Venkata V This workflow corresponds to n8n.io template #15220 — we link there as the canonical source.
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": "IhEqq3b5ajzUHUvv",
"meta": {
"templateCredsSetupCompleted": true
},
"name": "OCR to JSON Automation",
"tags": [],
"nodes": [
{
"id": "911f1b6a-f018-424d-a063-f5026125a015",
"name": "Webhook",
"type": "n8n-nodes-base.webhook",
"position": [
-1504,
96
],
"parameters": {
"path": "ocr-to-json",
"options": {},
"httpMethod": "POST",
"responseMode": "responseNode"
},
"typeVersion": 2
},
{
"id": "921c548f-926e-44c6-a6b1-84f4e5d97694",
"name": "Mistral OCR",
"type": "n8n-nodes-base.mistralAi",
"position": [
-1248,
96
],
"parameters": {
"options": {}
},
"credentials": {
"mistralCloudApi": {
"name": "<your credential>"
}
},
"typeVersion": 1
},
{
"id": "b01b4e52-22c6-4e94-a480-01bdc21539fa",
"name": "Normalize OCR Text",
"type": "n8n-nodes-base.code",
"position": [
-1024,
96
],
"parameters": {
"mode": "runOnceForEachItem",
"jsCode": "const raw = $json;\nconst ocrText =\n raw.text ||\n raw.content ||\n raw.extractedText ||\n raw.output ||\n (Array.isArray(raw.pages) ? raw.pages.map(p => p.markdown || p.text || '').join('\\n\\n') : '') ||\n JSON.stringify(raw, null, 2);\n\nreturn {\n json: {\n ocr_text: ocrText,\n ocr_raw: raw,\n received_at: new Date().toISOString()\n }\n};"
},
"typeVersion": 2
},
{
"id": "f87569b5-03eb-4afb-8e68-8b37d44f3410",
"name": "LLM Extract JSON",
"type": "n8n-nodes-base.httpRequest",
"position": [
-800,
96
],
"parameters": {
"url": "https://api.mistral.ai/v1/chat/completions",
"method": "POST",
"options": {
"timeout": 60000
},
"jsonBody": "={\n \"model\": \"mistral-small\",\n \"temperature\": 0,\n \"response_format\": { \"type\": \"json_object\" },\n \"messages\": [\n {\n \"role\": \"system\",\n \"content\": \"You extract structured invoice/document data from OCR text. Return strict JSON only. Use null for missing scalar values, [] for empty arrays, and numbers for numeric amounts. Convert dates to YYYY-MM-DD when reasonably possible. Include a confidence score from 0 to 1.\"\n },\n {\n \"role\": \"user\",\n \"content\": \"Extract structured data from the following OCR text.\\nReturn this JSON shape exactly:\\n{\\n \\\"document_type\\\": \\\"\\\",\\n \\\"invoice_number\\\": null,\\n \\\"invoice_date\\\": null,\\n \\\"supplier_name\\\": null,\\n \\\"currency\\\": null,\\n \\\"total_amount\\\": null,\\n \\\"line_items\\\": [],\\n \\\"confidence\\\": 0\\n}\\n\\nOCR text:\\n\\\" {{ $json.ocr_text.replace(/\\n/g, '\\\\n').replace(/\\\"/g, '\\\\\\\"') }}\"\n }\n ]\n}",
"sendBody": true,
"sendHeaders": true,
"specifyBody": "json",
"authentication": "predefinedCredentialType",
"headerParameters": {
"parameters": [
{
"name": "Content-Type",
"value": "application/json"
}
]
},
"nodeCredentialType": "mistralCloudApi"
},
"credentials": {
"openAiApi": {
"name": "<your credential>"
},
"mistralCloudApi": {
"name": "<your credential>"
}
},
"typeVersion": 4.2
},
{
"id": "2a9128ef-0483-455b-bc7e-c34ed5c0ac90",
"name": "Clean JSON",
"type": "n8n-nodes-base.code",
"disabled": true,
"position": [
-576,
96
],
"parameters": {
"mode": "runOnceForEachItem",
