AI Invoice Processing System

An AI-powered invoice processing workflow built with n8n, Ollama, and PostgreSQL.

This project automatically extracts structured information from PDF invoices using OCR and a local Large Language Model (LLM), stores the extracted data in PostgreSQL, and returns structured JSON through an HTTP webhook.

100% Local AI — No cloud AI APIs required.

---

Preview

---

Overview

This project demonstrates how to build an end-to-end AI document processing workflow using only local components.

The workflow receives a PDF invoice through an HTTP webhook, extracts its text, sends the extracted content to a locally hosted Ollama model (Qwen2.5:3B), converts the invoice into structured JSON, stores the data in PostgreSQL, and finally returns the parsed result through the webhook response.

Because every component runs locally, the workflow can be deployed without sending documents to external AI services.

---

Features

• PDF invoice upload through HTTP Webhook
• PDF text extraction
• AI-powered invoice understanding using Ollama
• Structured JSON generation
• PostgreSQL data storage
• HTTP JSON response
• Local AI deployment
• Docker-based PostgreSQL

---

Tech Stack

Technology	Purpose
n8n	Workflow Automation
Ollama	Local LLM Runtime
Qwen2.5:3B	Invoice Information Extraction
PostgreSQL	Database
Docker	Containerization
Extract from File	PDF Text Extraction

---

Workflow

The invoice processing pipeline consists of the following steps:

1. Receive a PDF invoice via Webhook.
2. Extract text from the uploaded PDF.
3. Send the extracted text to Ollama.
4. Convert invoice text into structured JSON.
5. Store invoice information in PostgreSQL.
6. Return JSON to the client.

---

Workflow Screenshot

---

Architecture

                PDF Invoice
                     │
                     ▼
               n8n Webhook
                     │
                     ▼
          Extract Text from PDF
                     │
                     ▼
           Ollama (Qwen2.5:3B)
                     │
                     ▼
          Structured Invoice JSON
              │               │
              ▼               ▼
       PostgreSQL       Webhook Response

---

Example JSON Output

Example:

{
  "vendor": "Nova Tech",
  "invoice_number": "#INV-2026-0842",
  "invoice_date": "2026-08-03",
  "currency": "USD",
  "total": 271.11,
  "items": [
    {
      "description": "Wireless Noise-Canceling Headphones",
      "quantity": 1,
      "unit_price": 149.99,
      "amount": 149.99
    }
  ]
}

---

PostgreSQL Storage

Each processed invoice is automatically stored in PostgreSQL.

Current table schema:

Column	Type
id	SERIAL
vendor	TEXT
invoice_number	TEXT
invoice_date	DATE
currency	TEXT
total	NUMERIC
items	JSONB
created_at	TIMESTAMP

---

Project Structure

ai-invoice-processing/
├── docker-compose.yml
├── README.md
├── .gitignore
├── workflow/
│   └── workflow.json
├── sql/
│   └── schema.sql
├── invoices/
│   └── invoice.pdf
├── examples/
│   ├── invoice.json
│   └── response.json
├── screenshots/
│   ├── workflow.png
│   ├── json-output.png
│   └── postgres.png
└── docs/

---

Installation

Clone the repository

git clone https://github.com/ibam28/ai-invoice-processing.git
cd ai-invoice-processing

Start PostgreSQL

docker compose up -d

Import the database schema

psql -U n8n -d invoices -f sql/schema.sql

Import the workflow

Open n8n and import:

workflow/workflow.json

---

API

Endpoint

POST /webhook/invoice-upload

Content-Type

multipart/form-data

Form Field

data

---

Sample Files

Input PDF

invoices/invoice.pdf

Expected JSON

examples/invoice.json

Webhook Response

examples/response.json

---

Roadmap

• PDF Upload
• PDF Text Extraction
• Ollama Integration
• Structured JSON Extraction
• PostgreSQL Storage
• Google Sheets Export
• Discord Notification
• Slack Notification
• Multi-page Invoice Support

---

License

This project is licensed under the MIT License.