ocr-invoice

name	ocr-invoice
description	Extract supplier invoice data from Odoo attachment (PDF/image) using an OpenAI-compatible vision model and optionally create the vendor bill automatically in Odoo.
homepage	https://github.com/nicolasramos/odooclaw
metadata	{"openclaw":{"emoji":"🧾","requires":{"env":["ODOO_URL","ODOO_DB","ODOO_USERNAME","ODOO_PASSWORD","VISION_API_BASE","VISION_MODEL","OPENAI_API_KEY"]}}}

OCR Invoice Skill

Complete vendor bill flow from an Odoo attachment:

1. Download ir.attachment (PDF or image)
2. Extract invoice data with a vision model (OpenAI-compatible API)
3. Normalize invoice JSON
4. Optionally create vendor bill (account.move, in_invoice) and attach original file
5. Optionally create employee expense (hr.expense) and attach original file
6. Optionally create mileage expense (hr.expense) and attach original file

When to use this skill

• When a user uploads an invoice (PDF/image) in Odoo chat.
• When attachment_id is available.

Tools

ocr-invoice

Extract invoice data without creating the vendor bill in Odoo.

{
  "attachment_id": 1234,
  "sender_id": 7,
  "company_id": 3,
  "allowed_company_ids": [3, 5]
}

ocr-create-vendor-bill

Extract and create vendor bill in Odoo.

{
  "attachment_id": 1234,
  "dry_run": false,
  "sender_id": 7,
  "company_id": 3,
  "allowed_company_ids": [3, 5]
}

• dry_run=true: extract and normalize only; do not create account.move.
• sender_id + company context: executes ORM as the Odoo user and respects multi-company (company_id, allowed_company_ids).

ocr-create-employee-expense

Extract and create employee expense in Odoo.

{
  "attachment_id": 1234,
  "dry_run": false,
  "sender_id": 7,
  "company_id": 3,
  "allowed_company_ids": [3, 5],
  "employee_id": 12,
  "product_id": 345
}

• dry_run=true: extract and normalize only; do not create hr.expense.
• employee_id and product_id are optional overrides.
• If employee_id is omitted, the skill resolves employee from sender_id.

ocr-create-mileage-expense

Extract and create mileage expense in Odoo.

{
  "attachment_id": 1234,
  "dry_run": false,
  "sender_id": 7,
  "company_id": 3,
  "allowed_company_ids": [3, 5],
  "employee_id": 12,
  "product_id": 345
}

• dry_run=true: extract and normalize mileage data only.
• Extracted fields include trip_date, origin, destination, distance_km, rate_per_km, and total_amount.
• If distance_km and rate_per_km are detected, the expense uses them as quantity and unit_amount.

Environment variables

Required:

• ODOO_URL
• ODOO_DB
• ODOO_USERNAME
• ODOO_PASSWORD
• VISION_API_BASE
• VISION_MODEL
• OPENAI_API_KEY

Optional:

• OCR_API_BASE (if an external /v1/ocr/invoice endpoint exists)
• OCR_TIMEOUT_SECONDS (default: 240)
• OCR_MAX_PAGES (default: 4)
• OCR_IMAGE_DPI (default: 170)

Example configuration

environment:
  - VISION_API_BASE=https://api.openai.com/v1
  - VISION_MODEL=gpt-4o-mini
  - OPENAI_API_KEY=sk-...

Also works with OpenAI-compatible endpoints (OVH, Groq, vLLM, etc.) that expose multimodal chat/completions.

Expected output

{
  "partner_name": "Proveedor SL",
  "partner_vat": "B12345678",
  "invoice_date": "2026-03-09",
  "invoice_date_due": "2026-04-09",
  "ref": "FRA-2026-001",
  "currency": "EUR",
  "amount_total": 121.0,
  "invoice_line_ids": [
    {
      "name": "Servicio",
      "quantity": 1,
      "price_unit": 100.0,
      "tax_percentage": 21
    }
  ]
}

Implementation notes

• For PDF files, it converts pages to images with pdftoppm.
• If OCR_API_BASE is set, it tries that endpoint first.
• During vendor bill creation:
  • looks up partner by VAT, then by name;
  • creates partner if missing;
  • assigns a default expense account when possible;
  • maps taxes by percentage (account.tax, purchase type);
  • attaches the original PDF/image to the created bill.