Open navigationClose navigation
DRLF

Detection, Research & Learning Framework

Turn public healthcare data into research you can explain.

Ask your AI agent to clone DRLF and guide you through setup. Then start a conversation about the public healthcare data you want to understand. The research skills and tools are already included.

Start with your AI agentSee an example

AvailableOpen source · GitHub

A question becomes a research planWorkflow illustration · synthetic scenario
You ask

Why does one prescriber appear across so many vaccine claims?

Investigate with DRLF

Start with the pattern. Work toward an explanation.

  1. Understand the data

    Check what the prescriber field actually represents.

  2. Test possible explanations

    Compare the pattern with pharmacy and protocol roles.

  3. Keep the evidence together

    Record sources, findings, and what remains unknown.

Example of a guided workflow after setup. Illustrative wording, not an app screenshot or a live investigation.

Start a conversation

Point your agent at DRLF. Bring your first question.

Use a coding agent that can work with files and run commands on your computer, such as Codex. Give it the repository link and ask it to help you get started.

DRLF includes seven research skills for finding suitable public healthcare data, investigating patterns, managing evidence, and learning from your work.

Explore the included skills
Your first messageSelect and copy this into your AI agent

Clone https://github.com/wittenauer-software/drlf locally. Read its AGENTS.md and getting-started guide, then help me complete setup and create a separate private research workspace. Once it is ready, help me choose a public healthcare data question and use the included skills to investigate it.

Your agent guides setup; you confirm the environment and access it needs. Then continue the conversation in your research workspace.

How it works

Start with a question.
Build understanding, step by step.

For researchers and analysts exploring public U.S. healthcare data. Once your workspace is ready, describe what you want to understand. Your agent can follow DRLF's included skills to guide the next steps, while you ask follow-up questions and review what it finds.

  1. 01

    Ask a research question

    Choose the pattern you want to understand. Set a manageable scope and decide what evidence would help answer it.

  2. 02

    Investigate with evidence

    Explore suitable public sources, compare like with like, and test the ordinary explanations behind unusual numbers.

  3. 03

    Review what you learned

    Keep findings, sources, and open questions together. Decide what deserves a closer look—and when the available evidence is enough.

DRLF supports research leads and human review. An unusual aggregate pattern alone does not establish fraud or an improper claim.

A useful result

More than a list of unusual numbers.

A finding you can follow back to its evidence.

In this synthetic exercise, a large number of vaccine claims leads to a better question: what role does the named prescriber represent?

The invented records support a protocol role across dispensing locations. The research explains the pattern while keeping transaction-level questions open.

Explore the synthetic exercise
Synthetic research exampleUnderstanding vaccine attributionIllustrated summary of the DRLF teaching exercise
The observation
One prescriber identifier appears across two dispensing locations.
The explanation
The invented protocol record supports a standing-order role, rather than one person's individual visits.
What remains unknown
The aggregate does not establish transaction-level dispensing or who received payment.
The reusable lesson
Understand the role behind an identifier before interpreting its scale.

Based on registered synthetic aggregate and protocol fixtures. No real provider, claim, or finding is shown.

What you can explore

Public data. Focused research questions.

Medicare Part B

Practitioner services

Compare service use, patient reach, and repeat activity within relevant provider groups and time periods.

Medicare Part D

Prescription-drug patterns

Explore provider and product trends, and investigate the operational roles behind prescriber attribution.

Medicare DMEPOS

Equipment and supplies

Examine supplier activity with attention to product types, billing units, rental status, and comparable suppliers.

Supporting sources include Open Payments and provider identity and status records. Coverage varies by source and workflow. See supported public data sources.

Build on what you learn

Keep the lessons that make your next question better.

Preserve what worked, what changed your interpretation, and which explanations mattered. Reviewed lessons can guide later research in your own workspace.

You decide when a lesson should change the workflow. Learning is a reviewed research record; it does not retrain the AI model.

  1. ResearchAsk and investigate
  2. ReflectRecord what changed
  3. ReviewCarry useful lessons forward

Get started

From the first message to your first question.

Start by sharing the repository with your agent. It can walk you through the setup guide and help prepare the workspace before you begin real research.

  1. 01

    Point your agent at the repo

    Ask it to clone DRLF locally and read the project instructions. The research skills come with the project.

    View DRLF on GitHub
  2. 02

    Let it guide setup

    Have your agent check the requirements and help create a separate private research workspace. Confirm its readiness before proceeding.

    Read the setup guide
  3. 03

    Start asking questions

    Continue in that workspace. Ask your first question, explore the findings together, and let each answer shape the next question.

    Explore the capabilities

Make it your own

Add skills for the data you work with.

The included skills give you a starting point for public healthcare research. You can also work with your agent to create and approve local skills for your own datasets and research questions in your private workspace.

Your agent can help add the connections and checks a new source needs. You review the changes and decide which data it can access.

Read the data-source extension guide
What do I need to run DRLF?

The current release supports Windows 11 x64 with PowerShell 7, Git, and Python 3.13. Database-backed workflows use Docker Desktop and PostgreSQL. The offline synthetic exercises do not require a database.

You supply and configure a compatible AI agent. Installation and live data acquisition require internet access. The setup guide contains the complete requirements.

How are research data and costs handled?

You manage research files and the database in your own workspace. Keep identifiable research out of the public DRLF checkout and public project channels. Your agent's configuration and provider determine what content it receives and how it is handled.

DRLF has no project telemetry or hosted research service. Its source is available under Apache-2.0; your chosen agent and other services may have separate costs. Third-party datasets retain their own terms.

Where can I get help?

Use the DRLF support guide for documentation and best-effort software support. Public reports must use synthetic examples. Do not submit identifiable research, allegations, fraud tips, or sensitive information to the project.

DRLF's two experimental research routes remain labeled as experimental in the capability guide.

Start exploring

Put a research question to work.

Give your agent the repository. Let it guide setup, then start exploring with the research skills already included.

Use the starter messageExplore all products