Concierge FAQs

Question 1

What is Concierge and what problem does it solve?

Accepted Answer

Concierge is a declarative framework designed to build reliable Agentic AI Services. It enables Large Language Models (LLMs) to navigate, interact, and transact with complex web applications by providing structured, defined workflows and controlled autonomy for AI agents.

Question 2

How does Concierge ensure control and reliability for AI agents?

Accepted Answer

Concierge provides granular control by allowing developers to define explicit workflows, stages, and tasks. It enforces rules, validates prerequisites before task execution, and manages transitions between stages, ensuring agents follow a predefined, reliable path through the application.

Question 3

What are 'Stages' and 'Tasks' in Concierge?

Accepted Answer

Stages represent logical sub-steps towards a goal within a workflow (e.g., 'checkout'). Each stage can contain multiple 'Tasks,' which are the smallest units of callable business logic (e.g., 'add_to_cart') that an agent can execute at that specific point in the workflow.

Question 4

How does Concierge manage persistent data across agent interactions?

Accepted Answer

Concierge utilizes a global 'State' management system. This state persists across all stages and tasks within a workflow, allowing critical context (like cart items, user details, or transaction data) to be maintained and propagated as the agent navigates through the application.

Question 5

What types of applications can be built with Concierge?

Accepted Answer

Concierge is ideal for building any application where LLMs need to interact with web services in a structured, controlled manner. Examples include e-commerce platforms (shopping, checkout), ride-sharing apps, food delivery services, travel booking systems, and payment processing workflows.

Concierge

概要

主な機能

ユースケース