Skip to content
View stef41's full-sized avatar
♦️
♦️

Block or report stef41

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
stef41/README.md

Hi, I'm Zacharie 👋

I build open-source AI/ML developer tools in Python — the missing utilities for anyone working with LLMs.

Everything is pip-installable, zero-dependency, works offline, and has comprehensive test suites.


🔥 Flagship Projects

🔍 lmscan — Open-source GPTZero alternative

PyPI Downloads Tests

Detect AI-generated text and fingerprint which LLM wrote it. 12 statistical features, 9 model signatures, multilingual support. <50ms, zero dependencies.

pip install lmscan
lmscan "paste any text here"
# → 82% AI probability, likely GPT-4 (confidence: high)

🎵 vibescore — Grade your vibe-coded project

PyPI Downloads Tests

One command, instant letter grade (A+ through F). Security, code quality, deps, testing. Python, JS/TS, Rust, Go.

pip install vibescore
vibescore .
# → Overall Grade: B+

🧰 AI/ML Developer Toolkit

📐 Data & Training

Package What it does Install
datacruxai Training data quality — dedup, PII detection, benchmark contamination pip install datacruxai
castwright Synthetic instruction data generation for fine-tuning pip install castwright
datamix Dataset mixing & curriculum optimization for LLM training pip install datamix
trainpulse Training health monitor — loss spikes, gradient explosions, plateaus pip install trainpulse

🔬 Models & Inference

Package What it does Install
infermark LLM inference benchmarking — TTFT, throughput, tail latencies pip install infermark
ckptkit Checkpoint Swiss Army knife — inspect, diff, merge SafeTensors/GGUF pip install ckptkit
modeldiffx Behavioral regression testing — diff model outputs across versions pip install modeldiffx
quantbenchx Quantization quality analyzer — layerwise sensitivity, mixed-precision pip install quantbenchx

🔤 Tokenization & Cost

Package What it does Install
tokonomix Token counting & cost management across all LLM providers pip install tokonomix
toksight Tokenizer analysis — vocab coverage, compression, fertility comparison pip install toksight

🛡️ Safety & Security

Package What it does Install
injectionguard Prompt injection detection for LLM apps (OWASP LLM Top 10) pip install injectionguard
vibesafex Catch bugs in AI-generated code — secrets, hallucinated imports, vulns pip install vibesafex

📊 By the Numbers

Metric Value
Packages on PyPI 14
Total tests 3,000+
Python versions 3.9 — 3.13
CI Green on every repo
External deps Zero (most packages)
License Apache 2.0 (all)

💡 Philosophy

  1. Ship real tools, not demos. Every project has a CLI, Python API, tests, CI, and is on PyPI.
  2. Zero deps where possible. No dependency hell. Pure Python stdlib.
  3. Offline first. Your code, your data, your machine. No API keys, no telemetry.
  4. Citable. Every repo has CITATION.cff for academic use.

⭐ If any of these tools are useful to you, a star helps others find them!

@stef41's activity is private