This document is a basic guide to Agent Assist resources. More details
about the resources described here can be found in the
[reference documentation](https://docs.cloud.google.com/gemini-enterprise-cx/agent-assist/reference-index.md.txt).
We recommend that you familiarize yourself with this guide before diving into
Agent Assist.

## Conversations

A [Conversation](https://docs.cloud.google.com/dialogflow/es/docs/reference/rest/v2/projects.conversations.md.txt)
is a resource that represents an exchange between two or more
[participants](https://docs.cloud.google.com/gemini-enterprise-cx/agent-assist/basics.md.txt#participants). Every time a
conversation participant's utterance is passed to Agent Assist as part
of a Conversation resource, a **message** is created and stored for the
conversation.

## Conversation profiles

Each conversation is created using a **conversation profile** . Each
conversation profile contains a set of parameters that control the
suggestions provided to the human agent during a conversation. You can also configure a conversation profile for use with a [virtual agent](https://docs.cloud.google.com/dialogflow/cx/docs.md.txt), which autogenerates a reply
to the end user.

You can create a conversation profile either by using the
[Agent Assist Console](https://docs.cloud.google.com/gemini-enterprise-cx/agent-assist/conversation-profile.md.txt)
or by [calling the API directly](https://docs.cloud.google.com/gemini-enterprise-cx/agent-assist/how-to.md.txt).

## Participants

Each conversation can have the following
[participants](https://docs.cloud.google.com/dialogflow/es/docs/reference/rest/v2/projects.conversations.participants.md.txt):

- `END_USER`: Participant is an end-user. This is the person the agent is conversing with.
- `AUTOMATED_AGENT`: Participant is an [automated agent](https://docs.cloud.google.com/dialogflow/cx/docs.md.txt).
- `HUMAN_AGENT`: Participant is a human agent.

## Virtual agents

Dialogflow agents are virtual agents that you can build and configure for
conversations with your end-users. You can design your system so that end-users
initially interact with a
[virtual agent](https://docs.cloud.google.com/dialogflow/cx/docs.md.txt) before being escalated to a
human agent. You can use either a
[Dialogflow ES](https://docs.cloud.google.com/dialogflow/es/docs/quick/build-agent.md.txt) or
[Dialogflow CX](https://docs.cloud.google.com/dialogflow/cx/docs/quick/build-agent.md.txt) agent.

## Conversation stages

A conversation with an end-user has multiple stages:

1. **Connection stage**: The end-user initiates a chat dialog.
2. **Virtual agent stage** : The end-user is connected to a [Dialogflow virtual agent](https://docs.cloud.google.com/dialogflow/cx/docs.md.txt). The virtual agent communicates with the end-user and attempts to resolve all issues and requests. You can set up your system to bypass the virtual agent and handoff stages.
3. **Handoff stage** : If the virtual agent cannot resolve all end-user issues, the conversation is [handed off](https://docs.cloud.google.com/gemini-enterprise-cx/agent-assist/handoff-cxas.md.txt) to a human agent.
4. **Agent Assist stage** : The end-user is connected to one or more human agents. [Agent Assist](https://docs.cloud.google.com/gemini-enterprise-cx/agent-assist.md.txt) supplies real-time document and response suggestions to the human agent that are relevant to the conversation.
5. **Termination stage**: The end-user is disconnected from the text chat.

## Agent desktop

To implement Agent Assist for text-based conversations you must
integrate the Agent Assist API into your agent desktop.