rlbot_flatbuffers 0.18.11

rlbot-flatbuffers

A Python module implemented in Rust for fast and safe serialization and deserialization of RLBot's flatbuffers

The goal of this project

A majority of the code is auto-generated by codegen/ upon first compile using the RLBot's schema as defined by the flatbuffers-schema submodule.

This includes the code generated by Planus (src/planus_flat.rs), the Python wrapper binds to the generated Rust code (src/python/), and the Python type hints (rlbot_flatbuffers.pyi).

Usage of this API should not significantly differ from RLBot v4 to reduce developer confusion, while not holding back changes that would make the API easier to work with.

Minimum support Python version

The crate used to generate Python binds (PyO3) supports all the way back to Python 3.7, however the minimum supported Python version is 3.10 for a few reasons:

1. RLBot v4 currently runs Python 3.11
2. The RLBot v5's Python interface has a minimum Python version of 3.11, but the difference between 3.10 and 3.11 doesn't mean much for these binds specifically.
3. Python 3.10 is the version of Python that added match/case
4. Python 3.7 & 3.8 are EOL, with 3.9's EOL date being 2025-10

Dev setup

• Ensure Python 3.10+ is installed
• Install uv (if not already installed)
  • curl -LsSf https://astral.sh/uv/install.sh | sh
• Sync the Python environment (this builds the Rust extension)
  • uv sync
• If you've made changes to the Rust code and uv sync doesn't rebuild, force reinstall:
  • uv sync --reinstall-package rlbot-flatbuffers

To use in another Python environment, like if testing python-interface, you can build a wheel:

• uv run maturin build --release
• (In another environment) pip install path/to/file.whl

The exact path of the wheel will be printed by maturin, just copy+paste it.

Running tests

Tests are written with pytest and live in the tests/ directory. Run them with:

uv run pytest

Basic usage

All classes and methods should have types hints readable by your IDE, removing the guesswork of common operations.

Creating

import rlbot_flatbuffers as flat

desired_ball = flat.DesiredBallState(
    physics=flat.Physics(
        location=flat.Vector3Partial(z=200),
        velocity=flat.Vector3Partial(x=1500, y=1500),
        angular_velocity=flat.Vector3Partial(),
    ),
)

desired_game_info = flat.DesiredGameInfoState(
    world_gravity_z=-100,
    game_speed=2,
)

desired_game_state = flat.DesiredGameState(
    ball_state=desired_ball,
    game_info_state=desired_game_info,
)

In the above code, we:

• Set the ball to:
  • Location (0, 0, 200)
  • Velocity (1500, 1500, 0)
  • Angular velocity of (0, 0, 0)
• Don't set the car states
• Set the game info state:
  • World gravity to -100
  • Game speed to 2x default
  • Don't set end match or paused
• Don't set any console commands

All values are optional when creating a class and have the proper defaults.

Reading values

import rlbot_flatbuffers as flat

def handle_packet(packet: flat.GamePacket):
    if packet.match_info.match_phase not in {
        flat.MatchPhase.Active,
        flat.MatchPhase.Kickoff,
    }:
        # Return early if the game isn't active
        return

    # Print the ball's location
    print(packet.ball.physics.location)

    for car in packet.players:
        # Print the every car's location
        print(car.physics.location)

The goal of the above was to feel familiar to RLBot v4 while providing a more Pythonic interface.

• Unions aren't custom types, rather a normal Python variable that can be 1 of a few types.

  • For example, previously BallInfo.shape had the separate type CollisionShape which contained the union's data but the type is now just BoxShape | CylinderShape | SphereShape

• Classes implement __match_args__ for easy destructuring via the match/case pattern.

  • Enums can still be used to match against the type, they just can't be destructured.

• Classes and enums properly implement __repr__, with __str__ being an alias.

• Enums implement __hash__, __int__ and __eq__.

• Lists no longer have num_x fields accompanying them, they are just Python lists of the appropriate length.

• Classes implement pack and unpack, which are used to serialize and deserialize data.

  • These are public methods that can be used directly for any purpose, for example saving flat.GamePacket to a file.

• Auto-generated python type stub (.pyi) generation that includes doc comments from the Flatbuffers schema