SupportFlow-AI 🛠️

An Intelligent Ticket Triage Agent built for the Elastic AI Agent Builder Hackathon.

SupportFlow-AI solves the "Support Bottleneck" by using an AI Agent to automate the analysis, categorization, and resolution-discovery of incoming support tickets directly within the Elastic ecosystem.

🌟 Key Features

• Automated Triage: Instantly determines Category, Priority, and Sentiment.
• Historical Discovery: Uses ES|QL to find similar resolved tickets and suggest proven solutions.
• Real-time Analytics: Provides instant queue statistics via natural language.
• Native Orchestration: 100% built using Elastic Agent Builder—no external backend required.

🏗️ Architecture

• Elasticsearch Serverless: Managed vector and text database.
• Agent Builder: Reasoning engine and tool orchestration.
• ES|QL (Elasticsearch Query Language): Powering the discovery tools for high-precision retrieval.

🛠️ Custom Tools (ES|QL)

This agent is equipped with two powerful tools:

1. search_similar_tickets: Finds historical matches using the QSTR function.
2. get_ticket_stats: Aggregates ticket volume and priority levels using STATS.

📂 Project Structure

• /agent: Exported Agent configuration (instructions & tool settings).
• /data: Elasticsearch index mappings and 30-ticket sample dataset.
• /queries: Raw ES|QL queries used for agent tools.

🚀 Getting Started

1. Create an Elasticsearch Serverless project.
2. Run the mapping in data/index-mapping.json via Dev Tools.
3. Import the data/sample-tickets.json using the Bulk API.
4. Import the agent/agent-config.json into the Agent Builder.

📺 Demo Video