Studio terms in plain language
| Term | Definition | Mental model |
|---|---|---|
| Workflow | The path followed for each message, from start to an answer or action. | A recipe with ordered steps |
| Block | A step that displays a message, calls a service, prepares an instruction, or selects a route. | One step in the recipe |
| Canvas | The workspace where blocks are placed and connected. | A workbench |
| Variable | A value available during execution, such as the message or an API result. | A reusable sticky note |
| LLM | The AI model that understands requests and writes answers. | A writer who needs a good brief |
| Instruction (prompt) | The model’s role, tone, limits, data, and tools. | The writer’s brief |
| RAG | Searching documents before answering to ground the response in your content. | A librarian checking sources |
| MCP | A standard for exposing and calling external tools. | A toolbox connected to the assistant |
| Tool | A concrete action a model can request through MCP. | A specialized tool in the box |
| Condition | A block that selects the next path according to a rule. | A railway switch |
| Alias (output name) | A readable name for block outputs, such as documents instead of http-1. | An easy-to-remember nickname |
| Success/failure branch | Two planned next steps depending on the result. | The normal route and plan B |
| Helper | A function used in an expression to transform or test a value. | A small ready-to-use operation |
| Scope | How long and where a variable remains available. | The sticky note’s lifetime |
| Data source | A collection of documents or tabular data used to find information. | A reference collection |
| Embedding model | A model that represents content so it can be searched by similarity. | An index of meaning |
| Contextual port | A connection that gives a block sources or tools prepared before it runs. | A preparation connection |