Glossary
Adapter — A thin integration layer that connects MIFY to an external platform (Dify, Flowise, Langflow, n8n).
Audit Event — An immutable record of a significant action in the system, used for compliance and debugging.
Baseline — A reference execution used for comparison with subsequent runs. Helps identify differences in outputs.
Batch — A group of workflow executions run together with shared concurrency and error handling settings.
Canvas — The visual drag-and-drop workflow editor where you build and connect nodes.
Credential — Sensitive information (API keys, tokens) stored securely and referenced by ID in node configurations. Never stored in workflow definitions.
DAG (Directed Acyclic Graph) — The required structure for all MIFY workflows. Data flows in one direction with no loops.
Deterministic — A node execution class where identical inputs always produce identical outputs. Fully cacheable and safely retryable.
Edge — A connection between two nodes in a workflow graph, defining data flow from source to target.
Execution — A single run of a workflow with specific inputs. Each execution has a unique ID and records all inputs/outputs.
Execution Run — An async GPU execution job (image, video, or vision) submitted to external providers.
HITL (Human-in-the-Loop) — A workflow pattern where execution pauses for human review or approval before continuing.
LiteLLM — An open-source proxy that provides a unified API for 90+ AI providers, used by MIFY for provider coverage beyond the 16 native providers.
MCP (Model Context Protocol) — A protocol for connecting AI models to external tool servers.
Marketplace — The MIFY platform for buying and selling AI workflows and digital assets with escrow protection.
Node — A single unit of work in a workflow graph. Nodes take inputs, perform operations, and produce outputs.
Opaque — A node execution class for external platform calls where behavior is a black box to MIFY.
Plugin — An extension package that adds custom nodes, adapters, or credentials to MIFY.
Probabilistic — A node execution class for AI models where output may vary between runs.
Provider Pack — A bundled provider configuration selectable at run time (fast/balanced/accurate tiers).
RAG (Retrieval-Augmented Generation) — A pattern that retrieves relevant documents from a knowledge base before generating AI responses.
Template — A pre-built workflow available for use or customization. MIFY includes 140 built-in templates.
Training Run — A model fine-tuning job submitted with a preset and dataset.
Trigger — The starting point of a workflow that defines how and when execution begins (manual, webhook, schedule, or event).
Trust Level — A classification of execution reproducibility: trusted, semi-trusted, or opaque.
Workflow — A sequence of connected processing steps represented as a DAG. Also called a graph.
Workspace — An organizational unit containing users, workflows, credentials, and executions.