AmigaGPT is a text and image generation program that runs on the classic AmigaOS. Utilising the power of the OpenAI's GPT-3 and GPT-4 architectures, this program brings state-of-the-art language modeling to your Amiga computer.
AmigaGPT uses the GPT-4 architecture developed by OpenAI to generate coherent, context-aware responses to your input.
AmigaGPT can access the powerful DALL-E models to generate images from a prompt. You can view and save the images right inside the app.
AmigaGPT takes full advantage of the latest AmigaOS 3.2 API to provide a smooth, native user experience without the need to have third party frameworks installed.
You can customise the look and feel of the application, including the ability to choose the fonts and a choice of opening in the Workbench screen or a custom screen.
AmigaGPT can use the Amiga's speech synthesis capability to read the generated text aloud with support for switching between the old Workbench 1.x v34 and the Workbench 2.0 v37 speech synthesisers.
Ensure you have the necessary system requirements:
- An OCS/ECS/AGA Amiga
- AmigaOS 3.9 and AmigaOS 3.X included in Amiga Forever supported but AmigaOS 3.2 or higher required for best experience. AmigaOS 4.1 version also included.
- Motorola 68020 or higher CPU or PPC for AmigaOS 4
- Internet access using a TCP/IP stack such as Roadshow (http://roadshow.apc-tcp.de/index-en.php)
- AmiSSL 5.9 or higher (http://aminet.net/util/libs/AmiSSL-5.9-OS3.lha)
- An OpenAI account with an active API key
- Optional: AmigaOS 3 only: A copy of the Workbench 1.x disk to install
narrator.devicev34 and a copy of the Workbench 2.0 disk to installnarrator.devicev37 - Optional: AmigaOS 4 only: Flite device (http://aminet.net/package/mus/misc/flite_device)
- Optional: For OpenAI voices, AHI needs to be installed (http://aminet.net/package/driver/audio/ahiusr_4.18)
- Install AmiSSL and a TCP/IP stack if not already done so
- Download the latest release of AmigaGPT
- Extract the
amigagpt.lhaarchive to your desired location
AmigaGPT supports reading the output aloud. This requires a file called narrator.device which cannot be included with AmigaGPT because it is still under copyright. Therefore, you must copy this file legally from your Workbench disks so that AmigaGPT will be able to synthesise speech. There are 2 versions of narrator.device supported, v34 and v37.
v34 is the original version that came with Workbench 1.x. v37 was an updated version included with Workbench 2.0.x. It has more features and sounds more natural, however it does sound quite different which is why AmigaGPT supports you installing both versions and your choice of version to be used can be selected in the Speech menu in the app.
Regardless of which version of narrator.device you choose to install (or both), AmigaGPT requires that you install the free third party translator.library v43. This works with both versions of narrator.device.
Since translator.library v43 is not available as a standalone install, you will need to install v42 and then patch it to v43.
- Download http://aminet.net/util/libs/translator42.lha and extract the archive to any convenient location on your Amiga such as
RAM: - Navigate to that directory and double click the
Installprogram - Run the installer using all the default settings
- Download http://aminet.net/util/libs/Tran43pch.lha and once again extract it to a location of your choice
- Navigate to that directory and double click the
Installprogram - Run the installer using all the default settings
- Reboot your Amiga - It will not work until the system is restarted
- Insert your Workbench 1.x disk and copy
df0:devs/narrator.deviceto{AmigaGPTProgramDirectory}/devs/speech/34
- Insert your Workbench 2.0.x (you cannot use 2.1 because the speech libraries were removed after version 2.0.4) disk and copy
df0:devs/narrator.deviceto{AmigaGPTProgramDirectory}/devs/speech/37
- If your OS does not come with AHI installed, you can get it from https://aminet.net/package/driver/audio/ahiusr_4.18
- AmigaGPT for AmigaOS 4 uses the Flite device to provide speech synthesis. Download it from http://aminet.net/package/mus/misc/flite_device.
- Extract the archive and run the installer
- Launch the application by double-clicking the AmigaGPT icon
- You may also launch the app in the command line but before you do, run the command
STACK 32768to give the program 32kb of stack since the default stack size for apps launched from the shell is 4kb and this is not enough for AmigaGPT and will cause random crashes due to stack overflow. This is not required when you launch the app by double clicking the icon since the stack size is saved in the icon
When launched, AmigaGPT presents you with a choice of opening the app in a new screen or opening in Workbench. If you open in a new screen you have the ability to create a screen for the app to open in. AmigaGPT supports anything from 320x200 all the way up to 4k resolution if using a video card for RTG. Bear in mind text will appear very tiny in resolutions above 1080p so you may want to increase the font size settings from the View menu when the app opens.
When launching for the first time you will need to enter your OpenAI API key before you can start chatting. If you haven't already done so, create an OpenAI account and navigate to https://platform.openai.com/account/api-keys to generate an API key for use with AmigaGPT.
There are 2 main modes of operation: Chat and Image Generation. You can switch between them via the tabs in the top left corner.
When the app has opened, you are presented with a text input box. You can type any prompt into this box and press "Send" to see the GPT model's response. The generated text appears in the box above the input. You can choose to have this text read aloud using the "Speech" menu option. You can also select which model for OpenAI to use in the "OpenAI" menu option.
To the left of the chat box is a conversation list which you can use to go to another saved conversation. New conversations can be created with the "New chat" button and conversations can be removed with the "Delete chat" button.
To generate images, simply select your desired image generation model from the "OpenAI" menu then type your prompt in the text box then hit the "Create Image" button. When it has been downloaded to your Amiga, you are then able to open the image to your desired scale, or save a copy of the file to a new location on your Amiga. Do note however that AmigaGPT will automatially save all your generated images until you delete them. This is just in case you would like to create a copy elsewhere.
