This project template demonstrates "Harmony," an intelligent AI customer support agent built using Voiceflow's Agent step enhanced with Custom Functions.

What it does:
Harmony provides context-aware, knowledgeable, and time-sensitive responses. It can handle greetings, check business operating hours, answer specific questions using a knowledge base, ask for clarification when needed, maintain conversation history, and identify handoff requests.

How it works:
The workflow captures user input ({last_utterance}) and utilizes three key custom JavaScript functions:

1. Conversation Memory Manager: Maintains rolling conversation history in {custom_memory} (with multiple format options available in the function code).

2. Chronos Global Time Formatter: Provides the current time ({formattedDateTime}) in a specific timezone (configured for Sydney) for accurate operating hours checks.

3. Knowledge Base Chunks Retriever [key outputs]: An optimized function called by the Agent step to search an NFI knowledge base, returning only essential outputs (kb_success, chunks_count, chunks) for better performance and reduced token usage.

The Agent step receives user input, memory context, and current time. Its detailed prompt guides it to either handle simple exceptions directly or execute the KB function. Based on the outcome, it synthesizes a spoken response and selects an appropriate output path (KB Answer Found, KB Searched - No Answer, Direct Answer Provided, etc.) to dynamically control the conversation flow.

Who it's for:
This template is ideal for Voiceflow developers looking for a practical example of integrating custom functions for memory management, real-time data (like time), and efficient Knowledge Base retrieval within the Agent step. It's a great starting point for building advanced customer support bots, internal helpdesks, or any AI agent requiring enhanced context and knowledge grounding.