Files

T

Ran into an issue with the previous model for having mllama archetecture so I changed it

2026-06-26 08:36:45 -05:00

Getting Started With AI

StepForge keeps AI local. It talks to your own Ollama server on your machine and does not send guide content to the cloud.

1. Install Ollama

Install Ollama from https://ollama.com and make sure the service is running.

On most systems you can verify it with:

ollama --version

The recommended default is:

ollama pull llama3.2:1b

That model is small enough to feel responsive on modest hardware, but still good enough for human-sounding titles and short text blocks.

If you want StepForge to send the screenshot itself to the model, pull a vision-capable model instead:

ollama pull gemma3

That model can inspect pictures as well as text, so it is better when you want the AI to read the UI directly from the screenshot.

If you need something even smaller, try:

ollama pull qwen3:0.6b

or:

ollama pull gemma3:270m

Those are lighter, but they are usually weaker at writing polished step text.

In StepForge, open Settings and find the AI section.

Set:

Enable AI text filling to on
Ollama host to your local Ollama server
Ollama model to llama3.2:1b for text-only mode, or llama3.2-vision if you want screenshot-aware AI

The default host is:

http://127.0.0.1:11434

Use the Test connection button in the AI settings section.

If the model is installed, StepForge should confirm the host and model.

AI is never automatic. After capture, use the AI button next to:

You can also use More -> Generate all text fields with AI to fill the whole step in one pass.

Capture titles are still generated automatically without AI.
AI generation only works when Enable AI text filling is turned on.
The app always uses local OCR around the click area first, then local AI only when you ask for it.
When the selected Ollama model supports vision, StepForge also sends the screenshot to the model so it can cross-check OCR and visual context.