Plek 1 Interactive Demo Playground

Context / Input Text Task Prompt / Question (Directive Lens) Presets (3-Parameter Directed Scenarios)

Department Routing Urgency Audit Safety Filter Logic Sanity Check Staff Helpfulness Double Charge

Candidate Options (Comma-Separated)

Model Decision Output

technical 50.0% Confidence

Model ID: Drenel/plek-1 Latency: ~12 ms

Run Plek-1 in Python (Hugging Face Transformers)

from transformers import AutoModel

# 1. Load Plek-1 directly from Hugging Face Hub
model = AutoModel.from_pretrained("Drenel/plek-1", trust_remote_code=True)

# 2. 3-Parameter Directed Decision Classification
result = model.predict(
    context="The checkout page crashed with error code 500",
    question="Which department should handle this issue?",
    choices=["technical", "customer care", "finance", "security", "legal"]
)

print(result["prediction"])     # Output: technical
print(result["confidence"])     # Output: 0.500
print(result["probabilities"])  # Calibrated probability distribution

# 3. 200,000+ Token Ultra-Long Context Processing
huge_document = open("large_contract.txt").read()
long_result = model.predict_long(
    long_context=huge_document,
    question="What is the overall legal risk level of this agreement?",
    choices=["Compliant Contract", "High-Risk Liability", "Non-Binding Terms"]
)