Veriafy - Universal File Classification Platform

Install Veriafy

If you haven't already, install Veriafy using the installation script:

curl -fsSL https://get.veriafy.com/install.sh | sh

Initialize Veriafy

Set up your local Veriafy environment:

veriafy init

This creates a ~/.veriafy directory with configuration and model storage.

Download a Model

Pull a classification model from the marketplace:

veriafy pull veriafy/nsfw-classifier

This downloads the official NSFW classifier model (~150MB).

Classify a File

Run classification on any image — Veriafy never accesses the actual content:

veriafy classify image.jpg --model veriafy/nsfw-classifier

Example output:

{ "vector_id": "v_8f3a2b1c4d5e6f7a", "file_type": "image", "extractor": "pdq_clip", "model": "veriafy/nsfw-classifier", "categories": { "safe": 0.98, "suggestive": 0.02, "explicit": 0.00 }, "action": "allow", "confidence": 0.98, "processing_time_ms": 4.2 }

Use the Python SDK

Integrate Veriafy into your Python applications:

pip install veriafy

Example code:

from veriafy import Veriafy

# Initialize client (local mode, no API key needed)
client = Veriafy()

# Classify a file
result = client.classify("document.pdf", model="veriafy/fraud-detection")

print(f"Vector ID: {result.vector_id}")
print(f"Categories: {result.categories}")
print(f"Confidence: {result.confidence}")
print(f"Action: {result.action}")  # 'allow', 'flag', or 'block'

# Batch processing
results = client.classify_batch(
    files=["doc1.pdf", "doc2.pdf", "doc3.pdf"],
    model="veriafy/fraud-detection"
)

for r in results:
    print(f"{r.file}: {r.action}")