KuchikiRenji/ai-medical-chatbot
Production-ready AI medical chatbot for healthcare. IBM WatsonX, OpenAI, RAG, Llama 3, Mixtral. Medical interviewer, fine-tuning, Gradio. Python 3.9+.
AI Medical Chatbot – Healthcare Conversational AI with RAG & LLMs
Production-ready AI medical chatbot for healthcare. Open-source medical consultation assistant using IBM WatsonX, OpenAI, RAG, and multiple LLMs (Llama 3, Mixtral, GPT).
Features · Quick Start · Installation · Usage · Project Structure · FAQ · Author
What is this?
AI Medical Chatbot is an open-source, production-ready healthcare conversational AI system. It helps with medical consultation assistance using Retrieval Augmented Generation (RAG), IBM WatsonX, OpenAI, and multiple foundation models (e.g. Llama 3, Mixtral, GPT-4). It includes a medical interviewer, fine-tuning pipelines, and Gradio web UIs.
Disclaimer: This software does not replace a doctor. It helps explore possible health-related information. Always consult a qualified healthcare professional for medical advice.
| Version | 2.0.0 |
| Status | Production-ready |
| License | Apache 2.0 |
Features
- Multi-model support – flan-ul2-20b, mt0-xxl-13b, gpt-neox-20b, flan-t5-xxl-11b, mpt-7b-instruct, OpenAI GPT-4/3.5, Meta Llama 3, fine-tuned variants
- RAG-powered answers – Retrieval Augmented Generation for context-aware medical responses
- Vector stores – Milvus, FAISS, ChromaDB for similarity search
- Medical interviewer – Chatbot that conducts structured medical interviews
- Gradio UIs – Web interfaces for chatbot and interviewer
- Production setup – Tests (pytest), type hints, PEP 8, Makefile, optional CI/CD
Quick Start
Prerequisites
- Python 3.9+
- UV (recommended) or pip
- API keys: OpenAI and/or IBM WatsonX
Installation
With UV (recommended):
pip install uv
git clone https://github.com/KuchikiRenji/ai-medical-chatbot.git
cd ai-medical-chatbot
make install
# Optional: make install-dev | make install-gpuWith pip:
git clone https://github.com/KuchikiRenji/ai-medical-chatbot.git
cd ai-medical-chatbot
python -m venv venv
source venv/bin/activate # Windows: venv\Scripts\activate
pip install -e .Environment
Create a .env in the project root:
OPENAI_API_KEY=your_openai_api_key_here
# Optional: WatsonX
WATSONX_API_KEY=your_watsonx_api_key_here
WATSONX_PROJECT_ID=your_project_id_here
# Optional: Milvus
REMOTE_SERVER=127.0.0.1
SYSTEM_MESSAGE=You are a helpful medical assistant.Usage
Run the medical chatbot:
make run-chatbot
# or: python 5-HuggingFace/app.pyRun the medical interviewer:
make run-interviewer
# or: python 8-Interviewer/hf/app.pyDevelopment commands:
make help # List all targets
make format # black + isort
make lint # flake8, pylint
make type-check # mypy
make test # pytest
make test-cov # tests + coverage
make clean # Remove artifactsProject Structure
| Component | Description |
|---|---|
| 1-Environment | Local environment and model setup |
| 2-Data | Medical datasets and preprocessing |
| 3-Modeling | RAG, feature engineering, model building |
| 4-Chatbot | Chatbot logic and references |
| 5-HuggingFace | Production chatbot app (Gradio, WatsonX) |
| 6-FineTunning | Fine-tuning medical LLMs |
| 7-Multimodal | Multimodal medical chatbot (e.g. images) |
| 8-Interviewer | Medical interviewer app |
Related resources
- WatsonX + Milvus: Watsonx-Assistant-with-Milvus-as-Vector-Database
- Custom LLM chatbot: Medical-Chatbot-with-Langchain-with-a-Custom-LLM
- Playground / demos: Medical-Llama3-Chatbot, AI-Medical-Chatbot, Medical-Interviewer
- Fine-tuned models: Medical-Llama3-8B, Medical-Llama3-v2, Medical-Mixtral-7B-v2k
- Empathy / NVC: Empathy Chatbot v1, empathyondemand
- Watsonx Medical MCP Server: watsonx-medical-mcp-server
FAQ
What is an AI medical chatbot?
An AI medical chatbot is a conversational system that uses natural language and (often) RAG/LLMs to help with health-related questions. It does not replace a doctor and should be used only as an informational aid.
Which models are supported?
The project supports IBM WatsonX models (e.g. flan-ul2, mt0-xxl, gpt-neox, flan-t5, mpt-7b), OpenAI GPT-4/GPT-3.5, and fine-tuned Llama 3 and Mixtral variants.
Do I need WatsonX and OpenAI?
You need at least one: OpenAI and/or IBM WatsonX. Configure the corresponding API keys in .env.
Is this for production use?
The codebase is structured for production (tests, type hints, Makefile). Deployment and compliance (e.g. HIPAA, local regulations) are your responsibility.
How do I run the medical interviewer?
Use make run-interviewer or python 8-Interviewer/hf/app.py. See 8-Interviewer.
Contributing
Contributions are welcome: bug reports, feature ideas, and pull requests. Please follow usual open-source guidelines for patches and issues.
License
This project is licensed under the Apache License 2.0. See LICENSE for the full text.
Author
KuchikiRenji
- Email: KuchikiRenji@outlook.com
- GitHub: github.com/KuchikiRenji
- Discord:
kuchiki_renji
Acknowledgments
- IBM WatsonX, OpenAI, and Hugging Face for models and infrastructure
- Tilburg University for empathy/NVC research collaboration
- The open-source community for contributions and feedback