
Auralyn AI is a real-time conversational voice assistant built for the AssemblyAI Voice Agent Hackathon. It allows users to speak naturally with an AI assistant while viewing live transcripts, assistant states, tool activity, and real-time external information through a modern web interface. Auralyn combines real-time voice interaction with practical live tools for weather, world time, and news. It uses a FastAPI WebSocket bridge to connect the browser interface with the Python backend, while the AssemblyAI Voice Agent API powers the real-time conversational experience. The assistant also stores conversation history in JSON and can load recent context into future sessions, enabling basic cross-session continuity.
30 Sep 2026

DevGuard AI is an autonomous repository rescue workflow built with IBM Bob 2.0 to help developers understand, repair, test, document, and verify unfamiliar codebases. Developers often inherit repositories with hidden bugs, weak validation, incomplete tests, unsafe configuration, duplicated logic, and poor documentation. Before they can safely modify the code, they must first spend significant time understanding the architecture and identifying risks. DevGuard AI turns IBM Bob 2.0 into a structured engineering workflow with focused stages for: - baseline repository analysis - architecture understanding - risk and defect detection - test-gap analysis - prioritized repair - validation and reliability improvements - documentation enhancement - final verification and health scoring For the hackathon, we created a realistic Flask task-management API containing controlled engineering defects. IBM Bob 2.0 was then used to analyze the repository, identify risks, repair high-priority defects, expand automated testing, improve documentation, and verify the final result. The repository health score improved from 26/100 to 91/100, while the automated test suite increased from 3 passing tests to 18 passing tests. All eight controlled defects were addressed, and the final verification status was PASS. DevGuard AI demonstrates a shift from simply asking AI to generate code toward using AI as a structured software-engineering system that can improve repository health in a measurable and repeatable way. The project includes IBM Bob task-session evidence in the required bob_sessions folder, a live dashboard, a public GitHub repository, automated tests, documentation, before-and-after reports, and a final verification summary. Business value: DevGuard AI can reduce the manual effort needed for developer onboarding, codebase understanding, debugging, test expansion, documentation improvement, and repository maintenance.
27 Sep 2026