Tottem: A Next.js App for Bookmark Management

Tottem is an open source experimentation, it aims combining personal productivity tool approach with (slow) social media capabilities to make users empowered and somehow emancipated.


Library management made social

I have two considerations in mind:

  • building a product based on ethic design
  • experimenting open-source web technologies and share it


First goal: experimenting the tools that are available to build web software in 2020. This documentation explains which technologies are used here and how they are organised.


Main technologies

It's a full-stack Typescript app with some code generation in order to have a type safe experience from end-to-end.

Here is a list of main technologies used:

Repository structure โ€” front-end

Inspired by those recommendations, this is how the codebase is structured:

โ”œโ”€โ”€ api # contains graphlq endpoint based on Apollo Server & Prisma2
โ”œโ”€โ”€โ”€โ”€ prisma # contains model definitions & database migration logbook
โ”œโ”€โ”€โ”€โ”€ src/schema # contains graphql resolvers using Prisma Client
โ”œโ”€โ”€ src
โ”œโ”€โ”€โ”€โ”€ generated # contains generated code (types, hooks, ...)
โ”œโ”€โ”€โ”€โ”€ pages # static and dynamic routes declaration used by NextJS
โ”œโ”€โ”€โ”€โ”€ components # shared generic component
โ”œโ”€โ”€โ”€โ”€ scenes # different parts of the application
โ”œโ”€โ”€โ”€โ”€โ”€โ”€ moduleName # Auth | Profile | Me ...
โ”œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€ components # module components. **Each** components can specify its own specific components, queries, ...
โ”œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€ queries.gql # All data queries and mutations are written in gql files
โ”œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€ hooks.ts # Most of the reusable logic is written in hooks
โ”œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€ index.tsx # Main scene file
โ”œโ”€โ”€โ”€โ”€โ”€โ”€โ”€โ”€ View.tsx # Sometime stateless component are isolated in View file for clarity or reusability
โ””โ”€โ”€โ”€โ”€ services # shared services as authentication, error management, ...

How is it typesafe from end-to-end?

  • Prisma provides a library Photon that generate a typesafe client to manipulate data depending on a unique schema declaration in schema.prisma file.

Example where Photon is used to retrieve the not softly deleted items from specific collection :

const items = (
    await ctx.photon.items.findMany({
        where: {
            collection: { id: collectionId },
            isDeleted: false,
        select: { id: true, position: true },
        orderBy: { createdAt: 'desc' },

Nexus provides a code-first graphql approach that allows you to generate graphql schema (schema.graphql file) based on your resolvers and object definitions. Nexus is fully compliant with prisma and offers a nice plugin to automatically declare resolvers based on your photon client.

graphql-codegen tool is configured to parse all .gql front-end files containing graphl queries and mutations. graphql-codegen uses remote and local (localSchema.gql) graphql schemas to generate every type and hook we need (putting them inside types.ts) so we can safely fetch and mutate data.

Note that when global state management is required, Apollo Client is used as much as possible.

Example where typesafe hook useGetItemsQuery is generated allowing to fetch data via Apollo Client smoothly

const { data } = useGetItemsQuery({
        variables: {

๐Ÿคฏ No typo anymore, much less files & context switching with typescript โœจ

Global state management

Most of the time, global state approach is used to avoid props drilling. Neither Redux or React Context API are used here. It has been implemented with Apollo Client. A local schema is defined in localSchema.gql

type Breadcrumb {
    title: String!
    href: String!

extend type Query {
    breadcrumbs: [Breadcrumb!]!

Then, we can define

  1. Query to read data from Apollo cache
query getBreadcrumbs {
    breadcrumbs @client {

codegraphql-code is configured to generate this simple hook query, used in Profile/TopBar for instance:

const { data } = useGetBreadcrumbsQuery()

2 Custom hooks to write data to cache with Apollo client

const useBreadcrumbs = (profileSlug: string) => {
    const client = useApolloClient()

    const setBreadcrumbs = ({ breadcrumbs }: GetCollectionProfileQuery) => {
            data: {

    const resetBreadcrumbs = () => {
            data: {
                breadcrumbs: [],

    return { resetBreadcrumbs, setBreadcrumbs }

SSR Workflow

NextJS provides SSR features that make user experience awesome. NextJS comes with the concept of pre-rendering built-in, that can take 2 forms:

  • Static Generation
  • Server-side rendering

When developing an app, pages are usually not static and need to be rendered on-demand depending on the context (ie. user). NextJS documentation is great. However, it can be a bit confusing and hard to deeply understand what's happening when we use NextJS / SSR. This is a sequence diagram aims explaining how NextJS SSR works in Tottem case:



The app is fully deployed on Zeit Now. Configuration can be found in now.json file. Merging on master trigger new deployment.

The API runtime is serverless and run via Now Serverless Functions. As described here, API performance with Prisma Client is pretty good even with coldstart.


Locally PG instance is needed with some var env set. In .env for instance

AUTH0_LOCAL_ID='auth0|5dc8800986c8ba0e74d73654' # set local user id by passing auth0

Then, two repositories are needed

git clone
cd tottem
npm install
npm run dev
cd tottem
cd api
npm install
npm run dev

Web app is available on http://localhost:3000 and graphql endpoint on http://localhost:4000/graphql


Second goal: designing a product human centered allowing people to build and manage their online and public library. A tool to manage and gather the content we love, in order to share it with friends & community ๐Ÿ˜‡


Why ?

More and more, people โ€” especially journalists, are losing their ability to choose which content to promote & amplify. Instead, automatic recommendation algorithms carefully choose the best (sic!) content to amplify. As reader, those deep learning algorithms create a unique & personal narrative stream of content in your social feed... How it can be the best ? Not really, it's just designed to maximize clicks and views. Of course, what else they can do?

Here, I'm happy to share some references

So, what if I want to explore durable book or article recommendations from a friend? What if I want to really dig into a specific subject?

How ?

Tottem aims combining personal productivity tool approach with (slow) social media capabilities to make users empowered and somehow emancipated.

Tottem aims to provide the same high quality user experience that most of modern productivity tools provide. Managing your library should be easy and enjoyable. With a great tool, great content could be made and shared.

The basic workflow:

  1. Collect everything in one Inbox.
  2. Organise into Spaces and Collection.
  3. Express yourself and explain your opinion
  4. Publish and share with your community



The UI/UX design is obviously inspired by Notion and Things 3. The design work made with Figma is available there



Download details:

Author: poulainv

License:  GPL-3.0 license


Tottem: A Next.js App for Bookmark Management
1.50 GEEK