Filip Mijic This workflow corresponds to n8n.io template #16370 — we link there as the canonical source.
This workflow follows the HTTP Request → Postgres recipe pattern — see all workflows that pair these two integrations.
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 →
{
"id": "VHF3xt11ExE3JiQW",
"meta": {
"builderVariant": "mcp",
"aiBuilderAssisted": true
},
"name": "Reindex Markdown Docs into a Supabase pgvector RAG Store (Schedule + Webhook)",
"tags": [],
"nodes": [
{
"id": "8a1f68ea-817f-425d-afdc-8c0ab1c33ddb",
"name": "Schedule: Daily Reindex",
"type": "n8n-nodes-base.scheduleTrigger",
"position": [
-48,
160
],
"parameters": {
"rule": {
"interval": [
{
"triggerAtHour": 2
}
]
}
},
"typeVersion": 1.3
},
{
"id": "b9cca7ec-ab1e-4afc-9d68-94fd1f3e3e77",
"name": "Fetch Sources (FAQ + Blog)",
"type": "n8n-nodes-base.httpRequest",
"position": [
176,
272
],
"parameters": {
"url": "https://kairesume.fit/api/rag/sources",
"options": {
"timeout": 30000,
"response": {
"response": {
"responseFormat": "json"
}
}
},
"authentication": "genericCredentialType",
"genericAuthType": "httpHeaderAuth"
},
"credentials": {
"httpHeaderAuth": {
"name": "<your credential>"
}
},
"executeOnce": true,
"typeVersion": 4.4
},
{
"id": "149533b2-93a7-4298-a95f-de3f1ba112c1",
"name": "Chunk and Batch",
"type": "n8n-nodes-base.code",
"position": [
400,
272
],
"parameters": {
"jsCode": "const TARGET_CHARS = 2800;\nconst OVERLAP_CHARS = 400;\nconst MIN_LETTERS_PER_CHUNK = 40;\nconst BATCH_SIZE = 32;\n\nfunction stripFrontmatter(md) {\n if (!md.startsWith(\"---\")) return md;\n const end = md.indexOf(\"\\n---\", 3);\n if (end === -1) return md;\n return md.slice(end + 4).replace(/^\\s+/, \"\");\n}\n\nfunction paragraphSplit(text, target) {\n if (text.length <= target) return text.length;\n const window = text.slice(0, target);\n const paragraphCut = window.lastIndexOf(\"\\n\\n\");\n if (paragraphCut > target * 0.6) return paragraphCut + 2;\n const sentenceCut = window.lastIndexOf(\". \");\n if (sentenceCut > target * 0.6) return sentenceCut + 2;\n return target;\n}\n\nfunction subdivideLongSection(text) {\n if (text.length <= TARGET_CHARS) return [text];\n const out = [];\n let i = 0;\n while (i < text.length) {\n const end = i + paragraphSplit(text.slice(i), TARGET_CHARS);\n out.push(text.slice(i, end).trim());\n if (end >= text.length) break;\n i = Math.max(end - OVERLAP_CHARS, i + 1);\n }\n return out;\n}\n\nfunction letterCount(s) {\n return (s.match(/[A-Za-z]/g) || []).length;\n}\n\nfunction chunkMarkdown(doc) {\n const body = stripFrontmatter(doc.content);\n const sections = body.split(/\\n(?=## )/).map(function (s) { return s.trim(); }).filter(Boolean);\n const pieces = [];\n for (const section of sections) {\n for (const piece of subdivideLongSection(section)) pieces.push(piece);\n }\n const out = [];\n let chunkIdx = 0;\n for (const piece of pieces) {\n if (letterCount(piece) < MIN_LETTERS_PER_CHUNK) continue;\n out.push({ source: doc.source, chunk_idx: chunkIdx, content: piece });\n chunkIdx++;\n }\n return out;\n}\n\nconst inputItem = items[0].json;\nconst sources = inputItem.sources || [];\nconst allChunks = [];\nfor (const doc of sources) {\n for (const chunk of chunkMarkdown(doc)) allChunks.push(chunk);\n}\n\nconst runStart = new Date().toISOString();\nconst totalBatches = Math.max(1, Math.ceil(allChunks.length / BATCH_SIZE));\nconst batches = [];\nfor (let i = 0; i < allChunks.length; i += BATCH_SIZE) {\n const batch = allChunks.slice(i, i + BATCH_SIZE);\n batches.push({\n json: {\n batch: batch,\n inputs: batch.map(function (c) { return c.content; }),\n run_start: runStart,\n batch_idx: batches.length,\n total_batches: totalBatches,\n total_chunks: allChunks.length,\n },\n });\n}\n\nreturn batches;\n"
},
"typeVersion": 2
},
{
"id": "ff79fd6e-7c61-4867-86ab-95b659420fbc",
"name": "Delete Stale Chunks",
"type": "n8n-nodes-base.postgres",
"position": [
848,
80
],
"parameters": {
"query": "DELETE FROM public.rag_chunks WHERE updated_at < $1::timestamptz RETURNING source, chunk_idx;",
"options": {
"queryReplacement": "={{ $(\"Chunk and Batch\").first().json.run_start }}"
},
"operation": "executeQuery"
},
"credentials": {
"postgres": {
