Tony

Tony is a graphical implementation of the Neuro API in python. Like Randy, he can answer actions/force commands automatically, but also allows the user to formulate the response Neuro would send themselves, as well as send non-forced actions whenever.

Note

I cannot guarantee that this implementation perfectly emulates what Neuro could/would do, or that it is error-free. There might be some things that I have overlooked.

Installation

A Python version of 3.10 or higher is required. Python versions 3.9 and below will not work.

The program can be installed from GitHub or from PyPI. If you want to modify the project and/or contribute, use the GitHub method.

Windows

Windows

From PyPI

This will install the package from PyPI using pip.

1. Run pip install neuro-api-tony

From GitHub

This will install the package in a virtual environment to not conflict with any global packages. Skip steps 2 and 3 if you don't want a virtual environment. All commands after step 1 should be run in the downloaded folder.

1. Clone the repository with git clone https://github.com/Pasu4/neuro-api-tony.git or download it from GitHub
2. Run python -m venv .venv
3. Run .\.venv\Scripts\activate
4. Run pip install -e .

MacOS

MacOS

Warning: This has not been tested, as I don't own a Mac.

From PyPI

This will install the package from PyPI using pip.

1. Run pip install neuro-api-tony

From GitHub

This will install the package in a virtual environment to not conflict with any global packages. Skip steps 2 and 3 if you don't want a virtual environment. All commands after step 1 should be run in the downloaded folder.

1. Clone the repository with git clone https://github.com/Pasu4/neuro-api-tony.git or download it from GitHub
2. Run python -m venv .venv
3. Run source ./.venv/bin/activate
4. Run pip install -e .

Fedora

Fedora

From PyPI

This will install the package from PyPI using pip.

1. Run sudo dnf install g++ gtk3-devel python-config
2. Run pip install neuro-api-tony

From GitHub

This will install the package in a virtual environment to not conflict with any global packages. Skip steps 2 and 3 if you don't want a virtual environment. All commands after step 2 should be run in the downloaded folder.

1. Run sudo dnf install g++ gtk3-devel python-config
2. Clone the repository with git clone https://github.com/Pasu4/neuro-api-tony.git or download it from GitHub
3. Run python -m venv .venv
4. Run source ./.venv/bin/activate
5. Run pip install -e .

Tested on Fedora 41 with Python 3.13.1.

Other Linux distributions

Other Linux distributions

Warning: Not all Linux distributions have been tested. You might have to install GTK+ in some form. On ubuntu-based systems, look for libgtk-3-dev. Also see https://github.com/wxWidgets/Phoenix/blob/wxPython-4.2.2/README.rst#prerequisites. If you run into further problems with a specific distribution, please submit an issue.

From PyPI

This will install the package from PyPI using pip.

1. Run pip install neuro-api-tony

From GitHub

This will install the package in a virtual environment to not conflict with any global packages. Skip steps 2 and 3 if you don't want a virtual environment. All commands after step 1 should be run in the downloaded folder.

1. Clone the repository with git clone https://github.com/Pasu4/neuro-api-tony.git or download it from GitHub
2. Run python -m venv .venv
3. Run source ./.venv/bin/activate
4. Run pip install -e .

Updating

Note

This repository was renamed from neuro-api-human-control to neuro-api-tony. GitHub should automatically redirect requests to the new URL, but in case you run into problems because of this, run git remote set-url origin https://github.com/Pasu4/neuro-api-tony.git in the repository.

If you have installed Tony by cloning the repository, pull the latest changes using git. If you downloaded from GitHub without cloning it, you're better off re-downloading and re-installing than updating the files manually. All commands should be run in the root folder of the repository.

1. Run git pull
2. Run .\.venv\Scripts\activate on Windows, or source ./.venv/bin/activate on Linux / Mac (Skip this step if you didn't set up a virtual environment)
3. Run pip install -e . to install any potential new dependencies

If you installed Tony from PyPI, you can update using pip install --upgrade neuro-api-tony.

Usage

This assumes you have set up a virtual environment during installation. Skip step 1 if you haven't, or if your virtual environment is already activated.

1. In the folder of the application, run .\.venv\Scripts\activate on Windows, or source ./.venv/bin/activate on Linux / Mac
2. Run neuro-api-tony

The application window should now open. The left panel will display the actions once they have been registered. They can be executed by clicking the "execute" button below the list. The right panel shows an event log, below that is a smaller panel with some controls. After sending an action command to the game, the next action can only be sent after the action/result command has been received. When the game sends an actions/force command, a window will open that only shows the applicable actions, and will only close once the action/result command indicates success.

By default, Tony opens a websocket server on port 8000 (websocket URL ws://localhost:8000), this can be changed with command line arguments.

Note

When working with the Unity SDK, you need to focus the Unity editor after sending an action for the game to receive the action.

