pswider/tula

# Technical Summary Tula implements health-specific skills on the OpenClaw framework, with intelligent model routing that directs medical imaging to specialized healthcare models (MedGemma, MedImageInsight), laboratory text extraction to domain-tuned parsers, and clinical synthesis to general reasoning models. It ingests health data across multiple channels—email with sender allowlists, photo capture via smartphone, FHIR R4 APIs, wearable device syncs, and Bluetooth home health devices—automatically classifying and routing each data type to purpose-built processing pipelines. Designed for self-hosted deployment on commodity infrastructure with Telegram as the primary interface, enabling low-bandwidth access in resource-constrained settings while maintaining HIPAA de-identification principles.