Glossary
Plain-language definitions of the AI terms we use across GOGOGO LLC, our products, and our blog — from "AI agent" and "orchestrator" to FAISS and rerankers.
A software program with a specific role, a defined toolset, and the autonomy to read context, decide on actions, and execute them. Distinct from a single-shot chatbot: an agent loops between thinking, calling tools, and observing results until the task is done.
An architecture in which several specialised AI agents collaborate to solve a problem — typically one orchestrator routes work to domain agents (HR, content, signage, analytics) which hand off intermediate results between each other instead of a single model doing everything alone.
The agent at the top of a multi-agent system. Reads the incoming intent, classifies it, picks which specialised agent should handle it, and tracks the conversation state across hand-offs. The orchestrator owns the run, not any one tool call.
An agent's ability to call external functions, APIs, or services to actually do something — generate an image, send a WhatsApp message, query a database — instead of producing only text. Every GOGOGO agent ships with a strictly-typed tool belt.
The act of one agent transferring control to another, carrying typed context with it. Hand-offs are how multi-agent systems stay coherent across a long task — each agent reads what the previous one decided, instead of re-deriving everything from scratch.
A pattern where the agent retrieves relevant documents, embeddings, or rows from a knowledge base before generating its answer. RAG keeps responses grounded in real data instead of model-internal memory, and is the basis of GoTrack's vision retrieval pipeline.
Facebook AI Similarity Search — an open-source vector index for fast approximate nearest-neighbour lookup. GoTrack uses FAISS to retrieve candidate product matches from CLIP embeddings before a reranker picks the final match.
A second-stage scoring model that takes the top N candidates from a fast retriever (like FAISS) and re-orders them using richer contextual signals — colour, hanger zone, time-of-day, recent activity, store-specific priors. GoTrack's reranker uses seven such signals.
Persistent state an agent reads and writes across turns. Short-term memory holds the current conversation; long-term memory holds tenant-specific knowledge — past decisions, document embeddings, preferred phrasing. Always permission-scoped to the tenant.
Built-in tracing of every span, tool call, and state transition inside the agent runtime. GOGOGO captures this by default — no "we'll add logging later". Required for replay-driven debugging and weekly improvement loops.
The rules an agent is not allowed to break — what data it can read, what actions it can execute without approval, what content it must refuse. Guardrails are enforced at the tool layer, not just the prompt.
A predefined route for an agent to bail out to a human (or to another agent) when confidence is low, the request is sensitive, or guardrails block automation. Designed up-front during the Design phase, not bolted on after launch.
An outbound HTTP request a system sends when an event happens. GOGOGO products fire webhooks for completed jobs, agent decisions, approvals, and system events — so customer ERPs, payroll, and data warehouses can react in real time.
An architecture where a single deployment serves many independent customers, each with isolated data, permissions, and configuration. All four GOGOGO products are multi-tenant by default — required for any business-grade AI platform.
Meta's official API for businesses to send and receive WhatsApp messages at scale, with templated messages, opt-in flows, and rich content. GoPeople is built on top of it to deliver HR workflows directly in the channel employees already use.
A content management system for screens — TVs, kiosks, LED walls — across many locations. Lets a central team schedule, version, and deploy content to thousands of screens with audit trail. GoVista is GOGOGO's digital signage CMS.
Computer-vision technique that identifies when a shopper picks up a specific product in a store — without barcodes, stickers, or RFID. GoTrack uses pickup detection to trigger live signage that swaps creative based on what's getting touched.
A pre-generation agent that rewrites a user's terse description into a richer, model-friendly prompt — adding style cues, composition hints, and resolution targets. Goddo's prompt enhancer is what makes a one-line idea become a production-grade visual.