# flylabs > One brain for every task. The complete fruit fly connectome (139,248 neurons, 50M signed synapses) as a fixed neural architecture, trained per task. Standard and custom adapters for input and output, optional classical co-processors. Every weight set is free up to 100k monthly active devices per platform. ## Engine - [FlyCore](https://flylabs.dev/docs/engine/): sparse connectome forward pass; 19,262 afferent, 118,497 intrinsic, 1,489 efferent neurons. Source: https://github.com/flylabs-dev/flycore. Install: pip install flycore. - [Adapters](https://flylabs.dev/docs/adapters/): how signals map onto afferent regions and how outputs are read. - [Training](https://flylabs.dev/docs/training/): imitation and reinforcement on the full connectome. ## Weight sets (available) - [Cursor: Mouse control](https://flylabs.dev/models/cursor/): Cursor turns a screen and a target into mouse motion: go to x,y, click, double-click, right-click, drag and drop. Trained on human recordings, so the path has the acceleration, the overshoot and the correction of a real hand. Platforms: Python, Rust, JavaScript. Sources: [Hugging Face](https://huggingface.co/flylabs/cursor), [Engine](https://github.com/flylabs-dev/flycore), [Docs](https://flylabs.dev/docs/cursor/). ## Weight sets (beta) - [Keys: Keyboard control](https://flylabs.dev/models/keys/): Keys produces key-down and key-up events for text and shortcuts with the inter-key timing of a person typing, including chords and held modifiers. Platforms: Python, Rust. Sources: [Hugging Face](https://huggingface.co/flylabs/keys), [Engine](https://github.com/flylabs-dev/flycore), [Docs](https://flylabs.dev/docs/keys/). - [Pad: Game controller](https://flylabs.dev/models/pad/): Pad maps game frames to a standard controller: two analogue sticks, triggers and buttons. Trained per game by imitation and reinforcement, on the same brain. Platforms: Python, Rust. Sources: [Hugging Face](https://huggingface.co/flylabs/pad), [Engine](https://github.com/flylabs-dev/flycore), [Docs](https://flylabs.dev/docs/pad/). - [Gaze: Object detection](https://flylabs.dev/models/gaze/): Gaze detects and tracks objects in video through the connectome's visual system, with a classical CNN co-processor for class labels. Built for motion, so it finds what moves first. Platforms: Python, JavaScript, Rust. Sources: [Hugging Face](https://huggingface.co/flylabs/gaze), [Engine](https://github.com/flylabs-dev/flycore), [Docs](https://flylabs.dev/docs/gaze/). - [Sono: Audio events](https://flylabs.dev/models/sono/): Sono detects sound events in a live stream: a door, a bark, a fall, a name. Audio enters through the auditory afferents of the connectome; the answer is an event with a timestamp. Platforms: Python, JavaScript, Rust. Sources: [Hugging Face](https://huggingface.co/flylabs/sono), [Engine](https://github.com/flylabs-dev/flycore), [Docs](https://flylabs.dev/docs/sono/). - [Track: Motion tracking](https://flylabs.dev/models/track/): Track follows a chosen object through a video with the connectome's motion detectors: no class labels, no re-detection, just a point and a velocity, every frame. Platforms: Python, JavaScript, Rust. Sources: [Hugging Face](https://huggingface.co/flylabs/track), [Engine](https://github.com/flylabs-dev/flycore), [Docs](https://flylabs.dev/docs/track/). - [Dictate: Speech to text](https://flylabs.dev/models/dictate/): Dictate pairs a compact speech co-processor with the connectome. The co-processor produces words; the brain decides segmentation, turn-taking and when a command is complete. Platforms: Python, JavaScript. Sources: [Hugging Face](https://huggingface.co/flylabs/dictate), [Engine](https://github.com/flylabs-dev/flycore), [Docs](https://flylabs.dev/docs/dictate/). - [Intent: Command understanding](https://flylabs.dev/models/intent/): Intent turns a short instruction plus a screen into a target embedding the control weights can act on. It is the bridge between language and motion, and it ships as a co-processor for Cursor and Keys. Platforms: Python. Sources: [Hugging Face](https://huggingface.co/flylabs/intent), [Engine](https://github.com/flylabs-dev/flycore), [Docs](https://flylabs.dev/docs/intent/). ## Weight sets (in development) - [Servo: Actuator control](https://flylabs.dev/models/servo/): Servo closes the loop with physical actuators: joint angles in, torques out, on the Rust runtime at 1 kHz. Trained in simulation on a body model, transferred to hardware. Platforms: Rust, Python. Sources: [Hugging Face](https://huggingface.co/flylabs/servo), [Engine](https://github.com/flylabs-dev/flycore), [Docs](https://flylabs.dev/docs/servo/). ## Categories - [Control weights](https://flylabs.dev/control/): Weights that move something: a cursor, a key, a game pad, a servo. The fly brain was built to turn perception into motion in milliseconds. - [Perception weights](https://flylabs.dev/perception/): Weights that watch and listen: objects in a video frame, events in audio, motion across a scene. Fed through the compound eye and the antennae of the connectome. - [Language weights](https://flylabs.dev/language/): Weights that turn speech and commands into something a machine can act on. A language co-processor does the words; the brain does the timing and the decision. ## Platforms - [FlyCore for Python and PyTorch](https://flylabs.dev/python/): The reference engine. Sparse connectome forward pass in PyTorch, training loops included, CUDA and Apple silicon. Install: pip install flycore - [FlyCore runtime for Rust and embedded](https://flylabs.dev/rust/): Inference only, no Python. One static library that runs any weight set on a robot controller, a desktop agent or a game process. Install: cargo add flycore - [FlyCore for the browser and Node](https://flylabs.dev/js/): Run a weight set in the browser on WebGPU, with a WebGL fallback, or in Node with the native runtime. Install: npm i @flylabs/flycore ## Resources - [Machine-readable catalog](https://flylabs.dev/catalog.json): every weight set and platform as JSON. - [Full content](https://flylabs.dev/llms-full.txt): every overview and use case in one file. - [Engine on GitHub](https://github.com/flylabs-dev/flycore) - [Weights on Hugging Face](https://huggingface.co/flylabs)