Skip to content

Latest commit

 

History

History
64 lines (42 loc) · 2.93 KB

SETUP_GUIDE.md

File metadata and controls

64 lines (42 loc) · 2.93 KB

Setup Guide: Flow Agentforce Chat

This guide covers the configuration for the three AI implementation options provided in this package.

🏗️ Architecture Comparison

Feature Agentforce Agent Chat Models API Chat Flow Integration
Backend Native Apex Controller Native Apex Controller Native Flow Engine
Authentication Current User Session Built-in Current User Session
Connected App ❌ Not Required ❌ Not Required ❌ Not Required
Use Case Persona-based Agents Direct LLM Interaction Structured Data Collection

1. Agentforce Agent Chat (flowAgentforceAgentChat)

The most common use case. It allows you to embed a full-featured Agentforce agent into a Flow screen.

Configuration Steps:

  1. Permissions: Assign users the Flow Agentforce Chat Access and the standard Einstein Agent User permission sets.
  2. Agent API Name: In the Flow/App Builder, set the agentId property to the API Name of your agent (found in Setup > Agents).

New Features (v2.2):

  • Agent Joining Indicator: Shows an animated "Agent is joining..." message immediately upon session start for better UX.
  • Auto-Start: If enabled, the component will wait autoStartDelayMs and then automatically send the defaultMessage.
  • Placeholder Mode: if autoStart is false, the defaultMessage appears as the input placeholder.
  • Smart Parsing: The Apex controller automatically cleans up JSON-formatted responses from the Atlas reasoning engine.

2. Models API Chat (flowAgentforceChat)

Direct interaction with LLMs without an Agent persona.

Requirements:

  • Einstein Licensing: Org must have Einstein 1 or Sales/Service Einstein credits.
  • Model Name: Defaults to sfdc_ai__DefaultGPT4Omni. Can be changed to any model configured in Einstein Studio.

3. Flow Integration System (flowRenderer)

Used to render a Salesforce Flow directly inside a chat session as an interactive component.

⚠️ Beta Note: In the current Beta release, this component is focused on UI rendering. Automated capture of Flow output variables into the agent session is planned for the GA release.

Key Components:

  • flowRenderer: The UI component that handles the multi-step Flow interaction.
  • FlowDetails: Apex class that holds configuration for the Flow to be rendered.
  • Launch Flow: Agent action used to trigger the UI from a conversation.

🔍 Common Troubleshooting

Authentication Issues

If you receive a 401 error, verify:

  1. The user has the Einstein Agent User permission set and the Flow Agentforce Chat Access permission set.
  2. The agent is Active in Agent Builder.

ID vs API Name

The agentId field must be the API Name (e.g., Technical_Support_Agent). If you use the 18-character ID starting with 0Xx, the call will return a 404 error.