🧠 Automating Firefly III Categorization with Node-RED and Lightweight AI

Managing personal or business finances in Firefly III can get tedious when every new transaction requires a manual category and budget. What if your finance app could intelligently tag these for you - instantly and without setting up an automation for each one?

That’s exactly what this guide will show you: how to use Node-RED and a lightweight AI model (like Phi-3 Mini or Mistral 7B) to automatically assign categories and budgets to new Firefly III transactions.

⚙️ The Goal

We’ll build a real-time automation that:

• Listens to Firefly III webhook events
• Checks if the transaction is a withdrawal
• Uses a small, OpenAI-compatible LLM to suggest a category and budget
• Updates Firefly III with those values
• Tags the transaction as ai-assisted

No heavy cloud models or GPU power needed — this can run on a Raspberry Pi or home server.

🧩 Tech Stack

• Firefly III (self-hosted finance manager)
• Node-RED (flow-based automation)
• Lightweight LLM (Phi-3 Mini, Mistral, or LLaMA 3.1 8B)
• OpenAI-compatible endpoint (local Ollama or LM Studio)

🔌 The Flow

Here’s the overview:

1. Firefly III Webhook → triggers on new or updated transactions.
2. Node-RED HTTP In node receives the webhook payload.
3. Logic function checks for missing category/budget.
4. Fetch categories and budgets from Firefly III API.
5. Send transaction details to the AI model using OpenAI /v1/chat/completions format.
6. Update Firefly III with the AI’s suggested category, budget, and an ai-assisted tag.

🧠 Why Use a Lightweight Model?

Most classification tasks — like deciding whether “Starbucks” belongs under Coffee or Dining Out — don’t require large LLMs.
A compact model like Phi-3 Mini or Mistral 7B Instruct is perfect: small, fast, and locally deployable.

Example compatible engines:

🚀 Node-RED Flow

You can import this JSON flow directly into Node-RED.
It listens for new Firefly transactions and enriches them automatically.

👉 View the complete flow on GitHub

Each time a transaction webhook arrives:

• Node-RED gets category & budget lists via the Firefly API.
• Builds a classification prompt.
• Calls your AI endpoint using an OpenAI-compatible API.
• Updates the transaction with the model’s suggestions.

All updated records are tagged ai-assisted for transparency and potentially future review.

🧰 Setup Instructions

1. Import the Flow in Node-RED → Menu → Import → Clipboard.
2. Replace placeholders:
   • YOUR_FIREFLY_URL
   • YOUR_FIREFLY_TOKEN
   • YOUR_LLM_API (e.g. http://localhost:11434/v1)
   • YOUR_LLM_TOKEN if needed
3. In Firefly III, create a webhook:
   • URL: http://your-node-red-host:1880/firefly/webhook
   • Events: transaction_journal_created, transaction_journal_updated
4. On your ollama node, run the model via the command ollama run phi3:instruct

Altenatively to the web hooks in firefly iii, if you have already followed this tutorial, you can simply add a "link out" from the existing webhooks, and a "link in" to the start of the new flow.

That’s it — Firefly III now has an AI-powered assistant classifying your spending.

🪶 Example Prompt

Transaction: Starbucks Coffee
Amount: 5.20
Notes: morning latte
Available Categories: Dining, Groceries, Coffee, Utilities
Available Budgets: Food, Entertainment, Essentials
Return ONLY:
Category: Coffee | Budget: Food

🧩 Optional Upgrade: Model Selection

You can easily add a decision node in Node-RED to select the best model based on system resources:

• Use phi-3-mini when memory is limited
• Switch to mistral-7b or llama-3.1-8b when GPU power is available

This makes your automation flexible across environments.

🔒 Privacy & Control

All AI processing stays within your environment — no cloud dependencies required.
You decide where your data goes and how it’s processed.

✅ Final Thoughts

This small automation bridges self-hosted finance with local AI intelligence.
It saves time, maintains privacy, and adapts to your system’s capabilities.

If you’re already using Node-RED and Firefly III, adding this AI layer takes your setup from organized to intelligent.

Have questions or improvements? Drop a comment below or reach out to me on any of my socials.

About the author

Tim Wilkes is a UK-based security architect with over 15 years of experience in electronics, Linux, and Unix systems administration. Since 2021, he's been designing secure systems for a telecom company while indulging his passions for programming, automation, and 3D printing. Tim shares his projects, tinkering adventures, and tech insights here - partly as a personal log, and partly in the hopes that others will find them useful.

Want to connect or follow along?

LinkedIn: [phpsytems]
Twitter / X: [@timmehwimmy]
Mastodon: [@timmehwimmy@infosec.exchange]


If you've found a post helpful, consider supporting the blog - it's a part-time passion that your support helps keep alive.

⚠️ Disclaimer

This post may contain affiliate links. If you choose to purchase through them, I may earn a small commission at no extra cost to you. I only recommend items and services I’ve personally read or used and found valuable.

As an Amazon Associate I earn from qualifying purchases.