digity 0.2.5

Python SDK for the Digity sensorized exohand

digity SDK

Python SDK for the Digity sensorized exohand — plug in the glove, stream live sensor data in three lines.

The Digity exohand is a sensorized glove that captures real-time hand motion data: joint angles for every finger, 6-axis IMU (accelerometer + gyroscope), and 6-channel capacitive touch per node. This SDK handles the USB connection, binary protocol decoding, and threading — so you get clean, typed Python objects at ~50 Hz without any low-level plumbing.

from digity import GloveStream, AnglesSensor

with GloveStream() as stream:
    for frame in stream:
        for sensor in frame.sensors:
            if isinstance(sensor, AnglesSensor):
                print(sensor.finger, sensor.samples[-1].angles_deg)

Use cases: motion capture, rehabilitation, robotics control, gesture recognition, HCI research, data collection.

---

Contents

• Installation
• Quick start
• GloveStream
• Data types
• Common patterns
• Error handling
• Real-time dashboard
• Agent mode — stream to cloud
• Publish over the network (ZMQ)
• Remote ZMQ receive
• Connection speed

---

Installation

# Core SDK — sensor data in Python only
pip install digity

# SDK + real-time local dashboard
pip install "digity[viz]"

# SDK + cloud agent relay
pip install "digity[agent]"

Requirements: Python 3.9 or later · Digity exohand connected via USB.

---

Quick start

Plug in the glove, then run:

from digity import GloveStream, AnglesSensor, ImuSensor, TouchSensor

with GloveStream() as stream:
    for frame in stream:
        print(f"side={frame.side}  node={frame.node_id}  seq={frame.seq}")

        for sensor in frame.sensors:

            if isinstance(sensor, AnglesSensor):
                angles = sensor.samples[-1].angles_deg
                print(f"  finger {sensor.finger} angles: {angles}")

            elif isinstance(sensor, ImuSensor):
                s = sensor.samples[-1]
                print(f"  finger {sensor.finger} acc={s.acc}  gyro={s.gyro}")

            elif isinstance(sensor, TouchSensor):
                print(f"  finger {sensor.finger} touch={sensor.channels}")

The with block handles connection and cleanup automatically. Press Ctrl+C to stop.

---

GloveStream

GloveStream(
    port=None,       # serial port e.g. "COM3" / "/dev/ttyUSB0" — None = auto-detect
    baud=921600,     # baud rate
    *,
    host=None,       # remote host IP for ZMQ mode
    zmq_port=5555,   # ZMQ port when using host
)

Method	Description
connect()	Open the connection. Called automatically by with.
disconnect()	Close the connection. Called automatically on with exit.
__iter__()	Yields GloveFrame objects. Blocks until the next frame arrives.

Specifying the USB port — the SDK auto-detects on all platforms. If auto-detection fails:

GloveStream(port="/dev/ttyUSB0")          # Linux
GloveStream(port="/dev/tty.usbserial-0001")  # macOS
GloveStream(port="COM3")                  # Windows

Stop from another thread:

import threading
from digity import GloveStream

stream = GloveStream()
stream.connect()

def stop_after(seconds):
    import time; time.sleep(seconds)
    stream.disconnect()

threading.Thread(target=stop_after, args=(10,), daemon=True).start()

for frame in stream:
    print(frame.seq)

---

Data types

GloveFrame

Every iteration of GloveStream yields one GloveFrame:

Field	Type	Description
ts	float	Host timestamp (seconds since epoch) when the frame arrived
side	str	"right" or "left"
group	str	"hand" or "arm"
node_id	int	PCB node number on the glove
seq	int	Packet counter 0–65535, wraps around
sensors	list[Sensor]	List of sensor readings in this frame

Sensor is Union[AnglesSensor, ImuSensor, TouchSensor].

---

AnglesSensor

Joint angles for one finger node.

Field	Type	Description
finger	int	Finger index — 0 = thumb, 1 = index, 2 = middle, 3 = ring, 4 = pinky
com	int	Communication line index
samples	list[AnglesSample]	One or more timestamped angle readings

AnglesSample

Field	Type	Description
ts_us	int	Sensor timestamp in microseconds
angles_deg	list[float]	Joint angles in degrees (5 values for hand group)

if isinstance(sensor, AnglesSensor):
    latest = sensor.samples[-1]
    print(latest.angles_deg)   # e.g. [12.3, 45.1, 30.0, 5.5, 2.0]

---

ImuSensor

6-axis IMU (accelerometer + gyroscope) for one finger node.

Field	Type	Description
finger	int	Finger index
com	int	Communication line index
samples	list[ImuSample]	One or more timestamped IMU readings

ImuSample

Field	Type	Description
ts_us	int	Sensor timestamp in microseconds
acc	tuple[int, int, int]	Accelerometer x/y/z — raw i16 counts
gyro	tuple[int, int, int]	Gyroscope x/y/z — raw i16 counts

if isinstance(sensor, ImuSensor):
    s = sensor.samples[-1]
    print(s.acc)    # e.g. (312, -128, 16384)
    print(s.gyro)   # e.g. (5, -12, 3)

---

TouchSensor

6-channel capacitive touch for one finger node.

Field	Type	Description
finger	int	Finger index
com	int	Communication line index
ts_us	int	Sensor timestamp in microseconds
channels	list[float]	6 values normalised 0.0–1.0
channels_raw	list[int]	6 raw ADC counts 0–4095

if isinstance(sensor, TouchSensor):
    print(sensor.channels)      # [0.0, 0.82, 0.0, 0.0, 0.41, 0.0]
    print(sensor.channels_raw)  # [0, 3358, 0, 0, 1680, 0]

---

Type annotation helper

from digity import Sensor  # Union[AnglesSensor, ImuSensor, TouchSensor]

def process(sensor: Sensor) -> None:
    if isinstance(sensor, AnglesSensor):
        ...

