Skip to content

deans-code/anamnesis

Repository files navigation

Anamnesis

🎥 Background

This application was created to test the capabilities of Google DeepMind's MedGemma models, specifically its 27B variant which supports complex text and multimodal medical knowledge and reasoning.

Warning

This project is a proof of concept and should not be used for any real medical purposes. It is intended for research and educational purposes only.

Warning

I believe that the use of AI in medical applications is a promising area of research, however, it is crucial to remember that AI models are not perfect and can make mistakes. It is important to use them with caution and always consult a medical professional for any health-related concerns.

Anamnesis, from Wikipedia:

The medical history, case history, or anamnesis (from Greek: ἀνά, aná, "open", and μνήσις, mnesis, "memory") of a patient is a set of information the physicians collect over medical interviews. It involves the patient, and eventually people close to them, so to collect reliable/objective information for managing the medical diagnosis and proposing efficient medical treatments.

The application is a simple web based chat interface which helps a user explore medical conditions or symptoms.

✅ Scope

  • Simple web interface.
  • Collect patient information.
  • Propose questions to ask the patient.
  • Define effective method of reviewing OWASP remediation implementation (manual, partially automated, fully automated?).

🔭 Future Gazing

TBC

🪲 Known defects

No known defects.

🔮 Use of AI

AI-assisted development

GitHub Copilot was used to assist in the development of this software.

Spec-driven development with OpenSpec

I am using OpenSpec for a specification-driven development approach, which means that I am writing specifications for the software before implementing features.

I am trialling the use of Claude Sonnet 4.6 for authoring the specifications and Qwen 3.6 (running locally via Ollama) for implementing the software based on the specifications.

I am using custom instructions, skills and agents to guide the implementation.

The choice of model for implementation may need to be dynamic based on the complexity of the individual specification, with more complex tasks being routed to a more capable model.

Note

Initial testing suggests that when completing more complex tasks such as the creation of the initial architecture, Claude Sonnet 4.6 is more effective. Qwen 3.6 and Gemma 4 running locally on Ollama, and GPT-5 Mini and Haiku 4.5 running on GitHub Copilot all struggled to follow guidance specified in the custom instructions, skills and agents.

Awsome GitHub Copilot

I am selecting useful add-ons from the Awsome GitHub Copilot collection, which can be installed using the GitHub CLI.

OWASP security review and remediation

I am using the Agent Owasp Compliance skill from Awesome GitHub Copilot to perform a security review based on the OWASP ASI Top 10.

I am then using OpenSpec to author a detailed remediation specification based on the identified issues.

Note

I am using using Qwen 3.6 (running locally via Ollama) for the OWASP review. The OpenSpec remediation specification is being created using Claude Sonnet 4.6. This is under test, initial indications are that Qwen 3.6 creates acceptable reviews but struggles to create detailed specifications based on the review.

Warning

I am using this software as a playground for establishing effective workflows. The OWASP review and remediation process is under development and may not be effective at identifying or remediating all security issues. It should not be relied upon for real world security reviews or remediation purposes but could be used as a basis for further development of an effective process.

🚀 Getting Started

💻 System Requirements

Software

GitHub CLI

For installing GitHub Copilot Skills, you will need the GitHub CLI:

https://github.com/cli/cli/blob/trunk/docs/install_windows.md

Hardware

A system capable of running Ollama is required.

Details of my personal system are below.

APU

Note

The hardware in use on my PC includes an Accelerated Processor Unit (APU) which combines CPU and GPU on a single chip. Recommendations for alternative hardware can be found here, performance will depend upon the models you choose to run (and other operational factors).

💾 System Configuration

Ollama

Follow the Ollama documentation to install and configure Ollama on your system.

You will need to pull the MedGemma:27b to run the application.

🔧 Development Setup

Clone the repository.

Open in Visual Studio code.

Build the projects.

⚡ Features

Simple chat interface for exploring medical conditions and symptoms.

The application is designed to help individuals understand potential medical conditions based on symptoms they input, and to suggest follow-up questions that a healthcare provider might ask to gather more information.

Warning

This software is designed for testing and educational purposes only. It is not intended for real medical use and should not be used as a substitute for professional medical advice, diagnosis, or treatment.

📎 Usage

Start the application.

Begin a conversation with the chatbot by entering symptoms or conditions you would like to explore.

The chatbot will ask follow-up questions to gather more information and provide insights based on the input.

You can end the session by clicking the "End Session" button.

A summary of your conversation will be displayed at the end.

👋 Contributing

This repository was created primarily for my own exploration of the technologies involved.

🎁 License

I have selected an appropriate license using this tool.

This software is licensed under the MIT license.

📖 Further reading

More detailed information can be found in the documentation:

About

Exploring the capabilities of MedGemma, a specialist open weight medical LLM by Google DeepMind.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors