Janus started when I noticed how support teams lose bandwidth on repetitive tickets. I wanted a system that could streamline intake, boost response quality, and reduce manual load.
I built a simple end to end helpdesk that lets users raise tickets, talk to an AI agent, and view updates in a clean dashboard. One agent classifies tickets by priority, type, and category. Another agent handles live chat using a knowledge base. I used MindsDB for the AI workflow, a vector store for retrieval, and a lightweight UI to keep delivery fast and usable.
I learned how much impact prompt design has on data quality, how retrieval improves answer relevance, and how important clean information flow is for scaling. I also had challenges like getting consistent JSON output, balancing accuracy with latency, and maintaining a reliable knowledge base.
The final result enables faster ticket resolution, consistent categorization, and actionable insights for admins. It creates a more efficient support pipeline and sets the foundation for future integrations and a more production ready backend.
Log in or sign up for Devpost to join the conversation.