How to set up filtering by tags?

[nktnet1]

I tried the following:

code

export async function GET(req: NextRequest) {
  const session = await getAuthSession();
  const pages = session
    ? source.getPages()
    : source.getPages().filter((page) => publicContentPaths.includes(page.url));
  const searchAPI = createSearchAPI('advanced', {
    indexes: pages.map((page) => ({
      id: page.url,
      title: page.data.title,
      description: page.data.description,
      structuredData: page.data.structuredData,
      tag: page.data.tag,
      url: page.url,
    })),
    // tag: true,
  });
  return searchAPI.GET(req);
}

After consulting the docs here:

• https://fumadocs.vercel.app/docs/headless/search/flexsearch#tag-filter

What am I missing? Thanks!

---

Current versions:

    "fumadocs-core": "13.4.7",
    "fumadocs-mdx": "10.0.1",
    "fumadocs-ui": "13.4.7",

[fuma-nama]

the data.tag is a custom property, you can pass anything to it, including a custom field on your frontmatter schema.

[nktnet1]

I'm trying to implement versioned docs similar to what you have here:

• https://fumadocs.vercel.app/docs/

I've been looking at the source code:

• https://github.com/fuma-nama/fumadocs/tree/dev/apps/docs

But can't seem to figure out where or how you're restricting the searches.

Can you show me how I can extend the frontmatter schema?

Basically, I have the following structure:

$ tree content/periods -L 1
content/periods
├── 24T3
└── 25T1

and I want to configure tags similar to Core/UI/MDX, where users can only search in one of those directories at a time.

Also, would I need to specify the tag in each .md and .mdx file, or is there a way to match it based those files using their parent directory as the tag?

Thank you!

[fuma-nama]

would I need to specify the tag in each .md and .mdx file, or is there a way to match it based those files using their parent directory as the tag?

just choose the way you want, page has its file path and frontmatter data.

to extend the frontmatter, see schema

[nktnet1]

Ah... I think I understand now. I thought tag was similar to title and description which was a metadata property of .md or .mdx file, but we could also do things like:

      tag: page.file.flattenedPath.substring(0, page.file.flattenedPath.indexOf('/')),

I couldn't find out where flattenedPath was documented in the docs, but was able to find it when logging things out. I think this solves my problem!

Thanks @fuma-nama ❤️!

[fuma-nama]

Also be careful, your code creates a new instance of search server for every request, you should only create two instances: one for authorised users, one for unauthorised

[nktnet1]

Thanks @fuma-nama!

I couldn't figure it out before when I was new to NextJS and fumadocs so went with this temporarily - also wasn't sure how to import "Page", but I guess I could just do:

ReturnType<typeof source.getPages>

I've now fixed it:

import { source } from '@/app/source';
import { createSearchAPI } from 'fumadocs-core/search/server';
import { NextRequest } from 'next/server';
import { publicContentPaths } from '@/app/(periods)/constants';
import { getAuthSession } from '../auth/[...nextauth]/auth';

const createSearchInstance = (pages: ReturnType<typeof source.getPages>) => {
  return createSearchAPI('advanced', {
    indexes: pages.map((page) => ({
      id: page.url,
      title: page.data.title,
      description: page.data.description,
      structuredData: page.data.structuredData,
      tag: page.file.flattenedPath.substring(0, page.file.flattenedPath.indexOf('/')),
      url: page.url,
    })),
    tag: true,
  });
};

const allPages = source.getPages();
const authorizedSearchAPI = createSearchInstance(allPages);
const unauthorizedSearchAPI = createSearchInstance(
  allPages.filter((page) => publicContentPaths.includes(page.url)),
);

export async function GET(req: NextRequest) {
  const session = await getAuthSession();
  const searchAPI = session ? authorizedSearchAPI : unauthorizedSearchAPI;
  return searchAPI.GET(req);
}

Thanks for pointing this out - yep, I definitely do not want to be doing that on every search 😹!

[fuma-nama]

btw InferPageType is what you need. I'll document it later