Handle Customer Support Queries with Cache-first RAG Using Redis, Langcache…

The workflow JSON

Copy or download the full n8n JSON below. Paste it into a new n8n workflow, add your credentials, activate. Full import guide →

{
  "meta": {
    "templateCredsSetupCompleted": true
  },
  "nodes": [
    {
      "id": "af6e85ef-2255-454f-8281-517b17a112c0",
      "name": "When chat message received",
      "type": "@n8n/n8n-nodes-langchain.chatTrigger",
      "position": [
        -80,
        736
      ],
      "parameters": {
        "options": {}
      },
      "typeVersion": 1.4
    },
    {
      "id": "f5884164-83b2-48d0-806d-381c1ba9b802",
      "name": "decompose_query",
      "type": "@n8n/n8n-nodes-langchain.agent",
      "position": [
        496,
        736
      ],
      "parameters": {
        "text": "={{ $('When chat message received').item.json.chatInput }}",
        "options": {
          "systemMessage": "=Analyze this customer support query and determine if it needs to be broken down into sub-questions.\n        \n        Original query: {{ $('When chat message received').item.json.chatInput }}\n        \n        Rules:\n        - If the query is simple and focused on ONE topic, respond with: SINGLE_QUESTION\n        - If the query has multiple distinct aspects that would benefit from separate research, break it into 2-4 specific sub-questions\n        - Each sub-question should be self-contained and cacheable\n        \n        If breaking down, provide ONLY the sub-questions, one per line, no numbering.\n        If keeping as single question, respond with exactly: SINGLE_QUESTION\n---\noutput should be json object with the follwoing structure:\n{\n\"questions\": [\"Q1\", \"Q2\", \"Q3\",\"Q4\"]\n}"
        },
        "promptType": "define",
        "hasOutputParser": true
      },
      "typeVersion": 3.1
    },
    {
      "id": "2c48cfdb-bba2-4d9e-88c5-72176c0d9595",
      "name": "Structured Output Parser",
      "type": "@n8n/n8n-nodes-langchain.outputParserStructured",
      "position": [
        704,
        1040
      ],
      "parameters": {
        "schemaType": "manual",
        "inputSchema": "{\n  \"$schema\": \"https://json-schema.org/draft/2020-12/schema\",\n  \"type\": \"object\",\n  \"properties\": {\n    \"questions\": {\n      \"type\": \"array\",\n      \"items\": {\n        \"type\": \"string\"\n      },\n      \"description\": \"List of generated questions.\"\n    }\n  },\n  \"required\": [\"questions\"]\n}\n"
      },
      "typeVersion": 1.3
    },
    {
      "id": "0c52729a-53e9-4f18-a4b4-845cc2384bec",
      "name": "Split Out",
      "type": "n8n-nodes-base.splitOut",
      "position": [
        800,
        736
      ],
      "parameters": {
        "options": {
          "disableDotNotation": false,
          "destinationFieldName": "question"
        },
        "fieldToSplitOut": "output.questions"
      },
      "typeVersion": 1
    },
    {
      "id": "6410933e-dde4-4de5-9f72-5444f0f51891",
      "name": "Search LangCache",
      "type": "n8n-nodes-base.httpRequest",
      "onError": "continueErrorOutput",
      "position": [
        1248,
        528
      ],
      "parameters": {
        "url": "={{ $('LangCache Config').item.json.langcacheBaseUrl }}/v1/caches/{{ $('LangCache Config').item.json.langcacheCacheId }}/entries/search",
        "method": "POST",
        "options": {},
        "sendBody": true,
        "sendHeaders": true,
        "authentication": "genericCredentialType",
        "bodyParameters": {
          "parameters": [
            {
              "name": "prompt",
              "value": "={{ $('Loop Over Items').item.json.question }}"
            },
            {
              "name": "similarityThreshold",
              "value": "={{ $('LangCache Config').item.json.similarityThreshold }}"
            }
          ]
        },
        "genericAuthType": "httpBearerAuth",
        "headerParameters": {
          "parameters": [
            {
              "name": "accept",
              "value": "application/json"
            }
          ]
        }
      },
      "credentials": {
        "httpBearerAuth": {
          "name": "<your credential>"
        }
      },
      "typeVersion": 4.3
