iii

iii is a software engineering framework designed to simplify the process of creating, extending, and observing distributed systems by eliminating the complexity typically associated with integrating multiple services. It introduces a model where new capabilities can be added as easily as importing a library, with integrations shipped in a manner similar to how packages are managed in Node.js. The framework is intended for developers who need to assemble and orchestrate diverse categories of services while minimizing the overhead of learning, configuring, deploying, and monitoring each new system.

A central feature of iii is its use of workers, which can be implemented in multiple programming languages including Node.js, Rust, Python, and Go. Developers can register functions in these languages, and once registered, these functions become discoverable and invokable from a unified system regardless of their runtime. The system maintains a registry of available workers and functions, automatically discovering new workers as they connect. Functions are organized into categories such as agent, tools, memory, content, state, durable, billing, and notify, supporting a wide range of tasks from data transformation and machine learning inference to billing and notification operations.

The developer console in iii provides visibility into the state and activity of the system, offering insights into workers, functions, triggers, states, streams, queues, traces, logs, and configuration. The platform supports various trigger mechanisms, including HTTP requests, events, and cron schedules, allowing functions to be invoked in response to different types of system activity. End-to-end observability is emphasized, enabling both developers and agents to monitor system context and activity in real time.

iii employs a unified protocol to facilitate interoperability between workers and agents, treating agents as workers that operate within the same protocol and tracing system. This approach allows for consistent orchestration and monitoring across all components, regardless of their underlying implementation language or task category.