VECTORIO
EN
EnglishDeutsch
Beta · design partners wanted

One billion rows · 32 ms · one notebook

Big data. No big cloud.

PROTOR computes so fast that most data jobs no longer need a cluster — on your own hardware, offline if you like, and every result with proof.

livePROTOR · SQL
SELECT key, SUM(value)
FROM events GROUP BY key  -- 1,000,000,000 rows
32.1 msone notebook · CPU only · measuredHow we measured
Start a pilotSee the numbers
cloudcluster32 ms1,000,000,000 rowsone notebook

Where teams start. Pick yours.

Defence, dual use, telecom

Sovereign & edge

“No cloud. Contested links. And the result has to hold up.”

Critical infrastructure, public sector, security

Regulated evidence

“Twenty-four hours to report. The evidence is in five silos.”

Data teams, finance, commerce, software

Cost & reproducibility

“The warehouse bill grows faster than the data. And last quarter's number won't reproduce.”

IT, service providers, groups, administrations

Directories & communication

“Two forests, three address books — and nobody knows who changed what.”

Showcase: instruments, satellites, imaging, media

Research & geodata

“The instrument outgrew the workstation. The queue outgrew your patience.”

At the counter

Done before the coffee.

Order a flat white and start the nightly pipeline over 100 million events. It is done on your notebook before the milk is foamed.

  1. 0 s
    You order a flat white

    The nightly pipeline starts: 100 million music-streaming events, 9.6 GB of CSV.

  2. 4.8 s
    100 million events read

    Sessions and the track catalogue join in as the events stream through.

  3. 34 s
    The pipeline is done

    21 steps: personal data masked, duplicates removed, quality checked, enriched and rolled up per day. 122,222 faulty rows set aside with their reason.

  4. ≈ 90 s
    Your coffee arrives

    Time to read the daily rollup — and change the next step.

Measured: a 21-step pipeline over 100 million music-streaming events (9.6 GB CSV), 33.8 s end to end on one Apple M3 Max notebook with 64 GB; coffee time illustrative.

0 sdone · 34 scoffee · ≈ 90 s
In the field

Compute where the data is born. Send only results.

Sensors, cameras, drones and vehicles run PROTOR on the device. Over 5G, satellite or a local radio link — and 6G as the networks arrive — only small signed results travel. The analysis keeps going when the link drops.

river sensorjunction cameradronefield sensorsvehicle5G · 6G · satelliteedge hubPROTOR on the deviceonly results travel
Raspberry Pi-class boards (beta)4–7 peers in one cluster, demonstratedMQTT · RTSP cameras · REST feedscontinues after a dropoutevery result signed

Environmental protection

River-level and air-quality sensors raise a local alert the moment a threshold breaks — the station keeps measuring when the link drops.

Traffic control

Cameras at the junction count and classify on the device; faces and plates are blurred before anything is stored.

Law enforcement

Body-camera footage is redacted on the device and sealed as evidence that anyone can verify later.

Agriculture

Field sensors and drone imagery turn into a field map at the farm, without a fast uplink.

NGOs and disaster relief

Damage reports, sensor feeds and images are processed in the crisis area on a single rugged box.

Defence and dual use

Sensor streams are fused at the edge; the same workflow runs in the vehicle, the command post and the data centre.

Workflows

See it work. For your case.

Four real workflows from our test suite — drawn on the canvas, saved as a file, each ending in a signed result on your hardware.

PROTOR · Body-camera footage, redacted on the devicerunning
Video File InputVideo.File.SubscriberFace BlurVideo.FaceBlurEvidenceEvidence CollectorPackageEvidence PackOutputBinary File Writer
  1. Video File InputVideo.File.Subscriber
  2. Face BlurVideo.FaceBlur
  3. EvidenceEvidence Collector
  4. PackageEvidence Pack
  5. OutputBinary File Writer
nodes
bodycam-input → Video.File.Subscriber
face-redact → Video.FaceBlur
evidence-collect → Forensic.EvidenceCollector
evidence-pack → Forensic.EvidencePack
evidence-output → BinaryFile.Publisher
18.7 ms per frame · 30 fps · 0 % frame drops · measuredfaces blurred on the devicesealed evidence package

Real workflows and node types; the designer view is illustrated and evolves with the product. Measured values come from our end-to-end test runs.