    },
    {
      "id": "65f6963c-6faf-46a4-971d-bf162a2cf49d",
      "name": "Is Cache Hit?",
      "type": "n8n-nodes-base.if",
      "position": [
        1488,
        448
      ],
      "parameters": {
        "options": {},
        "conditions": {
          "options": {
            "version": 3,
            "leftValue": "",
            "caseSensitive": true,
            "typeValidation": "strict"
          },
          "combinator": "and",
          "conditions": [
            {
              "id": "a84a9a1c-f13b-465c-abc2-f27b9e24f605",
              "operator": {
                "type": "boolean",
                "operation": "true",
                "singleValue": true
              },
              "leftValue": "={{ $json.data?.[0]?.similarity >= $('LangCache Config').item.json.similarityThreshold }}",
              "rightValue": ""
            }
          ]
        }
      },
      "typeVersion": 2.3
    },
    {
      "id": "0f16f03d-6fe1-4da8-9cc5-1fbba5012f30",
      "name": "LangCache Config",
      "type": "n8n-nodes-base.set",
      "position": [
        224,
        736
      ],
      "parameters": {
        "options": {},
        "assignments": {
          "assignments": [
            {
              "id": "79a40928-d50d-4998-90d8-cea319e9a1b7",
              "name": "langcacheBaseUrl",
              "type": "string",
              "value": "https://aws-us-east-1.langcache.redis.io"
            },
            {
              "id": "b9f02205-4a11-4814-9971-ee840b32537f",
              "name": "langcacheCacheId",
              "type": "string",
              "value": "b83aa61d58be484ebc37c64f1f30c2fa"
            },
            {
              "id": "cdc69c29-a34f-40e1-a9dd-193aacae9c69",
              "name": "similarityThreshold",
              "type": "number",
              "value": 0.75
            },
            {
              "id": "2bebfa5b-9f31-4a60-8121-422ce9a5b846",
              "name": "max_iterations",
              "type": "string",
              "value": "2"
            }
          ]
        }
      },
      "typeVersion": 3.4
    },
    {
      "id": "b142939c-814c-4b36-87bd-45a955f3c564",
      "name": "Loop Over Items",
      "type": "n8n-nodes-base.splitInBatches",
      "position": [
        1024,
        880
      ],
      "parameters": {
        "options": {}
      },
      "typeVersion": 3
    },
    {
      "id": "ea1352da-745e-4a86-8365-5ddfe8ed93ab",
      "name": "synthesize_response_node",
      "type": "@n8n/n8n-nodes-langchain.agent",
      "position": [
        1408,
        864
      ],
      "parameters": {
        "text": "=- Original query: \n        {{ $('When chat message received').item.json.chatInput }}\n- Information gathered:\n{{ $json.data.toJsonString() }}",
        "options": {
          "systemMessage": "= You are a helpful customer support assistant. Combine the following question-answer pairs \ninto a single, coherent, and comprehensive response to the user's original query.\n        \n1- If information gathered is enough to answer the question, Provide a natural, conversational response that:\n  - Directly addresses the user's question\n  - Integrates all relevant information smoothly\n  - Is helpful and actionable\n  - Maintains a professional, friendly tone.\n\n2- If information gathered is not enough or accurate to answer the question reply with: \n\"I apologize, but I couldn't find answers to your question. Please try rephrasing or contact support directly.\""
        },
        "promptType": "define",
        "hasOutputParser": true
      },
      "typeVersion": 3.1
    },
    {
      "id": "7548562c-2a2a-4c4b-8aae-5ed08d5dfb25",
      "name": "Aggregate",
      "type": "n8n-nodes-base.aggregate",
      "position": [
        1248,
        864
      ],
      "parameters": {
        "options": {},
        "aggregate": "aggregateAllItemData"
      },
      "typeVersion": 1
    },
    {
      "id": "aecc43ec-bb69-4d13-a7a4-00bb91453172",
      "name": "Redis Vector Store",
      "type": "@n8n/n8n-nodes-langchain.vectorStoreRedis",
      "position": [
        1440,
        -112
      ],
      "parameters": {
        "mode": "insert",
        "options": {},
        "redisIndex": {