"name": "<your credential>"
}
},
"typeVersion": 2.6
},
{
"id": "61871abe-9ff3-4024-a3e4-b5b939407b70",
"name": "Embed Batch (Supabase)",
"type": "n8n-nodes-base.httpRequest",
"position": [
848,
272
],
"parameters": {
"url": "https://jhccfsyytbkfmdvduvds.supabase.co/functions/v1/embed",
"method": "POST",
"options": {
"timeout": 60000,
"response": {
"response": {
"responseFormat": "json"
}
}
},
"jsonBody": "={{ JSON.stringify({ inputs: $json.inputs }) }}",
"sendBody": true,
"specifyBody": "json",
"authentication": "genericCredentialType",
"genericAuthType": "httpHeaderAuth"
},
"credentials": {
"httpHeaderAuth": {
"name": "<your credential>"
}
},
"typeVersion": 4.4
},
{
"id": "2d212395-3bb1-43eb-89d9-767c258339e8",
"name": "Build Upsert SQL",
"type": "n8n-nodes-base.code",
"position": [
1072,
272
],
"parameters": {
"jsCode": "const embedResponse = items[0].json;\nconst vectors = embedResponse.vectors || [];\n\nconst loopItem = $(\"Loop Batches\").item.json;\nconst batch = loopItem.batch;\nconst runStart = loopItem.run_start;\nconst batchIdx = loopItem.batch_idx;\n\nif (!Array.isArray(batch) || !Array.isArray(vectors) || batch.length !== vectors.length) {\n throw new Error(\"Vector count \" + vectors.length + \" does not match chunk count \" + (batch ? batch.length : 0) + \" for batch \" + batchIdx);\n}\n\nfunction sqlString(s) {\n return \"'\" + String(s).replace(/'/g, \"''\") + \"'\";\n}\n\nfunction vectorLiteral(v) {\n return sqlString(\"[\" + v.map(function (x) { return Number(x).toFixed(6); }).join(\",\") + \"]\");\n}\n\nconst valuesRows = batch.map(function (c, i) {\n return \"(\" + sqlString(c.source) + \", \" + Number(c.chunk_idx) + \", \" + sqlString(c.content) + \", \" + vectorLiteral(vectors[i]) + \"::vector, NOW())\";\n}).join(\",\\n \");\n\nconst sql = \"INSERT INTO public.rag_chunks (source, chunk_idx, content, embedding, updated_at)\\nVALUES\\n \" + valuesRows + \"\\nON CONFLICT (source, chunk_idx) DO UPDATE\\nSET content = EXCLUDED.content,\\n embedding = EXCLUDED.embedding,\\n updated_at = NOW();\";\n\nreturn [{ json: { sql: sql, batch_size: batch.length, run_start: runStart, batch_idx: batchIdx } }];\n"
},
"typeVersion": 2
},
{
"id": "4ba0f657-7da2-4925-be4f-1cde35ffd794",
"name": "Upsert Chunks",
"type": "n8n-nodes-base.postgres",
"position": [
1296,
352
],
"parameters": {
"query": "{{ $json.sql }}",
"options": {
"queryBatching": "single"
},
"operation": "executeQuery"
},
"credentials": {
"postgres": {
"name": "<your credential>"
}
},
"typeVersion": 2.6
},
{
"id": "ffd40eb7-91ba-4f6b-94a5-1b738cd26570",
"name": "Loop Batches",
"type": "n8n-nodes-base.splitInBatches",
"position": [
624,
272
],
"parameters": {
"options": {}
},
"typeVersion": 3
},
{
"id": "1837434c-8804-4074-8472-14b47a9d9a79",
"name": "Deploy Webhook",
"type": "n8n-nodes-base.webhook",
"position": [
-48,
352
],
"parameters": {
"path": "kairesume-deploy",
"options": {},
"httpMethod": "POST"
},
"typeVersion": 2.1
},
{
"id": "b05c9c6b-c3a6-4cf7-801b-9890c80cf0d4",
"name": "Sticky Note 1",
"type": "n8n-nodes-base.stickyNote",
"position": [
-48,
-480
],
"parameters": {
"width": 520,
"height": 520,
"content": "## \ud83d\udd01 RAG Corpus Reindex \u2192 Supabase pgvector\n\n**Who's it for:** Anyone running a retrieval-augmented (RAG) chatbot or search over their own docs who needs the vector store kept in sync automatically.\n\n**What it does:** On a daily schedule (and on a deploy webhook), it fetches your source markdown (FAQ + blog), chunks it on H2 boundaries (~2800 chars, 400 overlap), embeds each batch via a Supabase Edge Function (e.g. gte-small, 384-dim), bulk-UPSERTs the vectors into a Postgres `rag_chunks` table, then deletes stale rows from removed/shrunk sources.\n\n**How it works:**\n1. **Schedule** (daily) or **Deploy Webhook** triggers a run\n2. **Fetch Sources** pulls markdown docs from your API\n3. **Chunk and Batch** splits into ~32-chunk batches\n4. **Loop Batches** \u2192 **Embed Batch** \u2192 **Build Upsert SQL** \u2192 **Upsert Chunks** (ON CONFLICT updates)\n5. After the loop, **Delete Stale Chunks** removes rows older than this run\n\n**Idempotent** \u2014 re-running just refreshes `updated_at`. Runs on free tiers."
