DHTI ELIXIR BASE

Base library for creating new DHTI Elixirs.

What is DHTI? DHTI (Digital Health Tooling Interface) provides building blocks for interoperable, healthcare-focused AI tooling. This repository is the minimal foundation you extend to build your own "Elixir"—a reusable, testable component (chain/agent/server) that plugs into the wider DHTI ecosystem.

Status: WIP, but stable enough to scaffold new Elixirs and iterate quickly.

Overview

This package exposes lightweight, composable base classes:

• BaseChain – a LangChain-compatible chain scaffold with sensible defaults.
• BaseAgent – an agent scaffold that can bind tools and work with LangGraph.
• BaseGraph, BaseLLM, BaseServer, BaseSpace, and BaseDhtiModel – helpers to organize more complex Elixirs.

Use these to rapidly create a new Elixir with a consistent structure, tests, and docs.

Dependency Injection (mydi.py / di)

The codebase uses a simple dependency injection pattern via mydi.py and the get_di function. This allows you to:

• Configure FHIR endpoints, authentication, and other settings at runtime
• Swap out implementations for testing or customization
• Keep your code modular and maintainable

CDS Hook Module (Frontend Integration)

The cds_hook module now provides request parsing and context extraction for CDS Hooks workflows. It supports:

• Parsing various CDS Hooks request formats
• Extracting patient and order context

This enables integration with frontend CDS Hooks clients.

FHIR Module (Backend Integration: SMART on FHIR & FHIRPath)

The fhir module includes advanced backend features:

• SMART on FHIR: The SmartOnFhirSearch class provides secure, standards-based access to FHIR resources using OAuth2 and the SMART on FHIR protocol. It supports:
  • Patient-centric resource queries (Conditions, Observations, Procedures, MedicationRequests, AllergyIntolerances)
  • $everything operation for comprehensive patient data
  • Automatic token management and authentication
• FHIRPath Support: All resource queries can apply FHIRPath expressions for flexible data extraction and transformation.

CrewAI Integration (Optional)

The crewai module provides wrapper classes that make DHTI's LangChain-based components compatible with the CrewAI framework:

• CrewAILLMWrapper: Wraps BaseLLM and BaseChatLLM for use with CrewAI agents
• CrewAIAgentWrapper: Wraps BaseAgent to create CrewAI-compatible agents
• CrewAIChainToolWrapper: Wraps BaseChain as a CrewAI tool
• CrewAILangChainToolWrapper: Wraps LangChain tools for use in CrewAI

To use CrewAI integration, install with the optional dependency:

pip install dhti-elixir-base[crewai]

See the CrewAI documentation for detailed usage examples.

Project layout

• src/dhti_elixir_base/ – base classes and minimal utilities.
• tests/ – example tests to keep your Elixir robust.
• examples/ – quick patterns for chains/graphs.
• docs/ – MkDocs configuration for documentation.

Learn more

• DHTI home: https://github.com/dermatologist/dhti
• This template docs: https://dermatologist.github.io/dhti-elixir-base/

Give us a star ⭐️

If you find this project useful, give us a star. It helps others discover the project.

Contributors

• Bell Eapen |