yemrealtanay

yemrealtanay / llm-hasher

Public

Privacy-first PII tokenization middleware for LLM pipelines — local detection, encrypted vault, zero cloud dependency.

ai-safety data-privacy gdpr go llm
11
0
100% credibility
Found Apr 06, 2026 at 11 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Go
AI Summary

llm-hasher is a privacy tool that automatically detects and replaces sensitive personal information in text with unique placeholders before sending to AI language models, then restores the originals from responses.

How It Works

1
🔍 Discover llm-hasher

You find this helpful tool while looking for ways to keep names, addresses, and card numbers private when using AI chatbots or analyzers.

2
🚀 Get it started

You download and launch it on your computer with a few easy steps, and it sets up a local helper that runs quietly in the background.

3
📝 Hide sensitive details

You paste in your message or document full of personal info, and the tool spots things like emails, phones, or addresses and swaps them for safe stand-ins.

4
🤖 Share safely with AI

You send the cleaned-up version to your favorite AI service for summarizing or analyzing, knowing real details never leave your computer.

5
🔄 Reveal the real info

When the AI replies with stand-ins, you run it back through the tool to swap everything back to the original private details.

🛡️ Privacy achieved

Your conversations and documents get smart insights from AI without ever exposing personal information to the outside world.

Sign up to see the full architecture

4 more

Sign Up Free

Star Growth

See how this repo grew from 11 to 11 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is llm-hasher?

llm-hasher is a privacy-first PII tokenization middleware for LLM pipelines, built in Go. It detects sensitive data like credit cards, emails, names, and addresses using local regex and Ollama models, then replaces them with opaque tokens stored in an encrypted SQLite vault before sending text to any LLM. You detokenize responses via simple HTTP APIs like POST /v1/tokenize and /v1/detokenize, with zero cloud dependency.

Why is it gaining traction?

It stands out with fully local detection and storage—no data leaks to third parties—plus hybrid speed (regex for structured PII, LLM for context like addresses). Docker Compose spins up Ollama and the service in one command, and it doubles as a Go library for direct integration. Features like context-scoped deduplication, TTL expiry, and batch tokenization keep LLM reasoning intact without plaintext PII exposure.

Who should use this?

Backend engineers building LLM-powered customer support bots, call transcript analyzers, or document processors handling real user data. Compliance-focused teams at fintechs or healthcare apps needing on-prem PII scrubbing before cloud LLMs. Go devs wanting a lightweight hasher in RAG or chat pipelines without vendor lock-in.

Verdict

Worth a spin for privacy-sensitive LLM workflows—solid docs, quick Docker start, and MIT license make evaluation easy. At 11 stars and 1.0% credibility, it's early-stage with room for broader PII coverage and production hardening, but viable for prototypes today.

(187 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.