Workflow Configuration

The Workflow section defines how the AI agent manages conversation flow. It uses a state-machine model, where conversations are divided into states. Each state determines available actions, accessible tools, and transitions between conversation stages.

Workflow Concepts and Configuration

In Flametree, conversations are managed through a state-based workflow. Each conversation progresses through one or more states, where the agent follows defined instructions and gains access to specific tools.

The table below shows how conceptual workflow elements map to their YAML configuration.

Concept	Description	YAML Configuration
State	A phase of conversation (for example, greeting or information gathering)	`name`, `kind: StateConfig`
Initial State	Starting point of the interaction	`init_state: true`
State Logic	Defines what the agent does during this stage	`description`
Available Tools	Actions the agent can perform	`available_tools`
Process	Overall business scenario	`process_name`

For more details about Workflow configuration, please refer to the Workflow section of the documentation.

Example State Configuration

- process_name: SalesProcess
  name: SalesSupport
  kind: StateConfig
  init_state: true

  description: |
    - Answer questions about hardware
    - Collect contact info
    - Offer call scheduling if customer is interested
    - Use FAQ and Knowledge Base when relevant

  available_tools:
    SingleStatefulOutboundAgent:
      - SendChatMessage
      - FinishSession
      - SingleSearchFAQ

Workflow Editor Behavior

When editing workflow YAML, be aware of the following editor-specific behaviors.

Comment highlighting

Comments (#) starting from the 4th character onward may not appear highlighted in green. Placing the comment at the beginning of a line restores syntax highlighting. This is a visual limitation of the editor and does not affect execution.

Saving behavior

If the workflow YAML contains a syntax error, pressing Save discards all unsaved changes and restores the last valid version of the configuration. Always validate YAML syntax in an external editor before saving.

Required Tools by Functionality

Tools define what actions an AI agent can perform during a conversation, such as sending messages, retrieving information, or ending a session. In a workflow, tools are explicitly assigned to states, which determines which capabilities are available to the agent at each stage of interaction.

Functionality	Required Tools
Basic Chat	`SendChatMessage`, `FinishSession`
Knowledge Base	`SingleSearchFAQ`
Operator Handoff	`SuggestChatMessage`, `TransferToTheOperatorSkill`
Notifications	Custom tools for email/SMS integration

How Workflow Differs from Other Configuration Areas

The workflow works together with the Identity and Task sections of an AI agent, but it serves a different purpose.

Configuration Area	Defines	Purpose
Identity	Who the agent is	Identity establishes role, expertise, and how the agent presents itself to customers
Task	What the agent must achieve	Task defines objectives and operational boundaries
Workflow	How the agent behaves	Workflow controls conversation flow, decisions, and tool usage

Basic Single-State Workflow Example

Workflows may consist of a single state for simple agents or multiple states for more complex interaction scenarios.

- process_name: SalesProcess
  name: SalesSupport
  kind: StateConfig
  init_state: true

  description: |
    - Answer questions about products
    - Collect contact info
    - Offer call scheduling if the customer is interested
    - Use FAQ and Knowledge Base when relevant

  available_tools:
    SingleStatefulOutboundAgent:
      - SendChatMessage
      - FinishSession
      - SingleSearchFAQ    

Workflow Concepts and Configuration​

Example State Configuration​

Workflow Editor Behavior​

Comment highlighting​

Saving behavior​

Required Tools by Functionality​

How Workflow Differs from Other Configuration Areas​

Basic Single-State Workflow Example​

Related Resources​