Open-Source Medical Agent Workflows
MaziyarPanahi · x · 2026-07-14
This post outlines a reusable combo for medical agent workflows:
- GLM-5.2 runs via Unsloth's GGUF version with llama.cpp, or alternatively through Hugging Face Inference Providers.
- Gemma 4 handles structured extraction.
- OpenMed AI is used for PII de-identification and runs on MLX.
- pydantic serves as the orchestration layer for the agent loop.
The author adds that they have successfully configured GLM-5.2 into an agent capable of autonomously handling complete clinical cases: it can plan steps, invoke tools, and draft final disposal opinions, with all models sourced from Hugging Face. The post also mentions that a medical AI vendor leaked 1.4 million patient records this month due to cloud storage; the proposed solution mitigates this by de-identifying data via OpenMed first.
Related event: Open-Source Medical Agent Workflow Released(2 posts)→
More from coding & agent
- Soft Clamp cuts tool-call overuse in multi-teacher distillation, from 13.7% to 9.0% — antgroup · 2026-07-21
- Agent harness memory loss and compaction are still a major usability problem — adityaag · 2026-07-21
- SpecJudge runs locally on Ollama to pick the right-sized AI model for your project — jokiruiz · 2026-07-21
- A developer maps out six design rules for CLIs that humans and AI agents can both use — yujiezha · 2026-07-21
- A coding-agent skill that forces ADHD-friendly, answer-first output — ayghri · 2026-07-21
- A set of agent skills for CAD, robotics, and hardware design — earthtojake · 2026-07-21