          "__rl": true,
          "mode": "list",
          "value": "kb-3accd7ed",
          "cachedResultName": "kb-3accd7ed"
        }
      },
      "credentials": {
        "redis": {
          "name": "<your credential>"
        }
      },
      "typeVersion": 1.3
    },
    {
      "id": "2344ea11-5412-4d6d-9d59-005d5748a475",
      "name": "Redis Vector Store2",
      "type": "@n8n/n8n-nodes-langchain.vectorStoreRedis",
      "position": [
        2128,
        672
      ],
      "parameters": {
        "mode": "retrieve-as-tool",
        "options": {},
        "redisIndex": {
          "__rl": true,
          "mode": "list",
          "value": "kb-3accd7ed",
          "cachedResultName": "kb-3accd7ed"
        },
        "toolDescription": "Using search_knowledge_base tool for query"
      },
      "credentials": {
        "redis": {
          "name": "<your credential>"
        }
      },
      "typeVersion": 1.3
    },
    {
      "id": "84e92cea-0cf7-45c8-b346-d87126052efc",
      "name": "Embeddings OpenAI",
      "type": "@n8n/n8n-nodes-langchain.embeddingsOpenAi",
      "position": [
        2112,
        816
      ],
      "parameters": {
        "options": {}
      },
      "credentials": {
        "openAiApi": {
          "name": "<your credential>"
        }
      },
      "typeVersion": 1.2
    },
    {
      "id": "7a76f285-d661-4fd7-83cc-f20f050dbcc9",
      "name": "Embeddings OpenAI1",
      "type": "@n8n/n8n-nodes-langchain.embeddingsOpenAi",
      "position": [
        1120,
        144
      ],
      "parameters": {
        "options": {}
      },
      "credentials": {
        "openAiApi": {
          "name": "<your credential>"
        }
      },
      "typeVersion": 1.2
    },
    {
      "id": "bfe94569-e5c7-4282-b2f9-c710a076907d",
      "name": "Default Data Loader",
      "type": "@n8n/n8n-nodes-langchain.documentDefaultDataLoader",
      "position": [
        1408,
        144
      ],
      "parameters": {
        "options": {}
      },
      "typeVersion": 1.1
    },
    {
      "id": "27cd82d3-d938-49bc-92f2-213d80000e43",
      "name": "example Data",
      "type": "n8n-nodes-base.set",
      "position": [
        1216,
        -112
      ],
      "parameters": {
        "options": {},
        "assignments": {
          "assignments": [
            {
              "id": "8da432e2-ebc9-454b-b9b1-0968df56a715",
              "name": "raw_docs",
              "type": "array",
              "value": "=[     \"Our premium support plan includes 24/7 phone support, priority email response within 2 hours, and dedicated account management. Premium support costs $49/month.\",     \"Account upgrade process: Go to Account Settings \u2192 Plan & Billing \u2192 Select Upgrade. Available plans: Basic $9/month, Pro $29/month, Enterprise $99/month.\",     \"API rate limits by plan: Free tier 100 requests/hour, Basic 1,000 requests/hour, Pro 10,000 requests/hour, Enterprise unlimited with fair-use policy.\",     \"Data export options: CSV, JSON, XML formats supported. Large exports (>1GB) may take up to 24 hours to process.\",     \"Third-party integrations: Native support for Slack, Microsoft Teams, Zoom, Salesforce, HubSpot. 200+ additional integrations available via Zapier.\",     \"Security features: SOC2 compliance, end-to-end encryption, GDPR compliance, SSO integration, audit logs, IP whitelisting.\",     \"Billing and payments: We accept all major credit cards, PayPal, and ACH transfers. Enterprise customers can pay by invoice with NET30 terms.\",     \"Account recovery: Use forgot password link, verify email, or contact support with account verification details. Response within 4 hours.\" ]"
            }
          ]
        }
      },
      "typeVersion": 3.4
    },
    {
      "id": "b05e94c5-98ff-4be8-aa56-1739a99019f5",
      "name": "OpenAI Chat Model",
      "type": "@n8n/n8n-nodes-langchain.lmChatOpenAi",
      "position": [
        1072,
        1072
      ],
      "parameters": {
        "model": {
          "__rl": true,
          "mode": "list",
          "value": "gpt-4.1-mini"
        },
        "options": {},
        "builtInTools": {}
      },
      "credentials": {
        "openAiApi": {
          "name": "<your credential>"
