https://linear.app/seedhypermedia/issue/SHM-2056/project-funnel-analytics Our funnel identifies a user and measures the conversion rate from one step to the next step. We need to identify the pain points at each step to boost conversion rates. A Growth Funnel is a bit different from a UX Funnel. The Growth Funnel should be a single instance across the system. It is core to the strategy and to every member of the team. Once we discover our Growth Funnel, it seldom changes. UX Funnels are a design tool. They can change all the time with different experiments. Growth Funnel UX funnels

Outcome: New Users quickly understand the Basics of Seed Hypermedia Sketching Final Design Tasks

This is a small and focused project that is aimed to solve the double-subscribe button. If you want to see the roadmap for subscriptions, look at  We are ready to do this project now. It is a pretty simple frontend change. Unlike the grand plan for subscriptions, this project does not require a new identity system. Problem Solution Implementation Notes

Problem 01 Solution

This project explains the problems with our current list implementation and proposes additions to editor schema that should handle lists better.

https://linear.app/seedhypermedia/issue/SHM-1622/i-cant-paste-images-on-web-mobile-comment-box

https://linear.app/seedhypermedia/issue/SHM-2031/project-task-create-document

https://linear.app/seedhypermedia/issue/SHM-2020/project-navigation-infrastructure

https://linear.app/seedhypermedia/issue/SHM-2016/editor-test-coverage

We are still shaping the design methodology. Please be flexible and constructive. Currently, we have two Active Design Projects. The other one is . It is hard to have two active projects at the same time, but we want to unblock  with this one so he can start working on the cryptography and backend. A Design Project is different from a Development Project, which we have been using all these years. A Design Project involves deep design work, every team member is expected to participate, and it is usually composed of more than one Development Project. Our Design Process follows Google's Design Sprint Methodology. The goal is to provide a mix of methods for meetings and async work. We can use different techniques depending on the Design Project. However, we should always be aware of the 6 phases. It is very probable that every time we skip a phase, we will end up having to come back and redo it, because the outcome won't work. Below are the 6 phases applied to the current Identity System Design Project. Each document should be a method output. Let's see how we operate async! Everybody is expected to: Phase 1: Understanding Phase 2: Define Phase 3: Sketch Phase 4: Decide Phase 5: Prototype Phase 6: Validation Project Tasks

Problem Scope Solution Rabbit Holes

Problem Solution Scope Rabbit holes No goes Prior art Search Example: { // ...search result data } move document examples

This document tries to enumerate blobs that need to be synced in order to render a proper document

Problem Solution Conversations on Document Discussion and Activity Panels

Problem Solution

test

Problem Solution export const loader = dataLoader(async (load, request) => { // this server loader is entirely optional! if you skip this, everything will be loaded with AJAX requests like we do in the desktop app const doc = await load(request.params.profileId) await loadEmbeds(load, doc) }) export default const webPageContainer(params => { const { docId } = params return <DocPage id={docId} /> }) function DocPage({id}) { // shared with desktop const doc = useResource(id) return <DocumentContent doc={doc} /> } // deep within document content, it renders this component function ContentEmbed({id}) { // shared with desktop const doc = useResource(id) return <DocumentContent doc={doc} /> } In this example, the webPageContainer wraps the whole page with a client provider that has preloaded all the embed content. This means that the deep useResource in the <ContentEmbed> has access to the data that was preloaded by the server. Scope Rabbit Holes No Goes

Releasing by pushing a tag is not the best way to trigger app releases

User Stories covered by this project Problem

Linking to specific parts of comments is very important and we as users are starting to feel the need of doing it.

Notes Description The Profile button on the Mobile Navigation Menu is terrible!

Problem Solution

Problem Solution

This user story requires the user to perform a trillion clicks! They should be just two clicks! one in each window!

Create and share documents only among a selected group of accounts so that they can review and collaborate.

Problem Solution Proposed approaches Concerns and questions Notion floating toolbar Blocknote floating toolbar Notion and blocknote both use floating toolbars!

Status: Completed. This project is considered completed when we enabled signing comments on any Hypermedia web site. Now instead of going via the redirect flow to hyper.media users are able to create a key right on the web site they are looking at, and sign their comments with that key. Linking this key to your desktop app and/or other keys you might have will be covered by a separate project. Problem Solution Rabbit Holes No Gos

This document looks to answer to the following user story> If we want a way for users to be able to see all the activity for the documents they have subscribed to as well as the communities they belong to, the best thing we can do is adopt the feature of a home feed/ for "For You Page" that many social media platforms like: X, Medium, Substack amongst many other use to keep their users up to date with general information of what's happening. The feed needs to incorporate both information for the node as well as information form all subscribed sites. To start it would be ideal to have document related information (document creation and document updates) as well as information on discussions (new discussions and replies). Proposed design look The look for the design was inspired form Substack which uses a combination of document posts in the feed with discussion posts of a sort. Take in mind that this was adjusted to takin into account the needs and stories relevant to Seed users. Substack Replies Threading and Commenting on The Feed Clicking a thread will also open up the threaded discussion similar to how X does it. It will open the thread depending on where you want to reply.

