Ultrahost/ docs
All examples

Example 03

Experiment

Time-Series Analytics

Track any numeric metric over time, then share it, compare it, and keep it tidy.

Numbers that change are only useful with context. Keep each series next to a description of what it measures, relate series to each other, publish datasets and previews, and let a schedule take care of compression and cleanup.

Chapter one

When a new metric is born

Creating a metric is a small ritual. Once it exists, points can keep arriving for as long as it matters.

  1. 1
    Timeseries Database

    A time series is created

    The TS Store opens a new series with its own key, ready to receive points.

  2. 2
    Document Database

    Its key and meaning are kept

    A JSON document records the series key and its metadata, such as what it measures and in which unit.

  3. 3
    Graph Database

    Related series are linked

    If it belongs to a family of metrics

    Edges tie it to its siblings, such as different measurements describing the same event.

  4. 4
    Timeseries Database

    Points are added over time

    Each new reading is appended through the TS Store API, one point at a time or in batches.

  5. 5
    Authorization Layer

    Who can see it is decided

    Rules declare who may read the series, so sensitive numbers stay with the right people.

Chapter two

While nobody is looking

A schedule shrinks, reshapes, and packages the data, so it stays fast to query and easy to share.

A scheduled job wakes up

On a timer, with no one waiting for the result.

  • Compact

    Summarize and transform

    The Cron engine compresses older points and runs the ETL that reshapes the series.

    Cron

  • Share

    Package complete datasets

    Full datasets are stored as files, so they can be handed to others.

    File Server

  • Warehouse

    Keep the ETL results

    The File engine acts as a data warehouse for everything the ETL produces.

    File Server

  • Preview

    Save dashboard previews

    High-level views of the series are stored as files, so dashboards load instantly.

    File Server

Chapter three

When the data is put to work

Beyond charts, the shape of a series can feed models and reveal hidden similarities.

Someone wants to learn from it

An analyst, a model, or a curious question about how things repeat.

  • Train

    Feed neural networks

    Embeddings of the series are stored in the Vector engine, ready to train models.

    Vector DB

  • Compare

    Cluster similar segments

    Segments are grouped by shape, so repeating behavior shows up without being searched for.

    Vector DB

Make it yours

Variations worth trying

  • Start with one series and a dashboard preview, nothing else.
  • Add similarity only if you want to spot repeating shapes.
  • Replace file previews with live queries when freshness matters more than speed.

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 metrics tracker where each metric is a time series with metadata, related series, shareable datasets, restricted access, and scheduled upkeep.

Use these Ultrahost building blocks, in this flow.

When a new metric is born:
1. Timeseries Database: The TS Store opens a new series with its own key, ready to receive points.
2. Document Database: A JSON document records the series key and its metadata, such as what it measures and in which unit.
3. Graph Database (if it belongs to a family of metrics): Edges tie it to its siblings, such as different measurements describing the same event.
4. Timeseries Database: Each new reading is appended through the TS Store API, one point at a time or in batches.
5. Authorization Layer: Rules declare who may read the series, so sensitive numbers stay with the right people.

While nobody is looking:
- Compact (Cron): The Cron engine compresses older points and runs the ETL that reshapes the series.
- Share (File Server): Full datasets are stored as files, so they can be handed to others.
- Warehouse (File Server): The File engine acts as a data warehouse for everything the ETL produces.
- Preview (File Server): High-level views of the series are stored as files, so dashboards load instantly.

When the data is put to work:
- Train (Vector DB): Embeddings of the series are stored in the Vector engine, ready to train models.
- Compare (Vector DB): Segments are grouped by shape, so repeating behavior shows up without being searched for.

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

Next example

Knowledge Graph