        }
      },
      "typeVersion": 1.3
    },
    {
      "id": "b02d24fa-1de0-4240-b7bb-02f02c2619f8",
      "name": "Save to LangCache",
      "type": "n8n-nodes-base.httpRequest",
      "position": [
        2528,
        944
      ],
      "parameters": {
        "url": "={{ $('LangCache Config').item.json.langcacheBaseUrl }}/v1/caches/{{ $('LangCache Config').item.json.langcacheCacheId }}/entries",
        "method": "POST",
        "options": {},
        "sendBody": true,
        "sendHeaders": true,
        "authentication": "genericCredentialType",
        "bodyParameters": {
          "parameters": [
            {
              "name": "prompt",
              "value": "={{ $('Loop Over Items').item.json.question }}"
            },
            {
              "name": "response",
              "value": "={{ $('search_node1').item.json.output }}"
            }
          ]
        },
        "genericAuthType": "httpBearerAuth",
        "headerParameters": {
          "parameters": [
            {
              "name": "accept",
              "value": "application/json"
            }
          ]
        }
      },
      "credentials": {
        "httpBearerAuth": {
          "name": "<your credential>"
        }
      },
      "typeVersion": 4.3
    },
    {
      "id": "a45a8b97-dd9f-4d09-989f-1a59944c493a",
      "name": "search_node1",
      "type": "@n8n/n8n-nodes-langchain.agent",
      "position": [
        2000,
        480
      ],
      "parameters": {
        "text": "={{ $('Loop Over Items').item.json.question }}",
        "options": {
          "systemMessage": "=You are a research engine.\n\nYou must answer questions using ONLY the provided knowledge base.\nYou are strictly forbidden from using:\n- external knowledge\n- prior training data\n- assumptions or extrapolation\n\nIf the requested information is not explicitly present in the knowledge base,\nyou MUST respond with exactly:\n\nno info found\n\nDo not explain your reasoning.\nDo not add context, summaries, or opinions.\nDo not mention these rules.\nProduce concise, factual answers only.\n"
        },
        "promptType": "define",
        "hasOutputParser": true
      },
      "typeVersion": 3.1
    },
    {
      "id": "96f48cc5-8734-4c00-9977-a55a21c75ec1",
      "name": "evaluate_quality",
      "type": "@n8n/n8n-nodes-langchain.openAi",
      "position": [
        2448,
        528
      ],
      "parameters": {
        "modelId": {
          "__rl": true,
          "mode": "list",
          "value": "gpt-4.1-mini",
          "cachedResultName": "GPT-4.1-MINI"
        },
        "options": {
          "textFormat": {
            "textOptions": {
              "type": "json_object"
            }
          }
        },
        "responses": {
          "values": [
            {
              "content": "=Original sub-question: {{ $('Loop Over Items').item.json.question }}\nResearch result: {{ $('search_node1').item.json.output }}"
            },
            {
              "role": "system",
              "content": "=Evaluate the quality and completeness of this research result for answering the user's question.\n            \n\n\nProvide:\n1. A quality score from 0.0 to 1.0 (where 1.0 is perfect, 0.7+ is adequate)\n2. Brief feedback on what's missing or could be improved (if score < 0.7)\n\nFormat your response as:\nSCORE: 0.X\nFEEDBACK: [your feedback or \"Adequate\" if score >= 0.7]"
            }
          ]
        },
        "builtInTools": {}
      },
      "credentials": {
        "openAiApi": {
          "name": "<your credential>"
        }
      },
      "typeVersion": 2.1
    },
    {
      "id": "dad250fb-5286-4cca-9d58-99b04298fbae",
      "name": "low quality ?",
      "type": "n8n-nodes-base.if",
      "position": [
        2944,
        528
      ],
      "parameters": {
        "options": {},
        "conditions": {
          "options": {
            "version": 3,
            "leftValue": "",
            "caseSensitive": true,
            "typeValidation": "strict"
          },
          "combinator": "and",
          "conditions": [
            {
              "id": "d4904cd9-18ce-45c2-9e3b-e5bf7e339e48",
              "operator": {
