components of a Krixik pipeline

Components of a Krixik Pipeline

Krixik pipelines are composed of one or more sequentially connected modules. These modules are containers for a range of parameterizable AI models or support functions.

Let's examine each of the key terms in the above sentence.

A pipeline is a self-contained sequence of one or more modules that is consumed via a serverless API.

A module is a processing step with a unique input/output data footprint. Each module contains a parameterizable AI model or support function.

A model is a bespoke processing function contained within a module. Many of these are AI models, but some are simpler "support functions" for inter-pipeline data preparation or transformation.

Parameters can be set for each module when a pipeline is run and allow for further customization. Each has a default value, so setting them is optional. For instance, one parameterizable item is which specific AI model you want active within a given module.

--

New modules and models will constantly be added to the Krixik library. To see all available modules at any given time, use the available_modules property:

krixik.available_modules

Each module has its own documentation that details, among other things, available models for it. For example, here's documentation for the transcribe module.