ENI — Embedded Neural Interface
ENI is the EoS embedded neural interface — a low-latency, capability-secured stack that handles Neuralink-class 1024-channel acquisition at 30 kHz, EEG capture, signal conditioning, and a learned intent decoder, all on-device.
What ENI is
ENI bridges biosignal hardware and the rest of the EoS stack. It captures dense neural recordings (Neuralink-class 1024-channel arrays at 30 kHz, plus consumer EEG headsets), filters and spike-sorts in real time, and feeds a learned intent decoder that emits structured commands into eIPC.
Privacy and safety are first-class: raw neural data never leaves the device by default; the intent stream is gated by capability tokens; and a hardware kill-switch is wired into the EoS scheduler.
Features
The shape of ENI at a glance.
1024-Channel Capture
DMA-driven sampling at 30 kHz/ch with sub-millisecond pipeline jitter.
EEG & ECoG Support
OpenBCI, Muse, Emotiv, and custom analog front-ends via the EoS HAL.
Real-Time Spike Sorter
Online template-matching with drift compensation.
Intent Decoder
Configurable decoder (logistic / RNN / transformer) translating spikes → structured commands.
Capability-Gated Output
Intent stream is delivered through eIPC capabilities — apps can only see what they're authorized for.
Privacy By Default
Raw neural data is sandbox-only; no off-device transmission without explicit user consent.
Hardware Kill-Switch
Physical interlock wired into the kernel scheduler — stop signal acquisition in < 5 ms.
Calibration Toolkit
Per-user training UI with stored, re-loadable calibration profiles.
Research Tap
Optional, opt-in raw-data tap for clinical / research workflows under audit.
Open source on GitHub
ENI is Apache-2.0 licensed and developed in the open. Issues, discussions, and pull requests welcome.
In the EoS stack
ENI is the highlighted layer below.
Pairs well with
Sibling components that ENI commonly works alongside.
Ready to build with ENI?
Start with the docs, browse the source, or join the community.