nullpunkt 1.0.0

Nullpunkt behavioral anomaly detection API client

Nullpunkt

Python client for the Nullpunkt behavioral anomaly detection API by Nohvel.

Install

pip install nullpunkt

Quickstart

from nullpunkt import NullpunktClient

client = NullpunktClient(api_key='npk_...')

# Fit baseline from normal-behavior data
client.fit_baseline('normal_scene.mp4', wait=True)

# Run detection
result = client.detect('scene_to_analyze.mp4', wait=True)

print(f"Anomaly rate: {result.anomaly_rate:.1%}")
print(f"Anomalous frames: {result.anomaly_frames} / {result.total_frames}")

for event in result.events:
    print(f"  Frame {event.frame}: score={event.score:.3f}, agents={event.agents}")

Authentication

Pass your API key directly or set the NULLPUNKT_API_KEY environment variable:

export NULLPUNKT_API_KEY=npk_...

client = NullpunktClient()  # reads from environment

Get an API key at api.nohvel.com/portal.

Supported Input Formats

• Video files: .mp4, .avi, .mkv
• Pre-tracked CSV: frame, track_id, cx, cy, w, h
• Pre-tracked JSON: {"frames": [{"frame_id": 0, "detections": [...]}]}

Async Jobs

By default, fit_baseline() and detect() block until the job completes. For non-blocking usage, pass wait=False and poll manually:

job_id = client.detect('scene.mp4', wait=False)

while True:
    status = client.get_job_status(job_id)
    if status['status'] == 'complete':
        break
    time.sleep(10)

result = client.get_results(job_id)

Documentation

Full API reference: nohvel.com/docs

Support

[e-mail]