Analysis updated 2026-05-18
Get gentle reminders to stand up, drink water, or adjust your environment while working.
Study a working example of wiring LLM tool calling into a LangGraph state machine.
Run a desktop mascot companion driven by structured AI agent output.
Review a full trace of an agent's tool calls and reasoning for debugging.
| alan-lj/healthdesk_agent | aim-uofa/reasonmatch | airbone42/360-data-athlete | |
|---|---|---|---|
| Stars | 12 | 12 | 12 |
| Language | Python | Python | Python |
| Setup difficulty | moderate | hard | hard |
| Complexity | 4/5 | 5/5 | 4/5 |
| Audience | developer | researcher | general |
Figures from each repo's GitHub metadata at analysis time.
Requires a DeepSeek API key and a Python virtual environment before the agent can run.
HealthDesk Agent is a health reminder tool aimed at people who spend long hours sitting at a computer. Instead of a fixed alarm that fires at set intervals, it runs an AI agent that looks at your current office situation (how long you have been sitting, whether you have been drinking water, how comfortable your environment is) and then offers a small, calm suggestion you can act on right away. It shows up as a desktop mascot called Xiao Ling that you can click, drag around your screen, and chat with. The technical core is a ReAct agent pipeline built with LangGraph and the DeepSeek API. ReAct is a pattern where the AI reasons about what to do, picks a tool to call, observes the result, and repeats until it has enough information to give a final answer. In this project, the tools are things like checking your sitting duration, reading device health data, and searching a local knowledge base of health tips. Every agent run records a full trace showing which tools were called, what the model decided, and why it stopped, which makes it straightforward to review or debug the agent's behavior. The project runs as a Python backend using FastAPI and SQLite. It has three ways to display results: a full web dashboard that shows the whole agent pipeline on one page, a lightweight browser companion for demonstrating the chat interaction, and a real Windows desktop mascot built with Tkinter that sits over your other windows and can be dragged between screens. The mascot saves its position between sessions. Right now the project uses a simulation module to generate office-state data rather than real sensors. The README notes that real hardware (a millimeter-wave radar, depth camera, water cup sensor, or temperature sensor) could replace the simulation layer without changing the agent itself, because the agent reads from a standardized data structure. The README is written in Chinese and describes this as an open prototype for anyone interested in how LLM tool-calling integrates with LangGraph, not a finished commercial product. It explicitly says the agent does not provide medical diagnoses or treatment advice.
A LangGraph and DeepSeek powered AI agent that acts as a desktop mascot reminding office workers to rest, hydrate, and adjust their environment.
Mainly Python. The stack also includes Python, FastAPI, LangGraph.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.