GitHub Repository

Plug any physical randomness source into LLM token sampling. Works with vLLM, Transformers, and llama.cpp

1 starsPython

Entropick – Plug quantum/hardware RNGs into LLM token sampling

Name: Entropick – Plug quantum/hardware RNGs into LLM token sampling
Availability: InStock
Author: er777

by er777·Mar 9, 2026·2 points·0 comments

AI Analysis

●●SolidNiche GemWizardry

Swap software PRNG for hardware entropy in vLLM sampling, but niche use case with steep setup cost.

Strengths

•gRPC-first architecture cleanly separates entropy backend from inference stack; trivial to swap /dev/urandom for QCICADA or custom sources.
•Well-structured deployment profiles (urandom, OpenEntropy, template) lower onboarding friction; clear decision tree in README.
•Direct vLLM integration respects existing inference pipelines; no model retraining or inference-time overhead changes.

Weaknesses

•Extreme niche: QCICADA QRNG hardware is specialized lab equipment; practical audience is single-digit researchers.
•Unclear whether hardware entropy actually improves downstream LLM properties; benchmark vs. standard sampling missing.