Virtual deployment framework

A better way to build, deploy and use models

  1. 01 - Build trac models

    Use the trac model runtime in your IDE and build portable Python models that can be executed in any environment without modification.

  2. 02 - Deploy to trac

    When importing models, trac scans the code and creates a schema representation of the model. The code stays in the repository.

  3. 03 - Build flows

    A flow is the blueprint of a calculation (a graph) made up of multiple models. They’re built on the platform from model schemas.

  4. 04 - Runtime deployment

    When launching a calculation on - whether to run a flow or a single model - trac fetches all model code from the repository at runtime.

  5. 05 - Re-use model code

    The same versioned code can be used in other environments in parallel - e.g. notebook-style solutions for exploratory workflows.

Take a tour

See how the virtual deployment method works in practice.

The benefits of virtual deployment

Models are ‘promoted’ to production in a few minutes, with zero risk, zero code modification and no platform-level intervention.

Deploy in minutes!

Using the native capabilities of the repository, trac retains a complete and immutable version history of every model in use on the platform.

Complete auditability!

There's no single primary deployment, so experimental model runs can be configured and executed without affecting your ‘production’ processes.

Total flexibility, total control!

The platform remembers which model version was used in each job. Because that version is always available in the repository, the job can be recreated.

Re-run with just a few clicks!

Download the free desktop version and start exploring today.

Try Community Desktop

Portable models

The trac model runtime

We understand the pain involved in deploying and maintaining models, which is why we built the trac model runtime — a lightweight, open-source package for building portable, production-strength Python models that can be executed anywhere.

Explore model runtime

Self-descrbing models

Model schemas and the trac model API

Models declare their schema’s (inputs, outputs, parameters) and trac guarantees the model always receives appropriate inputs wherever it runs - taking responsibility for data marshalling out of the model code so developers can focus on building better models.

Build trac models