AmigaGPT for AROS x86_64
chat, image generation, text-to-speech

Ported to AROS from:
https://github.com/sacredbanana/AmigaGPT


First, you need to run an assign command in the system (possibly in user-startup)

assign AMIGAGPT: <selected directory>

You'll also need an API key from your provider. I tested it with OpenAI, but you can also use Gemini, Grok, etc.

Download link:
https://archives.arosworld.org/?function=showfile&file=network/misc/amigagpt.x86_64-aros-v11.zip

Tested on 2026.03-D03 - works good. Tested with GPT 5.6 Luna.

I created the key for Gemini, but it isn't being found or recognized. Do I need to name it something specific or place it in a specific location?

You must enter the API key in the settings

Chat / Chat Provider Settings

retrofaza, thanks, my mistake. Instead of pasting the code, I had created a Key file
Now I'm getting the error you see in the screenshot. What am I missing?

You need to choose a different model. For Gemini, click “Fetch models” and select, for example, “gemini-3.6-flash.”

Just one more thing: you'll probably get a “429 Too many requests” error. This happens when you simply generate a free key without providing billing information. It seems these APIs aren't meant to work completely for free.

The creator of AmigaGPT comments on this as follows:
https://github.com/sacredbanana/AmigaGPT/issues/40
https://github.com/sacredbanana/AmigaGPT/issues/1

I actually generated a key on OpenAI, and it only started working with the program after I added $5 to my account.

Actually, it seems to be just a paid service; I don't think many regular users will like it, given what the web offers for free.

Hi! I'm the creator of AmigaGPT. Nice to see this being ported to x86_64 AROS! Now that AROS has just been released for the Raspberry Pi, I was thinking of making official ports for both x86 and ARM.

As for the too many requests error, it can also happen for free Gemini accounts if you have "web search" turned on in your chat profile. Try turning it off. Google doesn't allow web search functionality for non-paying users

Hello nightfox!

Here are a few resources that can come in handy:

To get a cross-compiler for 64-bit, you can follow this tutorial: https://arosnews.github.io/how-to-cross-compile-aros-hosted-wsl/

Here is a thread with resources for developers: https://arosworld.org/infusions/forum/viewthread.php?thread_id=1201&pid=10891#post_4162

To get a cross-compiler for RPi, I please check this post: https://arosworld.org/infusions/forum/viewthread.php?thread_id=2019&rowstart=80#post_13667

Note that RPi target is still work in progress. Breaking changes can still happen which will mean you will have to re-build your software with refreshed SDK.

Is there a Docker image for it? If not, I can make one whenever I get around to trying out AROS dev. For AmigaOS and MorphOS scripts call my custom docker images. That way, even GitHub is be able to build the final AmigaGPT.lha file whenever there's a new commit to master. I just released AmigaGPT 3.2.0 a few hours ago, but it uses a new library I had to spend the weekend porting over https://github.co...anana/file

I think for the 64-bit x86_64 there is one here: https://github.com/BlitterStudio/aros-compiler-docker

For RPI, I don't think there is one.