734ai/ClinAssist-Edge

ClinAssist Edge is an offline, privacy-first AI assistant for clinicians in low-resource settings, powered by MedGemma and HAI-DEF models. It delivers differential diagnosis with risk flags, SOAP note automation, patient instruction translation, advanced RAG-based reasoning, and drug safety checks. # ClinAssist Edge ### Privacy-First Offline Clinical Intelligence Platform **ClinAssist Edge** is an offline-first clinical decision support system designed for resource-constrained environments. Built on the **MedGemma-2B** foundation model, it provides differential diagnosis, automated documentation, and safety monitoring without requiring internet connectivity or cloud infrastructure. ## System Architecture The platform utilizes a modular, privacy-preserving architecture optimized for edge deployment on consumer hardware. **Core Components:** 1. **User Interface Layer**: A Streamlit-based frontend optimized for low-latency interaction and high-contrast visibility. 2. **Application Logic**: Python-based backend handling input validation, session state management, and template rendering. 3. **Inference Engine**: Local execution of the MedGemma-2B Large Language Model, supporting 4-bit and 8-bit quantization for reduced memory footprint. 4. **Safety & Ethics Module**: Rule-based gating mechanisms to detect and flag high-risk outputs (e.g., contraindicated prescriptions) before they reach the user. 5. **Retrieval-Augmented Generation (RAG)**: Vector database (FAISS) implementation for grounding model outputs in local, reliable medical guidelines. ```mermaid graph TD User[Clinician] --> UI[Streamlit Interface] UI --> Backend[Application Backend] Backend --> Safety[Safety & Ethics Gate] Backend --> RAG[Local Knowledge Base] subgraph "Offline Inference Engine" Safety --> Model[MedGemma-2B] RAG --> Model end Backend --> Logs[Encrypted Audit Logs] Model --> UI ``` ## Key Capabilities * **Differential Diagnosis**: Generates ranked differential diagnoses with associated confidence scores using Bayesian uncertainty quantification. * **Drug Safety Analysis**: Performs real-time checks for drug-drug interactions, disease contraindications, and known allergies. * **Automated Documentation**: Converts unstructured clinical notes into standardized SOAP (Sub …