"jsCode": "const content = $json.choices?.[0]?.message?.content;\nlet data;\n\ntry {\n data = typeof content === 'string' ? JSON.parse(content) : content;\n} catch (e) {\n data = {\n document_type: \"unknown\",\n invoice_number: null,\n invoice_date: null,\n supplier_name: null,\n currency: null,\n total_amount: null,\n line_items: [],\n confidence: 0,\n parse_error: \"Model output was not valid JSON\"\n };\n}\n\nfunction toNumber(value) {\n if (value === null || value === undefined || value === \"\") return null;\n if (typeof value === \"number\") return value;\n const cleaned = String(value).replace(/[^0-9.-]/g, \"\");\n const num = Number(cleaned);\n return Number.isNaN(num) ? null : num;\n}\n\nfunction normalizeDate(value) {\n if (!value) return null;\n if (/^\\d{4}-\\d{2}-\\d{2}$/.test(value)) return value;\n\n const m1 = String(value).match(/^(\\d{1,2})[-\\/ ]([A-Za-z]{3,9})[-\\/ ](\\d{4})$/);\n if (m1) {\n const months = {\n jan:\"01\", january:\"01\", feb:\"02\", february:\"02\", mar:\"03\", march:\"03\",\n apr:\"04\", april:\"04\", may:\"05\", jun:\"06\", june:\"06\", jul:\"07\", july:\"07\",\n aug:\"08\", august:\"08\", sep:\"09\", sept:\"09\", september:\"09\", oct:\"10\", october:\"10\",\n nov:\"11\", november:\"11\", dec:\"12\", december:\"12\"\n };\n const day = m1[1].padStart(2, \"0\");\n const mon = months[m1[2].toLowerCase()];\n const year = m1[3];\n if (mon) return `${year}-${mon}-${day}`;\n }\n\n const m2 = String(value).match(/^(\\d{1,2})[-\\/](\\d{1,2})[-\\/](\\d{4})$/);\n if (m2) {\n return `${m2[3]}-${m2[2].padStart(2, \"0\")}-${m2[1].padStart(2, \"0\")}`;\n }\n\n return value;\n}\n\ndata.total_amount = toNumber(data.total_amount);\ndata.invoice_date = normalizeDate(data.invoice_date);\n\nif (!Array.isArray(data.line_items)) data.line_items = [];\nif (typeof data.confidence !== \"number\") {\n const c = Number(data.confidence);\n data.confidence = Number.isNaN(c) ? 0 : c;\n}\n\nreturn {\n json: data\n};"
},
"typeVersion": 2
},
{
"id": "6313b5c7-c852-4e94-a293-eafebe0cb192",
"name": "Confidence >= 0.5?",
"type": "n8n-nodes-base.if",
"position": [
-352,
96
],
"parameters": {
"options": {},
"conditions": {
"options": {
"version": 2,
"leftValue": "",
"caseSensitive": true,
"typeValidation": "strict"
},
"combinator": "and",
"conditions": [
{
"operator": {
"type": "number",
"operation": "gte"
},
"leftValue": "={{$json.confidence}}",
"rightValue": 0.5
}
]
}
},
"typeVersion": 2.2
},
{
"id": "d0a13ed4-3544-44de-bd43-7e918441ac39",
"name": "Respond OK",
"type": "n8n-nodes-base.respondToWebhook",
"position": [
-128,
0
],
"parameters": {
"options": {},
"respondWith": "json",
"responseBody": "={\n \"status\": \"ok\",\n \"data\": {{JSON.stringify($json)}}\n}"
},
"typeVersion": 1.1
},
{
"id": "e1bc80c8-92c0-4bef-a1fe-9d511b814396",
"name": "Respond Review",
"type": "n8n-nodes-base.respondToWebhook",
"position": [
-128,
192
],
"parameters": {
"options": {},
"respondWith": "json",
"responseBody": "={\n \"status\": \"review_needed\",\n \"data\": $json\n}"
},
"typeVersion": 1.1
},
{
"id": "6946957c-2bd7-45d8-8e3d-40132c9bbcc1",
"name": "Sticky Note",
"type": "n8n-nodes-base.stickyNote",
"position": [
-1568,
-80
],
"parameters": {
"color": 7,
"width": 224,
"height": 432,
"content": "This webhook node Accepts a PDF or image upload via Webhook as binary property \"data\""
},
"typeVersion": 1
},
{
"id": "b710116c-b9ce-4e3a-a64d-e20d7045fe5b",
"name": "Sticky Note1",
"type": "n8n-nodes-base.stickyNote",
"position": [
-1296,
-32
],
"parameters": {
"color": 7,
"width": 432,
"height": 336,
"content": "Runs OCR with the Mistral OCR node\n\nCreate/attach Mistral AI credentials on the \"Mistral OCR\" node\n\nNormalizes OCR text"
},
"typeVersion": 1
},
{
"id": "34158ea5-8ad8-4dd9-a5cf-9b52224234b7",
"name": "Sticky Note3",
"type": "n8n-nodes-base.stickyNote",
"position": [
-848,
-64
],
"parameters": {
"color": 7,
"width": 400,
"height": 400,
"content": "Sends OCR text to an LLM to extract structured JSON\n\nCreate/attach LLM AI credentials of your choice\n\nCleans and normalizes the JSON"
},
"typeVersion": 1
},
{
"id": "1ecfad88-526a-46cf-b39f-57dadabb27f0",
"name": "Sticky Note5",
"type": "n8n-nodes-base.stickyNote",
"position": [
-400,
-144
],
