add-doc-integration

Add a New Documentation Site Integration

You are adding a new documentation site to the go-dev-mcp MCP server. This codebase already has three working integrations (godoc, rustdoc, pydoc) that all follow the same pattern. Your job is to replicate this pattern for the new site.

Overview

Each documentation site integration consists of exactly 3 MCP tools:

1. search_{name} - Search/discover documents (modules, packages, crates, etc.)
2. read_{name} - Read a specific document page with pagination
3. search_within_{name} - Keyword search within a specific document page

These tools are implemented across 4 layers + golden tests.

Step-by-step Implementation

Phase 1: Explore the target documentation site

Before writing any code, you MUST understand the HTML structure of the target site.

1. Identify 2 key page types:

   • Search/index page: Where users discover available documents (e.g., module index, package listing)
   • Document page: Individual documentation page (e.g., a module reference, API docs)

2. Fetch sample HTML and inspect the DOM structure. Look for:

   • Root container element (e.g., section#module-*, div.rustdoc, article)
   • Headings (h1-h4) that structure the page
   • Definition lists or similar patterns for API entries (functions, classes, methods)
   • Code blocks (pre, code elements)
   • Signature elements (function/method signatures)
   • Navigation/link structure in the search/index page

3. Verify dq compatibility. The pkg/dq package supports:

   • Tag matching: "div", "section", "dl"
   • Class matching: "dl.py" matches <dl class="py function"> (checks individual classes)
   • ID matching with globs: "section#module-*" uses filepath.Match
   • Attribute matching: "a[href]"
   • Comma groups: "h2,h3,h4"

   For complex deeply nested structures (e.g., ul > li > ol > li), the dq.NewRecursiveNodeMatcher() supports multi-step patterns with > separators and recursive restart. See pkg/dq/dq.go for details.