The AI Flow Builder is Galantis AI’s active assistance mode — it engages while you are building a flow on the canvas, suggesting appropriate node configurations based on your stated goal and highlighting issues as they appear. It is powered by the LarAgent framework and operates as an intelligent layer on top of the standard visual flow editor.Documentation Index
Fetch the complete documentation index at: https://docs.digifist.com/llms.txt
Use this file to discover all available pages before exploring further.
What this covers
- How AI assistance activates during flow building
- Trigger and action suggestions
- How node-level highlighting works
- Optimization suggestions surfaced during building
- The LarAgent framework
- What AI assistance does and does not control
How AI assistance works during building
Galantis AI monitors the state of the flow canvas as you build. It does not require you to prompt it or switch to a separate interface — assistance surfaces contextually based on what nodes are placed and how they are configured. AI assistance operates in two modes during building: Goal-based suggestions — When you start a new automation, Galantis AI can accept a description of what you want the flow to accomplish and suggest a starting structure: the most appropriate trigger, a recommended delay, and an action sequence that fits the use case. This is most useful when you know the outcome you want but are less certain which trigger or condition structure achieves it. Inline configuration assistance — As you place nodes and configure them, Galantis AI monitors for incomplete or potentially problematic configurations and surfaces targeted suggestions at the node level. It does not wait for you to finish the entire flow before offering input.Trigger and action suggestions
When Galantis AI suggests a flow structure based on a goal, it maps the described use case to the available trigger types and action sequences in Galantis. Examples of how goals map to suggestions:| Goal described | Suggested trigger | Suggested structure |
|---|---|---|
| Recover abandoned checkouts | ABANDONED_CHECKOUT | Delay 30 min → Condition (Order Value) → Action (VIP or standard template) |
| Welcome new customers | CUSTOMER_CREATED | Delay 10 min → Action (welcome template) |
| Re-engage lapsed buyers | USER_ADDED_TO_SEGMENT | Delay 1 hour → Action (win-back template) |
| Notify subscribers when stock returns | BACK_IN_STOCK | Action (restock notification template) |
| Cross-sell after purchase | ORDER_PLACED | Condition (Product tag) → Delay 3 days → Action (recommendation template) |
Node-level highlighting
When Galantis AI detects an issue or opportunity at a specific node, it highlights that node visually on the canvas and surfaces a description in the node’s settings panel. Highlights fall into two categories: Error highlights — Configuration issues that will block activation. These are also caught by the formal validation step, but surfacing them during building means you can resolve issues as you go rather than encountering a list of errors at the end. Error-highlighted nodes show the specific problem — a missing template assignment, an unconnected branch, or a condition with incomplete logic. Advisory highlights — Optimization suggestions that will not block activation but represent a meaningful improvement opportunity. These are discussed in more detail below.Optimization suggestions
Beyond errors, Galantis AI surfaces optimization suggestions on flows that are structurally valid but could perform better. These suggestions draw on automation best practices: Missing delay after trigger — An automation that sends a message immediately on trigger — with no delay node — is flagged. For most use cases (abandoned checkout, new customer welcome, post-purchase), immediate dispatch produces a worse customer experience than a short delay. Galantis AI surfaces this when no delay exists between the trigger and the first action. Galantis AI also highlights incomplete or conflicting node configurations as optimization suggestions beyond blocking errors.The LarAgent framework
Galantis AI is powered by LarAgent — the AI framework underlying the flow builder’s intelligent assistance capabilities. LarAgent handles the interpretation of flow state, the mapping of merchant goals to node structures, and the evaluation logic behind both error detection and optimization suggestions. LarAgent operates on the structured JSON representation of the flow — the samenodes and edges arrays that define the flow canvas. It evaluates the graph structure, node configurations, and their relationships to identify issues and generate suggestions.
What AI assistance does and does not control
Understanding the boundaries of AI assistance prevents confusion about what Galantis AI changes versus what remains in merchant control: Galantis AI does:- Suggest trigger types and flow structures based on described goals
- Highlight nodes with configuration errors during building
- Surface optimization advisory suggestions on valid flows
- Run formal validation checks before activation
- Identify unapproved templates assigned to Action Nodes
- Create or submit templates — template creation and approval remain manual processes in the Templates module
- Automatically fix errors — it identifies and describes them, but all changes are made by the merchant on the canvas
- Send messages autonomously — all message dispatch happens through the same automation execution engine used by manually built flows
- Override merchant configuration — AI suggestions are suggestions, not enforced changes. A merchant can dismiss any advisory highlight and activate a flow that does not follow the suggested structure
Related guides
- Flow Validation — The formal validation checks run before activation
- Automations — Flow Builder — The underlying canvas, node types, and flow structure
- Automations — Triggers — All available trigger types Galantis AI can suggest
- Automations — Recipes — Pre-built flow examples that reflect the structures Galantis AI recommends