In the "Edit" menu, you'll find basic text editing commands like Cut, Copy, Paste and Clear. The "View" menu allows you to change the font used in the chat and the UI.
The "Project" menu includes an "About" option, which displays information about the program.
The latest DALL·E model released in Nov 2023.
The previous DALL·E model released in Nov 2022. The 2nd iteration of DALL·E with more realistic, accurate, and 4x greater resolution images than the original model.
GPT-4 is a large multimodal model (accepting text inputs and emitting text outputs today, with image inputs coming in the future) that can solve difficult problems with greater accuracy than any of our previous models, thanks to its broader general knowledge and advanced reasoning capabilities. For many basic tasks, the difference between GPT-4 and GPT-3.5 models is not significant. However, in more complex reasoning situations, GPT-4 is much more capable than any of our previous models.
| Model | Description | Max Tokens | Training Data |
|---|---|---|---|
| gpt-4o | Our most advanced, multimodal flagship model that’s cheaper and faster than GPT-4 Turbo. Currently points to gpt-4o-2024-05-13. | 128,000 | Up to Oct 2023 |
| gpt-4o-2024-05-13 | gpt-4o currently points to this version. | 128,000 | Up to Oct 2023 |
| gpt-4o-mini | Our affordable and intelligent small model for fast, lightweight tasks. GPT-4o mini is cheaper and more capable than GPT-3.5 Turbo. Currently points to gpt-4o-mini-2024-07-18. | 128,000 | Up to Oct 2023 |
| gpt-4o-mini-2024-07-18 | gpt-4o-mini currently points to this version. | Up to Oct 2023 | |
| gpt-4-turbo | The latest GPT-4 Turbo model with vision capabilities. Vision requests can now use JSON mode and function calling. Currently points to gpt-4-turbo-2024-04-09. | 128,000 | Up to Dec 2023 |
| gpt-4-turbo-2024-04-09 | GPT-4 Turbo with Vision model. Vision requests can now use JSON mode and function calling. gpt-4-turbo currently points to this version. | 128,000 | Up to Dec 2023 |
| gpt-4-turbo-preview | GPT-4 Turbo preview model. Currently points to gpt-4-0125-preview. | 128,000 | Up to Dec 2023 |
| gpt-4-0125-preview | The latest GPT-4 model intended to reduce cases of “laziness” where the model doesn’t complete a task. Returns a maximum of 4,096 output tokens. | 128,000 | Up to Dec 2023 |
| gpt-4-1106-preview | GPT-4 Turbo model featuring improved instruction following, JSON mode, reproducible outputs, parallel function calling, and more. Returns a maximum of 4,096 output tokens. This is a preview model. | 128,000 | Up to Apr 2023 |
| gpt-4 | Currently points to gpt-4-0613. | 8,192 | Up to Sep 2021 |
| gpt-4-32k-0613 | Snapshot of gpt-4-32k from June 13th 2023 with improved function calling support. | 32,768 | Up to Sep 2021 |
GPT-3.5 models can understand and generate natural language or code. Our most capable and cost effective model in the GPT-3.5 family is gpt-3.5-turbo which has been optimized for chat but works well for traditional completions tasks as well.
| Model | Description | Max Tokens | Training Data |
|---|---|---|---|
| gpt-3.5-turbo-0125 | The latest GPT-3.5 Turbo model with higher accuracy at responding in requested formats and a fix for a bug which caused a text encoding issue for non-English language function calls. Returns a maximum of 4,096 output tokens. | 16,385 | Up to Sep 2021 |
| gpt-3.5-turbo | Currently points to gpt-3.5-turbo-0613. Will point to gpt-3.5-turbo-1106 starting Dec 11, 2023. | 4,096 | Up to Sep 2021 |
| gpt-3.5-turbo-1106 | The latest GPT-3.5 Turbo model with improved instruction following, JSON mode, reproducible outputs, parallel function calling, and more. Returns a maximum of 4,096 output tokens. | 16,385 | Up to Sep 2021 |
You can either compile the code natively or with the Docker container.
If you would like to build this project from source you will need Bebbo's amiga-gcc toolchain here https://github.com/bebbo/amiga-gcc
Once installed, get the required other SDK's (AmiSSL, Translator, json-c) from https://github.com/sacredbanana/AmigaSDK-gcc and put these in your Amiga dev environment created in the above step.
Get this toolchain set up https://github.com/sba1/adtools
Once installed, get the required other SDK's (AmiSSL, Translator, json-c) from https://github.com/sacredbanana/AmigaSDK-gcc and put these in your Amiga dev environment created in the above step.
You may use pre-prepared Docker images that are able to compile both the AmigaOS 3 and AmigaOS 4 versions of the app.
Just install Docker on your machine and run the build_os3.sh or build_os4.sh scripts depending on which version of the app you want to build. If you want to perform a clean build, you can set the environment variable CLEAN=1 for example you can run CLEAN=1 ./build_os3.sh.
The build app will be saved to the /out directory.
AmigaGPT is licensed under the MIT License.
We welcome contributions to AmigaGPT! If you have a bug to report, a feature to suggest, or a change you'd like to make to the code, please open a new issue or submit a pull request.
- Cameron Armstrong (sacredbanana/Nightfox) https://github.com/sacredbanana/
- Mauricio Sandoval - Icon design
- Bebbo for creating the Amiga GCC toolchain https://github.com/bebbo
- OpenAI for making this all possible https://openai.com
- EAB and everyone in it for answering my questions https://eab.abime.net/
- Hyperion Entertainment for bringing us AmigaOS 3.2 https://www.hyperion-entertainment.com