nilrag-man: Semantic Man Page Search with nilRAG
By Anton Pyrogovskyi • 3 minutes read •
Table of Contents
Ever found yourself struggling to remember the exact tar
flag you need? Or flipping through man
pages trying to find the right section for a lesser-known Unix command? Enter nilrag-man
, a Python-based proof of concept built with Nillion’s privacy-preserving stack that brings the power of semantic search to local manpage content — all while keeping your data private and secure.
What is nilrag-man
?
nilrag-man
is a script that uses:
- nilRAG – Nillion’s Retrieval-Augmented Generation (RAG) framework
- nilDB – A secure, decentralized database layer
- nilAI – Nillion’s AI module for generating natural language responses
Together, they enable you to search and query local man pages semantically, so instead of memorizing flags or exact keywords, you can just ask questions like a human.
How It Works
Here’s a quick overview of how nilrag-man
functions under the hood:
- Indexing: It parses and ingests your system’s local
man
pages, chunking and embedding the content intonilDB
vianilRAG
. - Private Storage: All the indexed data is stored securely using
nilDB
. - Semantic Search: When you ask a question (e.g., “how do I recursively change file permissions?”),
nilRAG
retrieves the most relevant chunks. - Answer Generation: Finally,
nilAI
synthesizes a helpful answer, combining manpage data with semantic context.
This approach shows how you can use RAG-based agents in a completely private, decentralized environment.
Example Usage
Output:
To copy directories recursively, use the -r or -R flag with cp:
cp -r source_directory destination_directory
This tells cp to copy directories and their contents.
No more scrolling through man cp
or trial-and-error guessing!
Setup instructions are in the repo.
Why Use nilDB/nilRAG?
Traditional RAG setups typically rely on third-party APIs and cloud-hosted vector databases. nilrag-man
flips the script by proving that RAG can be locally-driven, secure and private:
- Your queries and indexed content stay private
- You get granular control over storage, embedding, and querying
- It’s open to extension — plug in more docs, more tools, more local datasets
A Proof of Concept with Potential
While nilrag-man
is a prototype, it demonstrates something powerful: you can build practical AI assistants for your own data, without sacrificing privacy. Whether you’re a power user, developer, or just sick of re-Googling awk
flags, this could be the beginning of a new way to interact with your tools. It’s not just about making man
pages easier to use — it’s a glimpse into what’s possible when semantic search meets decentralized, privacy-respecting AI.