Command line arguments

Copy-pasted from the help message:

-h, --help:
    Show this help message.

-a, --addr, --address <ADDRESS>:
    The address to start the websocket server on. Default is localhost.

-l, --log, --log-level <LOG_LEVEL>:
    The log level to use. Default is INFO. Must be one of: DEBUG, INFO,
    WARNING, ERROR, CRITICAL.

-p, --port <PORT>:
    The port number to start the websocket server on. Default is 8000.

-v, --version:
    Show the version of the program.

Actions panel

To execute an action, the game first needs to send an actions/register command. After that, an entry should appear in the left panel showing the name of the action and its description.

There are some buttons at the bottom of the panel:

• Execute: Opens a window where you can enter the content of the reply and send it via the "Send" button. The window will already contain some sample JSON that validates against the schema. By default, your input will be parsed and validated against the schema before sending, you can turn this off in the control panel.
• Delete: Manually unregisters the selected action. This is not something Neuro would normally do.
• Unlock: Unlocks the execute button while waiting for an action/result command. This is probably not something Neuro would normally do.

Forced actions

If an actions/force command is received, a "Forced action" window will open, showing all applicable actions and the query and state of the command. Executing actions from here works the same as from the main actions panel. Once the action has been sent, the window will close automatically. You can also close the window manually, this will ignore the forced action and allow you to execute any registered action.

Logs

The log panel on the top right has three different tabs:

• The system tab logs miscellaneous messages with color-coded tags:
  • Debug (gray): Things that usually should be handled internally by an SDK (e.g. action IDs), as well as some internals of the application. Debug messages alone are not a cause for concern.
  • Info (blue): Things that will likely not cause problems with Neuro, but might point to some other issue (e.g. action/result with no message).
  • Warning (yellow): Things that do not comply with the API specification, but which Tony can still tolerate (e.g. trying to register actions before sending startup). These will likely cause problems with Neuro.
  • Error (red): Things that make it impossible to process a command (e.g. receiving invalid JSON). These will definitely cause problems with Neuro.
  • Critical (dark red): Something went wrong and Tony will likely have to be restarted.
• The commands tab logs incoming and outgoing commands in a more condensed format than the raw tab.
• The context tab shows everything that Neuro would get to read directly, which is the content of context commands, the description of actions, the state and query of actions/force commands, and the message of action/result commands. Silent contexts are displayed in gray and ephemeral contexts in light blue. It has the following tags:
  • Context: Message is from a context command.
  • Silent: Message is from a silent context command.
  • State: Message is the state of an actions/force command.
  • Query: Message is the query of an actions/force command.
  • Ephemeral: Message is the query or state of an actions/force command with ephemeral context.
  • Action: Message is the description of an action, logged at registration.
  • Result: Message is from an action/result command. The color denotes whether the result indicates success (green) or failure (red).
• The raw tab shows the full data sent over the websocket, as well as who sent that data. If it is valid JSON, it will be formatted for easier viewing.

Control panel

Normal controls

The control panel has some checkboxes and buttons that change the behavior of the application (Some of these controls have been moved elsewhere).

• Ignore forced actions: If checked, will not open the "Forced action" dialog when an actions/force command is received. You have to execute the action yourself from the left panel. Since the forced action is ignored, you can execute any registered action.
• Auto-answer: If checked, will immediately send the pre-generated JSON of a random valid action instead of opening the "Forced action" window when an actions/force command arrives. This behavior is similar to what Randy does.
• L*tency: Will delay sending commands by the specified time. Must be non-negative and not greater than 10000ms.
• Log level: Will show only messages with an equal of higher log level than the selection. For example, selecting "Warning" will not show Debug or Info messages, but still show Warning, Error and System messages.
• Clear: Clears all log tabs and the export log.
• Export: Opens a dialog to save the logs to a file. If you submit a bug report, consider attaching this file. All messages will be included, regardless of configured log level.
• Maximize: Maximizes the log panel to fill the entire window. Can be restored using the restore size button.
• Don't validate: If checked, will allow you to send a message that does not validate against the schema or contains invalid JSON.

Experimental controls

These controls are proposed features in the Neuro API that will likely not work unless specifically implemented. Use at your own risk!

• Clear and reregister: Will unregister all currently registered actions and send an actions/reregister_all command to the game.
• Graceful shutdown: Will send a shutdown/graceful command to the game, indicating it should save the game and return to the main menu at the next opportunity.
• Cancel shutdown: Will send a shutdown/graceful command with its wants_shutdown field set to false to the game, signaling to cancel a previous shutdown request.
• Immediate shutdown: Will send a shutdown/immediate command to the game, indicating that the game will (not should!) be shut down within the next few seconds.

Known issues

• The last button on the control panel sometimes does not appear when the app is started. Resizing the window should make it appear.

Screenshots