Live streams

Recordings

Everything Kymoscope stores from a stream is a recording. Find them in Cloud, mark moments with MARK, and open them as datasets.

Everything Kymoscope stores from a live stream is a recording. A stream records by itself (always, or between a start and a stop condition) or when someone presses Record, as its source says. Every recording of the streams you can see is in the workspace’s Cloud window, grouped by project. The data travels and rests encrypted: your browser decrypts it.

How a stream records

The collector decides, per source, with record. The app only shows the result next to each stream: Always or Trigger for a stream its collector records (Automatic while the collector is older than 0.6.1), Manual for one recorded with Record.

record	What is stored
`always`	Everything the stream sends. The default on Team and Enterprise.
`trigger`	From a start condition until a stop condition, written like a cycle rule (Speed > 5, held 5 s). Each start to stop is one recording.
`manual`	Only while someone presses Record on the live stream. The default on other plans.

collector.yaml

sources:
  - name: Reactor 1 DCS
    opcua: { url: opc.tcp://dcs.customer.local:4840 }
    record: always

  - name: Car telemetry
    udp: { port: 20777 }
    record:
      trigger:
        start: Speed > 5
        stop: Speed < 1
        stop_after: 2min

recording:
  timezone: Europe/Madrid      # recordings are cut at its midnight (default UTC)
  split_gap: 30min             # no value for this long starts a new recording
  where: own                   # kymoscope | own: your organization's bucket

Every key is described in Collector: recording. Where the files are kept (Kymoscope storage or your own bucket) is in Historian: your own storage.

Find and open a recording

In the workspace, press Cloud in the left rail.
Under Projects, pick a stream. Type in the filter to find a project or stream by name.
Pick a range: Today, Last 7 days, Last 30 days, or Custom with From and To.
Coverage shows one row per day in the site’s time: each bar is a recording, each pin a mark. Click a bar, or tick rows under Recordings, to select them.
Press Open N recordings. Each opens as its own dataset, with absolute time on the X axis.

Open range opens the stream’s whole range as one dataset instead: summaries first, full detail as you zoom in, and new data as it arrives when the range ends now. It is off while recordings are selected, and for manual streams.

Selected, across projects

What you select stays in the Selected strip at the top of the window while you browse other streams and projects, so you can open the same session of two cars, or two shifts of two lines, together. × removes one; Clear removes them all. Once open, the dataset’s Cycles button cuts it into cycles.

Marks

Every collector stream has a channel mark. It is 0, and 1 for a single sample when someone marks the stream. A mark flags a moment; look before and after it when you analyse. It is stored and plotted like any channel, so you can chart it, alert on it, or start a cycle on it.

Watch the stream live in the workspace.
Press MARK at the right of the top bar. It marks the live stream on view.
The bar says Marked with the time. If the collector isn’t connected, it says nothing was marked.

Anyone who can watch the stream can mark it, viewer accounts included. In Cloud, marks show as pins on Coverage and as a count per recording, for manual recordings too: every MARK the collector received is kept with its time. A machine signal can mark too: set mark_tag on the source and a rise of that channel marks the stream (counted in Cloud for recordings its collector stores).

collector.yaml

sources:
  - name: Reactor 1 DCS
    opcua: { url: opc.tcp://dcs.customer.local:4840 }
    mark_tag: MarkButton        # a rise of this channel marks the stream

Keep, export, rename, trash

Action	What it does
Export	The recording as one file: CSV or MF4 (ASAM MDF 4), and Parquet for automatic recordings. Every channel, full resolution.
Keep	Automatic recordings: its full-resolution data is kept past the plan’s retention (up to 7 days per recording).
Rename	Gives the recording your own name. Clear the name to go back to the automatic one.
Trash	Manual recordings: moves it to the trash, where it can be restored for 30 days. Automatic recordings follow retention and Keep. In the trash (the Trash button at the bottom of Cloud), Delete forever removes one recording now, with its cameras, and Empty trash removes every one you may delete. Neither can be undone.

Automate it

What records, and how, lives in the collector’s collector.yaml: keep it in version control and roll it out like any configuration.
Marks from the process: mark_tag on a source (above).
Working with the data elsewhere: export a recording as Parquet and read it with pandas or DuckDB, see Open history in the app.