                "type": "boolean",
                "operation": "true",
                "singleValue": true
              },
              "leftValue": "={{ $json.SCORE < 0.7 && $('current_iteration').item.json.current_iterration >= $('LangCache Config').item.json.max_iterations\t}}",
              "rightValue": 0
            }
          ]
        }
      },
      "typeVersion": 2.3
    },
    {
      "id": "1cfea78d-da68-4c4e-af21-a3085f842a39",
      "name": "increase iteration",
      "type": "n8n-nodes-base.set",
      "position": [
        3360,
        480
      ],
      "parameters": {
        "options": {},
        "assignments": {
          "assignments": [
            {
              "id": "e5f9dc79-f8fe-4d03-a059-a1948821451c",
              "name": "current_iteration",
              "type": "number",
              "value": "={{ $('current_iteration').item.json.current_iterration + 1 }}"
            }
          ]
        }
      },
      "typeVersion": 3.4
    },
    {
      "id": "a650cdaf-4a8e-4938-af44-182d94b34193",
      "name": "Schedule Trigger",
      "type": "n8n-nodes-base.scheduleTrigger",
      "position": [
        1024,
        -112
      ],
      "parameters": {
        "rule": {
          "interval": [
            {}
          ]
        }
      },
      "typeVersion": 1.3
    },
    {
      "id": "e250b675-af0c-4c59-8749-d027da7b72f8",
      "name": "Simple Memory",
      "type": "@n8n/n8n-nodes-langchain.memoryBufferWindow",
      "position": [
        512,
        1056
      ],
      "parameters": {
        "sessionKey": "={{ $('When chat message received').item.json.sessionId }}",
        "sessionIdType": "customKey",
        "contextWindowLength": 10
      },
      "typeVersion": 1.3
    },
    {
      "id": "726bfd01-9dac-42ff-816c-151cbe58f897",
      "name": "Simple Memory1",
      "type": "@n8n/n8n-nodes-langchain.memoryBufferWindow",
      "position": [
        1952,
        784
      ],
      "parameters": {
        "sessionKey": "={{ $('When chat message received').item.json.sessionId }}",
        "sessionIdType": "customKey",
        "contextWindowLength": 10
      },
      "typeVersion": 1.3
    },
    {
      "id": "0677630f-5809-408c-8233-85d734f2a732",
      "name": "Simple Memory2",
      "type": "@n8n/n8n-nodes-langchain.memoryBufferWindow",
      "position": [
        1328,
        1056
      ],
      "parameters": {
        "sessionKey": "={{ $('When chat message received').item.json.sessionId }}",
        "sessionIdType": "customKey",
        "contextWindowLength": 10
      },
      "typeVersion": 1.3
    },
    {
      "id": "5f9e7ddc-9454-451d-8f19-28a145aaa9b2",
      "name": "current_iteration",
      "type": "n8n-nodes-base.set",
      "position": [
        1840,
        464
      ],
      "parameters": {
        "options": {},
        "assignments": {
          "assignments": [
            {
              "id": "0d7da9a5-ff39-4e6a-9105-6ac8d3a06cde",
              "name": "current_iteration",
              "type": "number",
              "value": "={{ $json.current_iteration ?? 1 }}"
            }
          ]
        }
      },
      "typeVersion": 3.4
    },
    {
      "id": "fdf91552-525c-4b3d-8e0f-de910525e648",
      "name": "Sticky Note",
      "type": "n8n-nodes-base.stickyNote",
      "position": [
        48,
        -800
      ],
      "parameters": {
        "width": 880,
        "height": 1120,
        "content": "# Customer support RAG workflow:\n## Workflow Overview\nCache-first **RAG** workflow for customer support.\n\n**Flow:**  \nChat \u2192 Decompose \u2192 Cache \u2192 Redis Search \u2192 Quality Check \u2192 Cache \u2192 Respond\n\n**Goals:** Fast, accurate, no hallucinations, cost-controlled.\n\n\nThis workflow provides an end-to-end Retrieval-Augmented Generation (RAG) solution for customer support use cases using n8n. It combines a cache-first strategy with semantic search to deliver fast, accurate answers from a structured knowledge base.\n\nIncoming chat messages are analyzed and, when needed, decomposed into smaller sub-questions. Each sub-question is first checked against LangCache to reuse previously generated answers and reduce repeated LLM calls. If no suitable cached result is found, the workflow retrieves relevant documents from a Redis vector store powered by OpenAI embeddings.