neuroskill

neuroskill is a command-line interface for the NeuroSkill real-time EXG analysis API. It communicates with a locally running Skill server over WebSocket or HTTP, giving you instant terminal access to EEG brain-state scores, sleep staging, session history, annotations, similarity search, and more.

This is the Python port of the TypeScript neuroskill CLI. It is a faithful port — same commands, same flags, same output format, same transport negotiation.

⚠️ Research Use Only. All metrics are experimental outputs derived from consumer-grade EXG hardware. They are not validated clinical measurements, not FDA/CE-cleared, and must not be used for diagnosis, treatment decisions, or any medical purpose.

Paper

---

Table of Contents

1. Features
2. Requirements
3. Installation
4. Quick Start
5. Transport
6. Commands
7. Output Modes
8. Global Options
9. Examples
10. How to Cite
11. License

---

Features

• Real-time EEG scores — focus, relaxation, engagement, meditation, cognitive load, drowsiness
• Consciousness metrics — Lempel-Ziv Complexity proxy, wakefulness, information integration
• PPG / HRV — heart rate, RMSSD, SDNN, pNN50, LF/HF, SpO₂, Baevsky stress index
• Sleep staging — automatic per-epoch classification and session-level summary
• Session history — list, compare, and UMAP-project all past recording sessions
• Annotations — create timestamped labels and search them by free text or EEG similarity
• Interactive graph search — cross-modal 4-layer graph (labels → EEG → labels)
• Dual transport — WebSocket (full-duplex, live events) and HTTP REST (curl-friendly)
• Pipe-friendly — --json flag emits clean JSON to stdout; informational lines go to stderr
• Cross-platform — Python ≥ 3.9, Windows / macOS / Linux

---

Requirements

Dependency	Version
Python	≥ 3.9
Skill server	running locally (auto-discovered via mDNS or lsof)

---

Installation

Via pip / uv (recommended)

pip install neuroskill

or with uv:

uv tool install neuroskill

After installation the neuroskill binary is available globally:

neuroskill status

From source

git clone https://github.com/NeuroSkill-com/neuroskill-py
cd neuroskill-py
uv sync
uv run neuroskill status

---

Quick Start

# Full device / session / scores snapshot
neuroskill status

# Pipe raw JSON to jq (or python -m json.tool)
neuroskill status --json | python3 -m json.tool
neuroskill status --json | jq '.scores'

# Stream broadcast events for 10 seconds
neuroskill listen --seconds 10

# Print full help with examples
neuroskill --help

---

Transport

neuroskill auto-discovers the Skill server port via:

1. --port <n> flag (skips all discovery)
2. mDNS (_skill._tcp service advertisement, 5 s timeout)
3. lsof / pgrep fallback (probes each TCP LISTEN port)

WebSocket (default)

Full-duplex, low-latency. Supports live event streaming. Used automatically when the server is reachable.

neuroskill status --ws          # force WebSocket

HTTP REST

Request/response only. Compatible with curl, Python requests, or any HTTP client.

neuroskill status --http        # force HTTP

# Equivalent curl call:
curl -s -X POST http://127.0.0.1:8375/ \
  -H "Content-Type: application/json" \
  -d '{"command":"status"}'

Auto (neither flag)

The CLI probes WebSocket first and silently falls back to HTTP. Informational messages go to stderr so JSON piping is never polluted.

---

Commands

Command	Description
status	Full device / session / embeddings / scores snapshot
session [index]	All metrics + trends for one session (0 = latest, 1 = previous, …)
sessions	List all recording sessions across all days
label "text"	Create a timestamped annotation on the current moment
search-labels "query"	Search labels by free text (text / context / both modes)
interactive "keyword"	Cross-modal 4-layer graph search (labels → EEG → found labels)
search	ANN EEG-similarity search (auto: last session, k = 5)
compare	Side-by-side A/B metrics (auto: last 2 sessions)
sleep [index]	Sleep staging — index selects session (0 = latest)
calibrations [list|get <id>]	List calibration profiles or inspect one by ID
calibrate	Open calibration window and start immediately
timer	Open focus-timer window and start work phase immediately
notify "title" ["body"]	Show a native OS notification
say "text"	Speak text aloud via on-device TTS
umap	3-D UMAP projection with live progress bar
listen	Stream broadcast events for N seconds
raw '{"command":"..."}'	Send arbitrary JSON and print full response