},
"typeVersion": 1
},
{
"id": "7a95922a-8886-4ecc-ab79-2e21727f5f04",
"name": "Sticky Note 2",
"type": "n8n-nodes-base.stickyNote",
"position": [
512,
-480
],
"parameters": {
"color": 4,
"width": 460,
"height": 520,
"content": "### \u2699\ufe0f Setup\n**Credentials (select placeholders after import):**\n- **[RAG_API]_HeaderAuth** \u2014 Header Auth for your sources API + Supabase embed function\n- **[SUPABASE]_Postgres** \u2014 Postgres connection to your Supabase database\n\n**Before running:**\n1. In **Fetch Sources**, set the URL to your docs API (returns `{ sources: [{ source, content }] }`)\n2. In **Embed Batch**, set your Supabase Edge Function URL (returns `{ vectors: [...] }`)\n3. Create the table:\n `rag_chunks(source text, chunk_idx int, content text, embedding vector(384), updated_at timestamptz, primary key(source, chunk_idx))`\n4. Adjust the schedule hour / embedding dimension as needed\n\n**Error handling:** the network nodes use **retry-on-fail (3\u00d7)**. Optionally set an *Error Workflow* in workflow settings to get alerts."
},
"typeVersion": 1
}
],
"active": true,
"settings": {
"timezone": "America/New_York",
"binaryMode": "separate",
"errorWorkflow": "Rsh29fD5MByBp7SN",
"availableInMCP": true,
"executionOrder": "v1"
},
"versionId": "34741ee4-690e-47ca-9404-532116c338a0",
"nodeGroups": [],
"connections": {
"Loop Batches": {
"main": [
[
{
"node": "Delete Stale Chunks",
"type": "main",
"index": 0
}
],
[
{
"node": "Embed Batch (Supabase)",
"type": "main",
"index": 0
}
]
]
},
"Upsert Chunks": {
"main": [
[
{
"node": "Loop Batches",
"type": "main",
"index": 0
}
]
]
},
"Deploy Webhook": {
"main": [
[
{
"node": "Fetch Sources (FAQ + Blog)",
"type": "main",
"index": 0
}
]
]
},
"Chunk and Batch": {
"main": [
[
{
"node": "Loop Batches",
"type": "main",
"index": 0
}
]
]
},
"Build Upsert SQL": {
"main": [
[
{
"node": "Upsert Chunks",
"type": "main",
"index": 0
}
]
]
},
"Embed Batch (Supabase)": {
"main": [
[
{
"node": "Build Upsert SQL",
"type": "main",
"index": 0
}
]
]
},
"Schedule: Daily Reindex": {
"main": [
[
{
"node": "Fetch Sources (FAQ + Blog)",
"type": "main",
"index": 0
}
]
]
},
"Fetch Sources (FAQ + Blog)": {
"main": [
[
{
"node": "Chunk and Batch",
"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.
httpHeaderAuthpostgres
Pro
For the full experience including quality scoring and batch install features for each workflow upgrade to Pro
About this workflow
This workflow reindexes Markdown documentation into a Supabase Postgres pgvector table by fetching source docs from an HTTP API, chunking and embedding them via a Supabase Edge Function, upserting the vectors, and deleting stale chunks on a daily schedule or on-demand webhook.…
Source: https://n8n.io/workflows/16370/ — original creator credit. Request a take-down →
Related workflows
Workflows that share integrations, category, or trigger type with this one. All free to copy and import.
My Workflow. Uses outputParserStructured, httpRequest, lmChatGoogleGemini, chainLlm. Scheduled trigger; 82 nodes.
Output Parser Structured, HTTP Request, Google Gemini Chat +15
Search Worflow Docker Complete. Uses documentDefaultDataLoader, textSplitterCharacterTextSplitter, vectorStoreSupabase, embeddingsOllama. Scheduled trigger; 71 nodes.
Document Default Data Loader, Text Splitter Character Text Splitter, Supabase Vector Store +14
This workflow automates end-to-end customer journey management by intelligently routing queries through multiple AI models (OpenAI, Claude) based on complexity and context. Designed for customer succe
HTTP Request, Agent, OpenAI Chat +7
This workflow implements a self-healing Retrieval-Augmented Generation (RAG) maintenance system that automatically updates document embeddings, evaluates retrieval quality, detects embedding drift, an
HTTP Request, Postgres, OpenAI Embeddings +5
This workflow automates end-to-end e-commerce order processing from intake through fulfillment by orchestrating multiple AI-powered validation stages and external system integrations. Designed for e-c
HTTP Request, Agent, OpenAI Chat +10