\n\nTo improve answer reliability, retrieved results are evaluated for quality. If the quality score does not meet the configured threshold, the workflow can re-run retrieval with adjusted context for a limited number of iterations. Once sufficient information is gathered, all results are consolidated into a single, concise response suitable for customer-facing communication.\n\nThis template is ideal for support chatbots, internal help desks, and knowledge-base-driven assistants where performance, cost control, and answer consistency are critical.\n\n## How it works\n\nWhen a chat message is received, the workflow determines whether it contains a single question or multiple topics. If necessary, the message is split into a small number of focused sub-questions.\n\nEach sub-question is first checked against LangCache using a similarity threshold. High-confidence cache hits are reused immediately to minimize latency and API cost. If no cache hit is found, the workflow queries a Redis vector store containing embedded knowledge-base documents. Retrieved results are then evaluated for quality and relevance.\n\nIf the quality score is below the defined threshold, the workflow retries retrieval for a limited number of iterations. Once acceptable results are available, all responses are merged and synthesized into one clear final answer that is returned to the user.\n\n## Setup steps\n\n1- Create and configure credentials in n8n for OpenAI, Redis, and LangCache.\n\n2- Open the LangCache Config node and set the cache ID, base URL, similarity threshold, and maximum iterations.\n\n3- Ensure your Redis vector index is populated with knowledge-base documents (or use the provided example loader).\n\n4- Review the OpenAI model and embedding nodes and adjust them to match your performance and cost requirements.\n\n5- Test the workflow using the chat trigger and verify cache hits, vector retrieval, and final response output."
      },
      "typeVersion": 1
    },
    {
      "id": "7994e6ea-45d6-44b6-8a1b-fc9b347f32eb",
      "name": "getScore",
      "type": "n8n-nodes-base.set",
      "position": [
        2768,
        528
      ],
      "parameters": {
        "options": {},
        "assignments": {
          "assignments": [
            {
              "id": "9da210db-b091-44b8-b936-d6263d78b512",
              "name": "SCORE",
              "type": "number",
              "value": "={{ $json.output[0].content[0].text.SCORE }}"
            },
            {
              "id": "d6f96594-6a30-4899-8bdf-388ba695c779",
              "name": "FEEDBACK",
              "type": "string",
              "value": "={{ $json.output[0].content[0].text.FEEDBACK }}"
            }
          ]
        }
      },
      "typeVersion": 3.4
    },
    {
      "id": "e6978f27-8803-4851-b148-fdbf9911648c",
      "name": "Sticky Note1",
      "type": "n8n-nodes-base.stickyNote",
      "position": [
        112,
        512
      ],
      "parameters": {
        "width": 320,
        "height": 656,
        "content": "#### Configuration (Edit First)\nUpdate in **LangCache Config**:\n- `langcacheBaseUrl`\n- `langcacheCacheId`\n- `similarityThreshold` (default `0.75`)\n- `max_iterations` (default `2`)"
      },
      "typeVersion": 1
    },
    {
      "id": "d44f3c0c-fb23-4bd1-97b4-96d32f446e8c",
      "name": "Sticky Note2",
      "type": "n8n-nodes-base.stickyNote",
      "position": [
        464,
        512
      ],
      "parameters": {
        "width": 464,
        "height": 656,
        "content": "## Query Decomposition\nSplits complex user input into focused questions to improve retrieval and caching.\n"
      },
      "typeVersion": 1
    },
    {
      "id": "981cbf1e-7179-4c77-8a79-b46d50dfdf90",
      "name": "Sticky Note3",
      "type": "n8n-nodes-base.stickyNote",
      "position": [
        992,
        384
      ],
      "parameters": {
        "width": 704,
        "height": 352,
        "content": "#### Cache-First Strategy\nEach question is checked in **LangCache** first.\n- Hit \u2192 reuse answer  \n- Miss \u2192 search Redis\n\nReduces latency and API cost."
