Platform model
Saptiva Studio is the console from which a team builds, runs, and manages its AI workloads. Everything you do in Studio belongs to a Client and happens inside these pieces:
Read it from left to right: in Studio you build and manage; Frida and InferNode turn that into real resources and calls; the infrastructure belongs to the Client (its GPUs, its downloaded models, its identity provider).
Client
The organization you belong to: a university, a company, a department. Each Client has its own members, agents, workflows, labs, and credentials. Nothing is shared between Clients. The name of the active Client appears in the top bar.
Member and role
Each person signs in with their own account and has a role within the Client: member or administrator. The role decides which sections you see and what you can change. Details in Roles.
Agent
A configured assistant: a model, instructions, generation parameters, tools, and knowledge sources. An agent can work alone or in a team of several agents. You test it from Studio and use it inside workflows.
Workflow
A flow of connected nodes that processes inputs (a message, a document, an API call) and produces outputs. Agents are one type of node. You build a workflow in the visual editor, publish it, and trigger it by webhook, WhatsApp, or API. Each run leaves a trace and a cost.
Lab
An isolated environment with CPU, memory, GPU, and disk for data science: notebooks, pipelines, training, experiments, and serving. Labs are provisioned by Frida, the Saptiva AI control plane, from a size and within the Client's quota. They open in their own work panel.
Models and InferNode
The Catálogo de modelos (Model catalog) lists the models available in your infrastructure and their storage. InferNode is the gateway that exposes those models (and those of external providers) through a single OpenAI-compatible API, applying limits, guardrails, and usage metering per Client. Agents and workflows call the models through InferNode.
Credential
A key to use an external provider (a model, a storage service, WhatsApp, an MCP server) or InferNode itself. You create, rotate, share, and revoke credentials from Studio; agents and workflows reference them without seeing the secret.