ADHD Manager Bot is a conversational productivity assistant built for Telegram, designed to support users with ADHD in managing tasks, reminders, and long-term memory — all in a friendly, casual tone. It supports both text and voice input using OpenAI’s Whisper, and replies naturally while using tools like QStash for reminders and Redis for memory. Trae IDE played a key role throughout development. Its AI code assistant offered fast, context-aware suggestions that felt remarkably accurate — far better than other tools I've used. With access to my codebase, I could iterate rapidly, fixing bugs and designing new features on the fly. I also used Trae's multi-model chat system to experiment with different LLMs inside the IDE itself, which helped shape the bot’s personality and tune the prompts for better memory usage and tone adaptation. Under the hood, the bot uses LangChain’s ZeroShotAgent architecture, enhanced with custom tools and memory, and deployed via AWS Lambda for low-cost serverless hosting. The architecture supports structured tool use while maintaining natural, assistant-like conversation flow. The bot handles casual voice messages just as well as structured commands — remembering your tone, matching your emoji use, and adjusting its responses accordingly. Overall, this project couldn’t have come together so quickly and smoothly without Trae IDE’s AI-first workflow. It enabled me to focus on solving real user problems instead of boilerplate. The result is a bot that feels more like a helpful friend than a productivity app.