32.1 ms1 billion rows grouped and summedsorted layout, one notebook
105–118 GB/saggregation throughputsorted · 45–80 GB/s unsorted
24.4 s103 million rows through a 21-stage workflowone notebook, end to end
10,000nodes in one real-time workflowBeethoven demo · 48 kHz · no underruns

Measured on one Apple M3 Max-class notebook, 16 cores, 64 GB, CPU only; 1 cold + 5 warm runs; logical decoded GB/s. Method and raw table →

Fewer tools. Fewer bills.

Most teams glue five products together to get one answer. PROTOR does the whole job in one runtime, with fewer licences and fewer moving parts.

What most teams run
  • Cloud warehouse
  • ETL / ingestion tool
  • Streaming platform
  • Orchestrator
  • Lineage & audit tool

5 vendors · 5 bills · 5 copies of your data

PROTOR
One runtimeBring in · clean · combine · check · deliver

On a notebook, a server, an edge box or your own cluster — with proof for every result.

What changes for you.

Answers you can defend

Every result comes with a record of how it was made. Hand it to an auditor, a customer or a regulator.

The same number, every time

Rerun last quarter's report and get last quarter's number — on any machine.

Big-data speed on a notebook

A billion rows in 32 ms. Many jobs run without a cluster — and without a cluster bill.

Your data stays home

From a small edge device to your own cluster, offline if you need it. Nothing leaves your building.

Local first

Your data. Your hardware. Your rules.

PROTOR runs where your data already is — on a notebook, a server, an edge box or a sealed network. That decides who can see the data, what happens when the line drops, and what you pay.

Confidential by default

Your data stays on hardware you own. Customer records, patient data and trade secrets go into the analysis and stay where they are.

Runs fully air-gapped

No connection needed — in a vehicle, a plant, a lab or a secure network. Same engine, same results.

Reliable without a network

No cloud outage, no rate limit, no queue for a cluster. When the machine runs, the analysis runs.

A licence you can plan

One licence for your own hardware. No per-query bill, no egress fee, no surprise at the end of the month.

From satellite to street. On one machine.

Munich's hottest streets, mapped from Sentinel-2 imagery: vegetation, surface temperature and heat — delivered as a signed map.

Munich Vegetation (NDVI), PROTOR pipeline output
Vegetation (NDVI)
Munich Surface temperature, PROTOR pipeline output
Surface temperature
Munich Surface albedo, PROTOR pipeline output
Surface albedo
Munich Urban heat index, PROTOR pipeline output
Urban heat index
All sample pipelines

Three products. One idea.

All three keep your data with you — and show where every result came from.

PROTOR · workflow canvas
PROTOR · workflow canvas
Data engine

PROTOR

ETL, SQL, streaming and workflows in one runtime — from a notebook to a cluster.

Meet PROTOR
Active DirectoryLDAPGoogle · CardDAVCSV · XML · ExcelEncrypted e-mailPREVIEWInsert124Update20Delete3Unchanged9,853Global addresslisttarget directory
Directory sync

Connectype

Address lists, users and groups in sync across directories and files — previewed, capped and audited. Built on PROTOR.

Meet Connectype
INDAGO Accelerate
INDAGO Accelerate
Search apps for Mac

INDAGO

Finds the exact page in your documents in milliseconds, on your Mac, with the source.

indago.one
Your first step

Start small. Keep what you have.

Each product starts next to the tools you already use. You see the result first and decide afterwards.

PROTOR

One job, next to your stack

Pick the nightly job with the biggest bill. Run it on a notebook beside your current platform and compare the numbers.

one job · one week
Connectype

Two directories, preview only

Point Connectype at a copy of two directories. Read the preview. Nothing is written until you agree.

one afternoon
INDAGO

One folder, one question

Add the folder you search most and ask the question you asked last week.

five minutes
Beta · design partners wanted

Bring us your hardest data problem. Let's solve it together.

A short pilot on your data and your hardware. We agree on what success looks like before we start.

Start a pilotSee the numbers