corruption. Germanic tribes invaded from the north while the empire was weakened.\"\nEVALUATION: This would receive a score of 1 because the facts directly address causes of the Roman Empire's fall.\n\nQUESTION: \"What factors contributed to the fall of the Roman Empire?\"\nFACTS: \"The Pythagorean theorem states that a² + b² = c² in a right triangle. Jupiter is the largest planet in our solar system.\"\nEVALUATION: This would receive a score of 0 because neither fact relates to the Roman Empire or its fall."
},
{
"content": "=QUESTION: {{ $('When fetching a dataset row').item.json.question }}\n\nFACTS: \n{{ $json.documents }}"
}
]
},
"jsonOutput": true
},
"credentials": {
"openAiApi": {
"id": "Ag9qPAsY7lpIGkvC",
"name": "JPs n8n openAI key"
}
},
"typeVersion": 1.8
},
{
"id": "b01d3427-4907-4715-8c15-669500012f89",
"name": "AI Agent",
"type": "@n8n/n8n-nodes-langchain.agent",
"position": [
-420,
200
],
"parameters": {
"options": {
"systemMessage": "You are a helpful assistant. Answer the user's questions using information from your vector knowledge base only.",
"returnIntermediateSteps": true
}
},
"typeVersion": 1.9
},
{
"id": "97e2db38-0114-449f-bb41-0ab490fc7327",
"name": "OpenAI Chat Model",
"type": "@n8n/n8n-nodes-langchain.lmChatOpenAi",
"position": [
-480,
440
],
"parameters": {
"model": {
"__rl": true,
"mode": "list",
"value": "gpt-4o-mini"
},
"options": {}
},
"credentials": {
"openAiApi": {
"id": "Ag9qPAsY7lpIGkvC",
"name": "JPs n8n openAI key"
}
},
"typeVersion": 1.2
},
{
"id": "5b3679ab-b0a6-4461-9276-89eddb16f699",
"name": "文書を抽出",
"type": "n8n-nodes-base.set",
"position": [
220,
80
],
"parameters": {
"options": {},
"assignments": {
"assignments": [
{
"id": "2d771c52-6637-4af0-9ff7-c87368041bfd",
"name": "documents",
"type": "string",
"value": "={{ $json.intermediateSteps.filter(x => x.action.tool == 'vector_knowledge_base')[0].observation }}"
}
]
}
},
"typeVersion": 3.4
},
{
"id": "0e4c4512-7b9f-4881-8ac7-4c7c24a4a446",
"name": "Simple Vector Store1",
"type": "@n8n/n8n-nodes-langchain.vectorStoreInMemory",
"position": [
-260,
440
],
"parameters": {
"mode": "retrieve-as-tool",
"toolName": "vector_knowledge_base",
"memoryKey": "rag_evaluation_demo",
"toolDescription": "Source of all knowledge to work with"
},
"typeVersion": 1.1
},
{
"id": "edde6c75-dea5-46cc-937e-68e4a925559b",
"name": "Embeddings OpenAI1",
"type": "@n8n/n8n-nodes-langchain.embeddingsOpenAi",
"position": [
-220,
620
],
"parameters": {
"options": {}
},
"credentials": {
"openAiApi": {
"id": "Ag9qPAsY7lpIGkvC",
"name": "JPs n8n openAI key"
}
},
"typeVersion": 1.2
},
{
"id": "e23289e7-a769-4aae-82e6-b7f9e6d903f0",
"name": "付箋6",
"type": "n8n-nodes-base.stickyNote",
"position": [
-980,
-40
],
"parameters": {
"color": 7,
"width": 2000,
"height": 820,
"content": "### Main workflow"
},
"typeVersion": 1
},
{
"id": "c90baefc-22bd-4782-a8eb-236bfb1d2751",
"name": "付箋5",
"type": "n8n-nodes-base.stickyNote",
"position": [
-440,
120
],
"parameters": {
"color": 7,
"width": 300,
"height": 200,
"content": "Make sure to enable 'Return intermediate steps' in the agent, to get the list of executed tools"
},
"typeVersion": 1
}
],
"pinData": {},
"connections": {
"b01d3427-4907-4715-8c15-669500012f89": {
"main": [
[
{
