Automated Flow Discovery
Understanding the heuristic logic used by the SyncFlo AI Agent to dynamically trigger and serve WhatsApp Flows during active sessions.
The Intent-to-Flow Mapping
The SyncFlo AI Agent does not use hardcoded triggers. instead, it utilizes a Runtime Discovery Protocol to match user intent with published flow assets.
1. Metadata Indexing
Upon publication, every flow's metadata (Name, Description, and Field keys) is indexed into a semantic vector database accessible by the agent.
2. Semantic Search
When a user makes a request (e.g., "I want to register for the workshop"), the agent performs a semantic lookup to find the most relevant flow id.
The Invocation Pipeline
The agent monitors the conversation stream to detect high-confidence action triggers that require structured data input.
The agent verifies that the targeted flow is in the `PUBLISHED` status within the Meta Business infrastructure.
The agent generates a specific `action` command that sends the flow link to the user, initiating the native UI session.
Optimizing for Discovery
To ensure high accuracy in AI discovery, adhere to the following naming conventions:
- Use descriptive, unique titles for your Flows (e.g. `Client_Booking_Flow` vs `New_Flow`).
- Provide clear component labels that describe the required data.
- Ensure the **Flow Status** is verified as `PUBLISHED` after any visual updates.