Boomi Architect key terms

The core Boomi build artifact: a visual sequence of shapes that moves data between sources and targets, applies logic, and handles errors. It is deployed to a runtime and triggered by a schedule, event, or API call.

A single step inside a process, such as a connector call, map, decision, branch, or try/catch block. Shapes pass documents along and define the order of execution.

The unit of data that flows between shapes in a process. It carries a payload plus properties and dynamic process properties, and can appear as a flat structure or a nested hierarchy.

A transformation shape that defines how source profiles map to destination profiles, including functions for concatenation, lookups, and conditional logic. It is the primary tool for converting between data formats.

A reusable definition of a data structure, such as an XML schema, JSON layout, flat file layout, or database table. Profiles give maps and connectors a shared understanding of fields and types.

A reusable operation that reads from or writes to a specific system, such as a database, file share, or SaaS application. Connectors are configured once and reused across processes.

A configured, named link to an external system that holds endpoint details and credentials. Connections are referenced by operations and are typically shared across many processes.

A specific action on a connection, such as GET, query, or upsert, along with its request and response profiles. Operations separate what action is performed from where the system lives.

A lightweight Boomi runtime that executes deployed processes on a server or in the cloud. It provides the execution engine, local storage, and connectivity to the platform.

A clustered runtime made of at least two nodes that share work and provide high availability. If one node fails, the remaining nodes continue processing, unlike a single Atom.

A Boomi-managed, multi-tenant runtime environment that hosts customer processes without requiring on-premises infrastructure. It offers shared or dedicated options with different isolation and scaling characteristics.

The act of packaging a process, its dependencies, and configuration and releasing it to one or more runtimes. Deployments can be versioned and rolled back.

A versioned bundle containing a process and all its dependent components, such as maps, profiles, and connections. Packaging makes deployments reproducible and traceable.

A configurable value attached to a process that can be set per environment or per deployment without editing the process itself. It is commonly used for endpoints, credentials references, and feature flags.

A property that exists only for the lifetime of a single process execution and can be read or written by shapes at runtime. It is useful for passing context such as record counts or correlation IDs.

A temporary in-memory or on-disk store where a process can put documents and retrieve them later in the same execution. It supports patterns like batching, sorting, and multi-pass processing.

A shape that wraps a group of steps and routes failures to a separate error path instead of stopping the process. It enables graceful handling, retries, and cleanup logic.

A shape that splits execution into parallel paths or routes to one path based on conditions. It is used for conditional logic and for processing independent streams concurrently.

A Boomi component that defines API resources, methods, and operations for exposing or consuming RESTful services. It can be deployed to a gateway and combined with authentication and policies.

A runtime layer that receives API requests, enforces policies, and forwards traffic to backend processes or services. It centralizes authentication, throttling, and routing for published APIs.

A rule applied at the gateway such as authentication, rate limiting, or payload validation. Policies are configured declaratively and enforced before the request reaches the backend.

A single request-and-response definition within a REST API component, binding an HTTP method and path to a process or backend action. Operations define the contract the API exposes.

A standard, machine-readable description of an API's paths, operations, parameters, and schemas. Boomi can import it to generate API components and export it to document published APIs.

An authorization framework where a client obtains a time-limited access token from an authorization server and presents it to the API. It is commonly used for delegated and machine-to-machine access.

A gateway policy that caps how many requests a client may make in a time window. It protects backends from overload and enforces fair use across consumers.

A profile describing delimited or fixed-width text data, including field positions, delimiters, and headers. It lets processes parse and produce files such as CSV without custom code.

A profile describing the structure of a JSON payload, including nested objects, arrays, and data types. It is used by maps and connectors to validate and transform JSON data.

A built-in or custom transformation used inside a map, such as string manipulation, date formatting, or lookup. Functions let maps perform complex value-level logic declaratively.

A reference table used by maps or processes to translate one value into another, such as a code to a description. Lookups can be stored as components and reused across maps.

A shape that runs a separate process as a sub-process, either synchronously or asynchronously. It promotes reuse by letting a parent process delegate a well-defined unit of work.

The practice of standardizing and correcting data during integration, such as trimming whitespace, normalizing dates, and removing duplicates. It improves downstream data quality and consistency.

The discipline of maintaining a single, authoritative version of core business entities such as customers or products. Integration processes synchronize these records across systems to reduce conflicting copies.

An AI-assisted capability that proposes mappings between source and destination profiles based on field names and data patterns. Users review and adjust suggestions before accepting them into a map.

A model trained on large text corpora that generates or interprets natural language and structured content. In integration, LLMs can classify, summarize, or extract data from unstructured inputs.

The instruction and context sent to a language model to produce a response. Well-designed prompts specify the task, the expected output format, and constraints on the answer.

A store that holds numeric embeddings of text or other content and supports similarity search. It enables retrieval of relevant context to ground model responses.

A numeric vector representation of content such as text, capturing semantic meaning so that similar items are close together. Embeddings are what vector databases index and search.

A pattern where relevant documents are retrieved from a store and inserted into a model prompt so answers are grounded in supplied content. It reduces hallucination and lets responses reflect private data.

A software component that uses a model to decide which actions or tools to invoke toward a goal, then observes results and iterates. Agents can call APIs, query data, and hand off to processes.

A mechanism where a model requests that an external function or API be executed with specific arguments, and the result is returned to the model. It lets agents act on systems rather than only generate text.

An open protocol that standardizes how AI applications discover and invoke external tools, data sources, and prompts. It gives agents a consistent interface instead of bespoke integrations per tool.

A constraint that limits what an AI system may input, output, or do, such as content filters, allowlists, or approval steps. Guardrails reduce harmful or unauthorized agent behavior in production.

A design where a person reviews or approves an AI-generated decision or action before it takes effect. It is used where errors carry high cost or regulatory risk.