"parameters": {
"color": 7,
"width": 480,
"height": 544,
"content": "Returns either:\n - status: ok\n - status: review_needed\n\n- The workflow returns review_needed when confidence is below 0.5.\n"
},
"typeVersion": 1
},
{
"id": "a4e06f24-9938-463c-bc2c-df9ac9ef2779",
"name": "Sticky Note2",
"type": "n8n-nodes-base.stickyNote",
"position": [
-2016,
-288
],
"parameters": {
"width": 400,
"height": 752,
"content": "N8N AI LLM Unstructured Invoice data PDF OCR recognition to JSON output API\n\nWhat this workflow does\n1. Accepts a PDF or image upload via Webhook as binary property \"data\"\n2. Runs OCR with the Mistral OCR node\n3. Normalizes OCR text\n4. Sends OCR text to an LLM to extract structured JSON\n5. Cleans and normalizes the JSON\n6. Returns either:\n - status: ok\n - status: review_needed\n\nSetup\n1. Import the workflow JSON into n8n\n2. Create/attach Mistral AI credentials on the \"Mistral OCR\" node\n3. Create/attach your choice LLM AI credentials on the OCR text to JSON converson node\n4. Activate the workflow\n5. POST a file to:\n /webhook/ocr-to-json\n\nNotes\n- This starter is tuned for invoices/documents but can be adapted for receipts, purchase orders, or forms.\n- Depending on your installed n8n version, the Mistral node parameter names may need minor adjustment after import.\n- The workflow returns review_needed when confidence is below 0.5.\n"
},
"typeVersion": 1
}
],
"active": true,
"settings": {
"binaryMode": "separate",
"executionOrder": "v1"
},
"versionId": "8e7cf8a0-3ab0-4a9b-aad8-f0d2178af437",
"connections": {
"Webhook": {
"main": [
[
{
"node": "Mistral OCR",
"type": "main",
"index": 0
}
]
]
},
"Clean JSON": {
"main": [
[
{
"node": "Confidence >= 0.5?",
"type": "main",
"index": 0
}
]
]
},
"Mistral OCR": {
"main": [
[
{
"node": "Normalize OCR Text",
"type": "main",
"index": 0
}
]
]
},
"LLM Extract JSON": {
"main": [
[
{
"node": "Clean JSON",
"type": "main",
"index": 0
}
]
]
},
"Confidence >= 0.5?": {
"main": [
[
{
"node": "Respond OK",
"type": "main",
"index": 0
}
],
[
{
"node": "Respond Review",
"type": "main",
"index": 0
}
]
]
},
"Normalize OCR Text": {
"main": [
[
{
"node": "LLM Extract JSON",
"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.
mistralCloudApiopenAiApi
Pro
For the full experience including quality scoring and batch install features for each workflow upgrade to Pro
About this workflow
Accepts a PDF or image upload via Webhook as binary property "data" Runs OCR with the Mistral OCR node Normalizes OCR text Sends OCR text to an LLM to extract structured JSON Cleans and normalizes the JSON Returns either: status: ok status: review_needed Import the workflow JSON…
Source: https://n8n.io/workflows/15220/ — 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.
Who is this for? Event organizers, conference planners, and marketing teams fighting registration drop-off who want 4-field forms with LinkedIn-level attendee intelligence. What problem is this workfl
Data Table, HubSpot, Email Send +4
Use Case: Analyze images with multiple subjects. In this use case I have a bookshelf and am extracting and verifying book titles/authors from a bookshelf photo.
OpenAI, HTTP Request
<section> <h2>🌊 What it Does</h2> <p> This workflow <strong>automatically classifies uploaded files</strong> (PDFs or images) as <span>floorplans</span> or <span>non‑floorplans</span>. It filters out
HTTP Request
This n8n workflow provides a ready-to-use API endpoint for extracting structured data from images. It processes an image URL using an AI-powered OCR model and returns the extracted details in a struct
HTTP Request
AI Keyword & Entity Extractor (Ollama). Uses httpRequest. Webhook trigger; 7 nodes.
HTTP Request