AI Agents

Chat Assistant overview: What Chat Assistant is, what parts it consists of, and how the agent works at a high level

Core concepts and terms: Basic entities: account, channel, chat, dialog, agent, step, signal, and others

Chat events and the dialog lifecycle: Which events in the chat start the agent, the dialog owner, the manager command

The agent's work cycle: What happens during a single run: iterations, available steps, stopping

Agent settings: Channels, working hours and schedules, reply style, reminders, context clearing, catalog, knowledge base training

Agent instructions: The prompt and its slots, the work dashboard, where to place which instructions

Scenario steps: What a step consists of: events, visibility, data, and actions

Scenario: How steps are assembled into a scenario, how the agent works through it, the dashboard, choosing the structure

Actions: What the agent does through steps: saving data, messages, transitions, handover to a manager

Signals: Background recognition of conversation states and automatic reaction to them

Conditions: Rules for the availability of steps, actions, and signals based on conversation and environment data

Delayed actions: Launching steps by a timer: reminders, resuming work

Schedules: Working hours and custom schedules that affect the agent's behaviour

Knowledge base: Documents the agent searches for information to answer the client

Catalog subagent: Working with the product catalog and orders

Agent versions and statuses: Versioning the agent configuration, statuses, safe changes

Testing and debugging: How to check the agent's behaviour and find the causes of problems

Anti-patterns: Frequent configuration mistakes and why they lead to failures

Billing: What you are charged for: the active chat, dynamic cost, periods, invoices

Client handling history: A log of the agent's work for each chat: the course of reasoning and steps, delayed launches