"node": "c381a3c3-12b6-44fe-a2ed-6adf6f997534",
"type": "main",
"index": 0
}
]
]
},
"c381a3c3-12b6-44fe-a2ed-6adf6f997534": {
"main": [
[
{
"node": "5b3679ab-b0a6-4461-9276-89eddb16f699",
"type": "main",
"index": 0
}
],
[
{
"node": "a21f98f0-25eb-44cc-a04a-33181d7e5288",
"type": "main",
"index": 0
}
]
]
},
"eefe3a5a-722a-4cce-b84b-65c0520a225f": {
"main": [
[
{
"node": "ea83d68c-f499-4864-918c-5ccaccf555a5",
"type": "main",
"index": 0
}
]
]
},
"66dd59b0-b2e4-43c9-b1f5-7f5a65568f5e": {
"ai_embedding": [
[
{
"node": "6bf59dc8-fe59-49cf-971f-e22dac5d0500",
"type": "ai_embedding",
"index": 0
}
]
]
},
"5b3679ab-b0a6-4461-9276-89eddb16f699": {
"main": [
[
{
"node": "a0f6211f-6202-462c-8fa3-52e10ad89b5d",
"type": "main",
"index": 0
}
]
]
},
"b521aa40-c8da-4485-b834-50a129c27d3c": {
"main": [
[
{
"node": "b01d3427-4907-4715-8c15-669500012f89",
"type": "main",
"index": 0
}
]
]
},
"97e2db38-0114-449f-bb41-0ab490fc7327": {
"ai_languageModel": [
[
{
"node": "b01d3427-4907-4715-8c15-669500012f89",
"type": "ai_languageModel",
"index": 0
}
]
]
},
"ea83d68c-f499-4864-918c-5ccaccf555a5": {
"main": [
[
{
"node": "6bf59dc8-fe59-49cf-971f-e22dac5d0500",
"type": "main",
"index": 0
}
]
]
},
"edde6c75-dea5-46cc-937e-68e4a925559b": {
"ai_embedding": [
[
{
"node": "0e4c4512-7b9f-4881-8ac7-4c7c24a4a446",
"type": "ai_embedding",
"index": 0
}
]
]
},
"213b230d-0f40-4e71-b0bf-4783274e2cdb": {
"ai_document": [
[
{
"node": "6bf59dc8-fe59-49cf-971f-e22dac5d0500",
"type": "ai_document",
"index": 0
}
]
]
},
"0e4c4512-7b9f-4881-8ac7-4c7c24a4a446": {
"ai_tool": [
[
{
"node": "b01d3427-4907-4715-8c15-669500012f89",
"type": "ai_tool",
"index": 0
}
]
]
},
"e1cf0122-c668-41d2-9316-a173d425d897": {
"main": [
[
{
"node": "b01d3427-4907-4715-8c15-669500012f89",
"type": "main",
"index": 0
}
]
]
},
"677b90c8-aa6f-4cc5-88df-d8e635b0bde2": {
"main": [
[
{
"node": "b521aa40-c8da-4485-b834-50a129c27d3c",
"type": "main",
"index": 0
}
]
]
},
"a0f6211f-6202-462c-8fa3-52e10ad89b5d": {
"main": [
[
{
"node": "fd8628a1-9f31-4f42-88df-b87eda7216ee",
"type": "main",
"index": 0
}
]
]
},
"64ec7e8c-63f9-4198-9e8d-61345c564c13": {
"ai_textSplitter": [
[
{
"node": "213b230d-0f40-4e71-b0bf-4783274e2cdb",
"type": "ai_textSplitter",
"index": 0
}
]
]
},
"4a7de6bd-17ac-40df-9b27-2d8f6fdbeb07": {
"main": [
[
{
"node": "eefe3a5a-722a-4cce-b84b-65c0520a225f",
"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？
上級 - エンジニアリング, 人工知能
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
OpenAIを使ったRAG応答の精度評価：ドキュメント・ベースライン指標
OpenAIによるRAG応答の精度評価：ドキュメントベースライン指標
Set
Evaluation
Http Request
+
Set
Evaluation
Http Request
25 ノードJimleuk
エンジニアリング
評価指標例：正確性（AIによる判断）
評価指標サンプル：正確性（AIによる判断）
Set
Evaluation
Agent
+
Set
Evaluation
Agent
13 ノードDavid Roberts
エンジニアリング
評価指標の例：ツールが呼び出されたかどうかを確認
評価指標の例：ツールが呼び出しされているかチェック
Set
Evaluation
Agent
+
Set
Evaluation
Agent
15 ノードDavid Roberts
エンジニアリング
n8nノードの探索（可視化リファレンスライブラリ内）
n8nノードを可視化リファレンスライブラリで探索
If
Ftp
Set
+
If
Ftp
Set
113 ノードI versus AI
その他
AI スマートアシスタント: Supabase ストレージと Google Drive ファイルとの対話
AIワンチャットボット：SupabaseストレージとGoogle Driveのファイルと対話
If
Set
Wait
+
If
Set
Wait
62 ノードMark Shcherbakov
エンジニアリング
テキスト、音声、画像、PDF をサポートする RAG を備えた AI 駆動型 WhatsApp チャットボット
テキスト、音声、画像、PDF をサポートする AI 搭載の WhatsApp チャットボット (RAG)
Set
Code
Switch
+
Set
Code
Switch
35 ノードNovaNode
エンジニアリング
