Processes

Processes are the atomic building blocks of Colibri. Each process is a single-purpose unit built on the convee library with:

• Clear inputs and outputs — Typed interfaces for predictable behavior

• Standardized errors — Every failure is wrapped in a typed ColibriError with diagnostics

• Plugin extensibility — Extend behavior via plugins that act on inputs, outputs, errors, or combinations

In Colibri, all processes are prefixed with P_ (e.g., P_BuildTransaction).

Process Structure

Each process follows the same pattern:

import { P_BuildTransaction } from "@colibri/core";

const process = P_BuildTransaction();

// Run standalone
const result = await process.run(input);

// Or add plugins
process.addPlugin(handleErrorGraciouslyPlugin);

Plugins can hook into:

• processInput — Transform or validate input before execution

• processOutput — Transform or enrich output after execution

• processError — Handle or transform errors

• Combinations of the above

When to Use Processes Directly

For most use cases, use Pipelines which compose these processes automatically. Use processes directly when you need:

• Fine-grained control over individual steps

• Custom process ordering

• To build your own pipelines

• To run a single step in isolation