      },
      "typeVersion": 1
    },
    {
      "id": "8c35a870-1f2b-4207-9827-6c8aeae2550b",
      "name": "Sticky Note4",
      "type": "n8n-nodes-base.stickyNote",
      "position": [
        1728,
        384
      ],
      "parameters": {
        "width": 688,
        "height": 784,
        "content": "#### Redis Vector Retrieval\nRuns only on cache miss.\nUses embeddings to retrieve relevant knowledge from Redis."
      },
      "typeVersion": 1
    },
    {
      "id": "bb7f7b45-eee6-41ac-89d7-7b39062d04af",
      "name": "Sticky Note5",
      "type": "n8n-nodes-base.stickyNote",
      "position": [
        2432,
        384
      ],
      "parameters": {
        "width": 704,
        "height": 384,
        "content": "## Quality Evaluation\nEach answer is scored (`0.0 \u2013 1.0`).\n- \u2265 `0.7` \u2192 accept  \n- < `0.7` \u2192 retry if allowed"
      },
      "typeVersion": 1
    },
    {
      "id": "8b0d0a12-df24-4a78-a8f2-5d987c8da340",
      "name": "Sticky Note6",
      "type": "n8n-nodes-base.stickyNote",
      "position": [
        3168,
        384
      ],
      "parameters": {
        "width": 384,
        "height": 784,
        "content": "## Retry Control\nRetries are limited by `max_iterations` to avoid loops and high cost."
      },
      "typeVersion": 1
    },
    {
      "id": "f1a6f21c-fba8-42eb-b479-b6e10b2c8339",
      "name": "Sticky Note7",
      "type": "n8n-nodes-base.stickyNote",
      "position": [
        2432,
        800
      ],
      "parameters": {
        "width": 704,
        "height": 368,
        "content": "## ## Save to Cache\nOnly high-quality answers are saved to **LangCache** for future reuse."
      },
      "typeVersion": 1
    },
    {
      "id": "da62019e-6e0e-41cd-9918-fe295ddc9659",
      "name": "Sticky Note8",
      "type": "n8n-nodes-base.stickyNote",
      "position": [
        992,
        768
      ],
      "parameters": {
        "width": 704,
        "height": 400,
        "content": "## Generate the respoonse"
      },
      "typeVersion": 1
    },
    {
      "id": "70286448-0f9d-44d6-ba62-666efd454d5b",
      "name": "Sticky Note9",
      "type": "n8n-nodes-base.stickyNote",
      "position": [
        1024,
        -192
      ],
      "parameters": {
        "width": 752,
        "height": 480,
        "content": "## Prepare the Knowledge Base -  Example Data"
      },
      "typeVersion": 1
    }
  ],
  "connections": {
    "getScore": {
      "main": [
        [
          {
            "node": "low quality ?",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "Aggregate": {
      "main": [
        [
          {
            "node": "synthesize_response_node",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "Split Out": {
      "main": [
        [
          {
            "node": "Loop Over Items",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "example Data": {
      "main": [
        [
          {
            "node": "Redis Vector Store",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "search_node1": {
      "main": [
        [
          {
            "node": "evaluate_quality",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "Is Cache Hit?": {
      "main": [
        [
          {
            "node": "Loop Over Items",
            "type": "main",
            "index": 0
          }
        ],
        [
          {
            "node": "current_iteration",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "Simple Memory": {
      "ai_memory": [
        [
          {
            "node": "decompose_query",
            "type": "ai_memory",
            "index": 0
          }
        ]
      ]
    },
    "low quality ?": {
      "main": [
        [
          {
            "node": "increase iteration",
            "type": "main",
            "index": 0
          }
        ],
        [
          {
            "node": "Save to LangCache",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "Simple Memory1": {
      "ai_memory": [
        [
          {
            "node": "search_node1",
            "type": "ai_memory",
            "index": 0
          }
        ]
      ]
    },
    "Simple Memory2": {
      "ai_memory": [
        [
          {
            "node": "synthesize_response_node",