---

Output Modes

Flag	Behaviour
(none)	Human-readable colored summary to stdout
--json	Raw JSON only — pipe-safe, no colors
--full	Human-readable summary and colorized JSON

---

Global Options

--port <n>       Connect to explicit port (skips mDNS discovery)
--ws             Force WebSocket transport
--http           Force HTTP REST transport
--json           Output raw JSON (pipeable to jq / python -m json.tool)
--full           Print JSON in addition to human-readable summary
--poll <n>       (status) Re-poll every N seconds
--mode <m>       Search mode for search-labels: text|context|both (default: text)
--k <n>          Number of nearest neighbors for search / search-labels
--ef <n>         HNSW ef parameter for search-labels (default: max(k×4, 64))
--k-text <n>     (interactive) k for text-label search (default: 5)
--k-eeg <n>      (interactive) k for EEG-similarity search (default: 5)
--k-labels <n>   (interactive) k for label-proximity search (default: 3)
--reach <n>      (interactive) temporal reach in minutes around EEG points (default: 10)
--dot            (interactive) Output Graphviz DOT format
--context "..."  (label) Long-form annotation body stored with the label
--at <utc>       (label) Backdate to a specific unix second (default: now)
--voice <name>   (say) Voice name (e.g. Jasper); uses server default when omitted
--profile <p>    (calibrate) Profile name or UUID to run (default: active profile)
--seconds <n>    (listen) Duration in seconds (default: 5)
--trends         (sessions) Show per-session metric trends
--no-color       Disable ANSI colors (also honours NO_COLOR env var)
--version        Print CLI version and exit
--help           Show full help with examples

---

Examples

# Device snapshot
neuroskill status

# Pipe scores to jq or python
neuroskill status --json | jq '.scores.focus'
neuroskill status --json | python3 -c "import sys,json; d=json.load(sys.stdin); print(d['scores']['focus'])"

# Poll status every 5 seconds
neuroskill status --poll 5

# Latest session metrics + trends
neuroskill session 0

# List sessions, show per-session metric trends
neuroskill sessions --trends

# Label the current moment
neuroskill label "started meditation"
neuroskill label "breathwork" --context "box breathing 4-4-4-4, 10 min"
neuroskill label "retrospective note" --at 1740412800

# Search past labels
neuroskill search-labels "meditation" --mode both --k 10

# 4-layer interactive graph search
neuroskill interactive "focus" --k-eeg 10 --reach 15
neuroskill interactive "anxiety" --dot | dot -Tsvg > graph.svg

# Sleep staging for latest session
neuroskill sleep 0

# Compare last two sessions
neuroskill compare

# ANN EEG similarity search
neuroskill search

# UMAP projection
neuroskill umap

# Calibration management
neuroskill calibrations
neuroskill calibrations get 3
neuroskill calibrate --profile "Eyes Open/Closed"

# Focus timer
neuroskill timer

# TTS and notification
neuroskill say "Calibration complete."
neuroskill say "Break time." --voice Jasper
neuroskill notify "Session done" "Great work!"

# Stream events for 30 seconds
neuroskill listen --seconds 30

# Send arbitrary JSON command
neuroskill raw '{"command":"status"}'
neuroskill raw '{"command":"search","start_utc":1740412800,"end_utc":1740415500,"k":3}'

# Force HTTP + specific port
neuroskill status --http --port 8375

---

How to Cite

If you use neuroskill or the Skill EEG platform in academic work, please cite it as:

BibTeX

@software{neuroskill2025,
  title        = {neuroskill: A Command-Line Interface for the Skill Real-Time EEG Analysis API},
  author       = {Nataliya Kosmyna and Eugene Hauptmann},
  year         = {2026},
  version      = {0.0.1},
  url          = {https://github.com/NeuroSkill-com/neuroskill},
  note         = {Research use only. Not a validated clinical tool.}
}

If you are citing the underlying Skill EEG analysis platform specifically:

@software{skill2025,
  title        = {NeuroSkill: Real-Time EEG Analysis Platform},
  author       = {Nataliya Kosmyna and Eugene Hauptmann},
  year         = {2026},
  url          = {https://neuroskill.com},
  note         = {Consumer-grade EEG processing pipeline with WebSocket and HTTP APIs.
                  Research use only. Not FDA/CE-cleared.}
}

---

License

GPLv3 — see LICENSE.