This document looks to answer the following user story: Users currently viewing a document's feed can only see activity specific to that document. The only way for them to see activity from the entire site is to go to the home document, which may not contain content they are interested in. They simply want to see the entire site's activity. The best solution for this is for the feed to be able to display both the document's feed and the entire site's feed. Library as a Document Directory and Activity Hub

User Story: . Refactor By subscribing to a site, a user will get notified of the latest activity on that site. Readers: Commenter: Owner/Author

How to give access to statistics to site owners depending if they self host or not.

This Is related to the story Find what is relevant. The immediate application would be to search for relevant content given some keywords. Documents relating to those keywords should pop up even if those words were never written literally. Also, searching in different languages. Searching for terms or concepts in a language, would yield results in another language. Other applications, and taking advantage of LLM models (out of the scope of this project) to generate the keywords themselves (reading the current document and coming up with relevant keywords) Then the semantic search would find similar documents/fragments. The LLM would be exposed via an Openai API compatible way so that the frontend could use it however it wants. The backend would implement the MCP tools and the model itself. Data Embedding First we have to vectorize the texts using an embedding model. These models translate text to a numeric representation so that we can apply a distance algorithm later given a query (also needs to be vectorized). This vectorizing process is multilingual so we can support major languages and the resulting numeric data would be the same independently of the language used. This is one of the benefits over traditional search, where no matter what languages you use in the query, you will find results with similar ideas no matter the language. The golang go-llama-cpp binding does not support new architectures (like bert) since it is based on old versions ob llama.cpp and its not currently maintained. The ollama engine is a separate process that needs to run in parallel. So the only real option to ship an embedding model with the daemon is to use the onnx runtime for golang We should load a dynamic library (.so in linux and Mac, dll in windows) and then load the models. A good enough model is multilingual-e5-small, especially in its qint8 version. Its just 118 MB and the tokenizer about 17MB. As long as we ship those three files along with the daemon (library, model, tokenizer) we are good to go. The process of vectorization is slow (even though we can do it in batches) so it will slow down the reindex process. However, as the search text is already stored in the db, we can populate the embeddings table on the background once the reindex process has finished. On a live index (for example when we create a new change) the embedding overhead is minimal. Vector search

This is a WIP document. It's related to the user story of . Problem Solution Rabbit Holes

We want to bring back Feeds. Problem Solution API Support

users want a way to show a list of discussions in the document content

TBD 

Problem Solution

Objective Import content from WordPress into Seed Hypermedia, turning posts/pages into documents in a site? Import workflow Possible libraries Edge cases Questions Should I just try calling API with /wp-json/? Apparently API key is only needed for private content or authenticated actions. Each post includes:    is my understanding of the project correct?

Solution to make recent relevant matches appear first

Invite link generation When clicking on invite link you will get a generated link with the selected role. If you want to change duration for the link, number of people that can use it or role you will be able to do this on the invite settings. Discord invites where taken as a reference Invite Membership

The commentor role is someone who is allowed to view and write comments even if this permissions have been disabled in the document. Joining Experience For Users Document Access Panel

Problem Solution Design Idea

Problem Solution Rabbit Holes / Concerns

Problem Solution

Problem Solution I believe we can use a library like DownShift to create the primitives for all the search instances we need. Scope Rabbit Holes No-Gos

We need to be able to copy comment URLs, open comment URLs on web and desktop, and we should be able to embed documents Owner: 

Now that we have comment IDs we can include back comments in search results

A contact is a relation between two accounts so for each account i am logged in, I pick a name (not the publishable name) of the other account. But right now, we don't search for contacts. We need to change the search API to take the account we are logged as and the backend should find the contacts matching the search input. Questions. What's a contact in the permanent data?

   

Problem Solution Behavior Downsides Scope

Problem Solution In the original window where the comment draft exists, there is an iframe which has loaded the identity provider hyper.media/hm/embed/auth. Internally it is polling the indexeddb for the containing origin's abilities. When they change, they are and communicated via postmessage so the publisher knows what sort of content can be signed in the iframe. When the comment is submitted, air.com passes in the unsigned comment blob to the identity provider iframe (again, via postmessage). The identity provider iframe will make sure the ability is present and valid. Then will sign it. It will pass the signed comment back to the air.com who requested it. At this point the signed comment needs to be saved to at least one server, and it can be sent to air.com or hyper.media, or both, whoever is willing to distribute it. For now, air.com will be the one responsible for saving and distributing the comment blob, which will result in a good UX because the comments will be refetched after saving to show the user the comment they just created. Why this is Secure Concerns

Problem Solution Alternative: Seed Phrase with Local Storage Cross-Domain Strategy Implementation Strategy

Problem Solution Scope Rabbit Holes

Problem Solution Scope

Problem Solution Scope Rabbit Holes No Goes

Problem Solution Scope Rabbit Holes No Goes