            "type": "ai_memory",
            "index": 0
          }
        ]
      ]
    },
    "Loop Over Items": {
      "main": [
        [
          {
            "node": "Aggregate",
            "type": "main",
            "index": 0
          }
        ],
        [
          {
            "node": "Search LangCache",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "decompose_query": {
      "main": [
        [
          {
            "node": "Split Out",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "LangCache Config": {
      "main": [
        [
          {
            "node": "decompose_query",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "Schedule Trigger": {
      "main": [
        [
          {
            "node": "example Data",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "Search LangCache": {
      "main": [
        [
          {
            "node": "Is Cache Hit?",
            "type": "main",
            "index": 0
          }
        ],
        [
          {
            "node": "current_iteration",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "evaluate_quality": {
      "main": [
        [
          {
            "node": "getScore",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "Embeddings OpenAI": {
      "ai_embedding": [
        [
          {
            "node": "Redis Vector Store2",
            "type": "ai_embedding",
            "index": 0
          }
        ]
      ]
    },
    "OpenAI Chat Model": {
      "ai_languageModel": [
        [
          {
            "node": "decompose_query",
            "type": "ai_languageModel",
            "index": 0
          },
          {
            "node": "synthesize_response_node",
            "type": "ai_languageModel",
            "index": 0
          },
          {
            "node": "search_node1",
            "type": "ai_languageModel",
            "index": 0
          }
        ]
      ]
    },
    "Save to LangCache": {
      "main": [
        [
          {
            "node": "Loop Over Items",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "current_iteration": {
      "main": [
        [
          {
            "node": "search_node1",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "Embeddings OpenAI1": {
      "ai_embedding": [
        [
          {
            "node": "Redis Vector Store",
            "type": "ai_embedding",
            "index": 0
          }
        ]
      ]
    },
    "increase iteration": {
      "main": [
        [
          {
            "node": "current_iteration",
            "type": "main",
            "index": 0
          }
        ]
      ]
    },
    "Default Data Loader": {
      "ai_document": [
        [
          {
            "node": "Redis Vector Store",
            "type": "ai_document",
            "index": 0
          }
        ]
      ]
    },
    "Redis Vector Store2": {
      "ai_tool": [
        [
          {
            "node": "search_node1",
            "type": "ai_tool",
            "index": 0
          }
        ]
      ]
    },
    "Structured Output Parser": {
      "ai_outputParser": [
        [
          {
            "node": "decompose_query",
            "type": "ai_outputParser",
            "index": 0
          }
        ]
      ]
    },
    "When chat message received": {
      "main": [
        [
          {
            "node": "LangCache Config",
            "type": "main",
            "index": 0
          }
        ]
      ]
    }
  }
}

Credentials you'll need

Each integration node will prompt for credentials when you import. We strip credential IDs before publishing — you'll add your own.

• httpBearerAuth
• openAiApi
• redis

About this workflow

An end-to-end Retrieval-Augmented Generation (RAG) customer support workflow for n8n, using a cache-first strategy (LangCache) combined with a Redis vector store powered by OpenAI embeddings. This template is designed for fast, accurate, and cost-efficient customer support…

Source: https://n8n.io/workflows/12400/ — original creator credit. Request a take-down →

More AI & RAG workflows → · Browse all categories →

Related workflows

Workflows that share integrations, category, or trigger type with this one. All free to copy and import.