---

Common patterns

Record data to CSV

import csv
from digity import GloveStream, AnglesSensor

with GloveStream() as stream, open("recording.csv", "w", newline="") as f:
    writer = csv.writer(f)
    writer.writerow(["ts", "finger", "a0", "a1", "a2", "a3", "a4"])

    for frame in stream:
        for sensor in frame.sensors:
            if isinstance(sensor, AnglesSensor):
                a = sensor.samples[-1].angles_deg
                writer.writerow([frame.ts, sensor.finger] + a)

Detect a closed fist

from digity import GloveStream, AnglesSensor

with GloveStream() as stream:
    finger_angles = {}

    for frame in stream:
        for sensor in frame.sensors:
            if isinstance(sensor, AnglesSensor):
                finger_angles[sensor.finger] = sensor.samples[-1].angles_deg

        if len(finger_angles) == 5:
            avg_flex = sum(finger_angles[f][1] for f in range(1, 5)) / 4
            if avg_flex > 60:
                print("Fist detected!")

Read a saved recording

import json
from pathlib import Path

session_dir = Path("~/digity-recordings/admin/user01_grasp_2025-04-22T10-30-00").expanduser()
meta = json.loads((session_dir / "meta.json").read_text())
print(f"Duration: {meta['duration_s']:.1f}s  Frames: {meta['frame_count']}")

with open(session_dir / "recording.jsonl") as f:
    for line in f:
        frame = json.loads(line)
        print(frame["ts"], frame["side"])

---

Error handling

from digity import GloveStream, GloveNotFoundError

try:
    with GloveStream() as stream:
        for frame in stream:
            ...
except GloveNotFoundError:
    print("Glove not found — check the USB cable")
except KeyboardInterrupt:
    pass

GloveNotFoundError is raised at connection time if the glove port cannot be found.

---

Real-time dashboard (digity[viz])

digity[viz] ships a full local dashboard that visualises live hand data in 3-D, records labelled sessions, and serves as a local relay for remote agents — no internet required.

pip install "digity[viz]"

digity-viz              # opens as a desktop window
digity-viz --port COM3  # explicit serial port
digity-viz --browser    # open in system browser instead

Or from Python:

import digity.viz
digity.viz.start()

The dashboard runs at http://localhost:5001/chiros/ and opens automatically.

Dashboard pages:

Page	Description
Viewer	Live 3-D hand model with real-time joint angles
Record	Start/stop sessions, download .jsonl files, manage recordings
Setup	Recordings folder, agent token, update checker

Session recordings are saved to ~/digity-recordings/ by default. Each session creates a directory containing meta.json (duration, frame count, task label) and recording.jsonl (one JSON frame per line at ~50 Hz).

---

Agent mode (digity[agent])

Stream the glove directly to the cloud dashboard — no separate agent.exe needed.

pip install "digity[agent]"

# copy your token from the dashboard → Setup page
digity-agent --token YOUR_TOKEN

# or via the digity-viz command
digity-viz --agent --token YOUR_TOKEN

# explicit port or custom server URL
digity-agent --token YOUR_TOKEN --port COM3
digity-agent --token YOUR_TOKEN --url https://app.digity.de/chiros

The agent auto-detects the glove, connects to the server, and streams frames in real time. It reconnects automatically if either connection drops.

If digity-agent is not on PATH (common with Windows Store Python), use:

python -m digity.viz --agent --token YOUR_TOKEN

---

Publish over the network (ZMQ)

GlovePublisher broadcasts frames over a ZMQ PUB socket so multiple processes or machines can subscribe to the same glove.

from digity import GlovePublisher

# Standalone — blocks until Ctrl+C
with GlovePublisher() as pub:
    pub.run()

# Embedded in a custom loop
from digity import GloveStream, GlovePublisher

with GlovePublisher() as pub, GloveStream() as stream:
    for frame in stream:
        pub.publish(frame)
        # ... your processing here

GlovePublisher(
    zmq_port=5555,     # port to bind
    bind="tcp://*",    # bind address
)

Method	Description
start()	Open ZMQ PUB socket. Returns self.
stop()	Close socket.
publish(frame)	Publish one frame. Non-blocking, thread-safe.
run(port, baud)	Read from serial and publish. Blocks until Ctrl+C.

---

Remote ZMQ receive

On any machine that can reach the publisher over the network:

from digity import GloveStream

with GloveStream(host="192.168.1.10") as stream:        # default port 5555
    for frame in stream:
        print(frame.side, frame.sensors)

# Custom ZMQ port
with GloveStream(host="192.168.1.10", zmq_port=5556) as stream:
    ...

No USB connection is needed on the receiving machine.

---

Connection speed

The glove streams at approximately 50 Hz. Each for frame in stream call blocks until the next frame arrives (up to 1 second timeout before checking for errors).

The SDK uses a background thread and an internal queue so your processing code never stalls the serial buffer. If your code is slower than 50 Hz, frames are dropped to keep the stream live rather than accumulating memory.

---

Updating

pip install --upgrade digity
pip install --upgrade "digity[viz]"    # if using the dashboard
pip install --upgrade "digity[agent]"  # if using agent mode

---

License

© Digity. All rights reserved.