8 projects
ko-copycheck
Offline plagiarism checker for Korean prose submissions: verbatim winnowing plus paraphrase sentence similarity, static HTML report.
known-secret-guard
Egress check for secrets you already know: catches base64/URL/hex/HTML/split variants that pattern scanners miss.
litellm-lint
Static linter for LiteLLM proxy config.yaml: typo'd keys, missing env vars, dangling fallbacks and aliases. Unofficial.
toolcall-conformance
Check how an OpenAI-compatible server (Ollama, llama.cpp, vLLM, LM Studio, LiteLLM...) handles tool calls on the wire.
hwpdown
Pure-Python HWP 5.0 / HWPX reader that renders LLM-friendly Markdown (merged cells, footnotes, numbering) from the public file-format specs.
awsglance
Read-only, single-process dashboard for one AWS account: who changed what, risky settings, and what to check now.
embed-ab
A/B-test two embedding models on your own documents: Recall/nDCG/MRR with confidence intervals, per-query wins and losses, top-k overlap, and a CI gate.
cache-autopsy
Find out why your LLM prompt cache missed: offline diff, usage-aware hit rate, missed cost and CI gates for Anthropic and OpenAI request logs.
