FlyEye v0.5.0-alpha.9 — Publication & Citation Release

FlyEye v0.5.0-alpha.9 — Publication & Citation Release

FlyEye Developer Kit is now publicly distributed and archived for external reuse and citation.

Public distribution

• Live demo: https://fly.fde.fan
• GitHub: https://github.com/tubban1/fly-eye
• Zenodo DOI: https://doi.org/10.5281/zenodo.22829797
• Hugging Face: https://huggingface.co/datasets/Tubban/flyeye-escape-neuron-v1
• PyPI: https://pypi.org/project/flyeye-graphs/0.1.0/
• npm: five packages under the @fly-eye scope, version 0.1.0

Published developer kit

• @fly-eye/graph-core@0.1.0
• @fly-eye/runtime@0.1.0
• @fly-eye/escape-neuron-v1@0.1.0
• @fly-eye/connectome-tools@0.1.0
• @fly-eye/benchmarks@0.1.0
• flyeye-graphs==0.1.0

Compact neuron graph

• 3,376 neurons
• 78,797 signed edges
• 959,092-byte FLYGRAPH asset
• manifest, report, aggregate fast profile, citation metadata, license metadata, and checksums

Citation

FlyEye Contributors (2026). FlyEye Developer Kit: reusable Drosophila connectome graphs and runtime tooling (v0.5.0-alpha.9). Zenodo.

DOI: 10.5281/zenodo.22829797

https://doi.org/10.5281/zenodo.22829797

Scientific boundary

MaleCNS-derived connectivity data and FlyEye modeled dynamics remain explicitly separated. The wiring is connectome-derived; camera encoding and runtime neural dynamics are models.

FlyEye v0.5.0-alpha.8 — Developer Kit

FlyEye v0.5.0-alpha.8 — Developer Kit

This release turns FlyEye from a single interactive camera experiment into a
reusable developer surface for Drosophila connectome projects.

Highlight: escape-neuron-v1

A compact neuron-level escape corridor is generated reproducibly from the pinned
MaleCNS-derived source graph:

Asset	Neurons	Edges	Bytes
pinned 70K source graph	70,000	798,715	9,864,604
flyeye.escape-neuron.v1	3,376	78,797	959,092

The compact profile retains:

• 126 LC4/loom seed neurons;
• 26 selected direct escape descending-neuron targets;
• directed seed-to-output paths within the declared hop budget;
• all signed internal edges among retained nodes;
• original 70K graph indices;
• MaleCNS body IDs;
• source commit/hash and derived graph hash;
• extraction parameters and scientific-boundary metadata.

Six foundation packages

@flyeye/escape-neuron-v1

Self-contained compact neuron-level graph package.

@flyeye/graph-core

FLYGRAPH v1 parser, validators, loaders, ID helpers and adjacency utilities.
Contains no behavior model.

@flyeye/runtime

Explicitly modeled group-level escape dynamics. Separates modeled dynamics from
real connectome data.

flyeye-graphs (Python)

Standard-library Python loader with optional NetworkX conversion.

@flyeye/connectome-tools

CLI for graph inspection and deterministic pathway extraction.

@flyeye/benchmarks

Compatibility vectors and graph-integrity checks. These are software-contract
benchmarks, not biological ground truth.

Fly Eye application

v0.5 also includes:

• camera/fingertip/optical/brain self-check;
• camera-motion lockout and stable re-arming;
• hand-independent optical fingertip tracking;
• Replay 2.0 with perception and brain layers;
• model-relative causal timeline;
• Sneak Up and Scare Fast modes;
• server-validated scoring and leaderboard API;
• privacy-safe local result cards;
• Fly Vision 2.0;
• optional neuron-level advanced inspector.

Licensing

• FlyEye software: MIT.
• MaleCNS-derived graph assets: upstream CC BY 4.0 attribution preserved in
  profile manifests and DATA_LICENSE.md.

Scientific boundary

Real connectivity data and modeled dynamics are intentionally separate. The
neuron graph preserves selected signed MaleCNS-derived edges. Camera encoding,
group dynamics, thresholds, readouts and visual explanations are modeled unless
separately validated.

Documentation

Start with:

• docs/FLYEYE_DEVELOPER_KIT.md
• docs/FLYEYE_GRAPHS.md
• DATA_LICENSE.md
• CITATION.cff