MCP Relearn

This project provides a Python-based search functionality for static websites generated by Hugo using the Relearn theme. It enables AI-assisted search over documentation content, supporting integration with AI clients like Jan and local Ollama instances.

The library implements search engines using different backends:

• Lunr: Full-text search using the lunr.js algorithm

• RapidFuzz: Fuzzy string matching for approximate results

Scope

This project is designed as an MCP (Model Context Protocol) server that:

• Fetches and indexes search data from Hugo/Relearn-generated sites

• Provides semantic search capabilities over documentation content

• Supports multiple search backends for different use cases

• Integrates with AI chat clients for natural language queries

Related MCP server: linked-docs

Prerequisites

• Any MCP client able to provide tools to LLMs as MCP servers.

• Python 3.14 runtime environment.

• A static website generated by Hugo and using the fabulous Relearn theme.

Licensing

See LICENSE

How to actually use this

Using "Jan" and a remote Ollama instance

Instructions on how to set up a lean working environment using Jan as an AI Chat client and MCP host together with an Ollama instance – preferably self-hosted – can be found in README-jan.md.