Ultrahost/ docs
All examples

Example 01

Experiment

Searchable & Connected Data

Documents that can be found by keyword, by meaning, and by how they relate to each other.

Start with a plain document, then let it grow senses. It gains a shape, a way to be found by words or by meaning, links to its neighbors, and rules about who can touch it. Most knowledge-heavy products end up here.

Chapter one

When a document is saved

A single save quietly sets off a chain of small jobs, each handled by a different block.

  1. 1
    Authorization Layer

    Can this person store it?

    Before anything is written, the system checks that the author is allowed to create documents here.

  2. 2
    Document Database

    The document is stored

    The JSON document is saved and gets an id the rest of the system can point to.

  3. 3
    Schemas & Ontology Versioning

    It is checked for shape

    The document is compared against its declared schema, so malformed ones are caught early.

  4. 4
    Graph Database

    References become edges

    If it mentions other documents

    Each reference is stored as a link between documents, so relationships can be followed later.

  5. 5

    At the same time

    It is made findable

    Two copies of its meaning are kept, one for each way of looking.

    • Vector DB

      Stored for meaning

      An embedding is saved so similar documents can be found, even in different words.

    • Search Engine

      Indexed for keywords

      The text is indexed so exact words and phrases can be found quickly.

  6. 6
    Federated Auth & User Management

    The author is recorded

    Who wrote it, and who may edit it later, is remembered with the document.

Chapter two

When someone looks for it

The same document can now be reached from three directions.

Someone starts a search

A question, a few words, or a document they already have in hand.

  • By similarity

    Find what means the same

    Ask for documents that are close in meaning, even when no words match.

    Vector DB

  • By keywords

    Find what says the same

    Look up exact words, phrases, and filters, with typo tolerance and ranking.

    Search Engine

  • By neighbors

    Find what sits nearby

    Start from one document and follow its links to the ones around it.

    Graph Database

Make it yours

Variations worth trying

  • Drop the graph and ship a searchable library first.
  • Swap keyword search for similarity only, if people search by idea.
  • Add per-team visibility rules once more than one team shares the space.

Starting point

Ask your coding agent for your version

Use this prompt as a starting point, then change whatever does not fit your idea.

Build a document library where every document is searchable, related to other documents, and protected by permissions.

Use these Ultrahost building blocks, in this flow.

When a document is saved:
1. Authorization Layer: Before anything is written, the system checks that the author is allowed to create documents here.
2. Document Database: The JSON document is saved and gets an id the rest of the system can point to.
3. Schemas & Ontology Versioning: The document is compared against its declared schema, so malformed ones are caught early.
4. Graph Database (if it mentions other documents): Each reference is stored as a link between documents, so relationships can be followed later.
5. It is made findable, at the same time:
   - Vector DB: An embedding is saved so similar documents can be found, even in different words.
   - Search Engine: The text is indexed so exact words and phrases can be found quickly.
6. Federated Auth & User Management: Who wrote it, and who may edit it later, is remembered with the document.

When someone looks for it:
- By similarity (Vector DB): Ask for documents that are close in meaning, even when no words match.
- By keywords (Search Engine): Look up exact words, phrases, and filters, with typo tolerance and ranking.
- By neighbors (Graph Database): Start from one document and follow its links to the ones around it.

Treat this as a starting point. Keep the parts that fit my idea and adapt the rest.

Next example

Intelligent File Management