Quick start#

Pick what you want to do. Each path takes 5 to 20 minutes and needs nothing you do not already have.

What HEART is, in one minute#

A HEART is a small, standard part that lives inside a product. It knows whether the product is doing its main job. When the job stops, it notices with no network and no AI, raises an alert with a standard fault code, and tells a person.

• Its API is the same on every product: a public Card (who it is), a Pulse (how it is now) and alerts (what is wrong).
• HEART Hub at heartcortex.ai lets the owner watch it from anywhere, on a private page.
• An AI model is optional. It may explain and propose, but only a person, or rules the owner set, can approve a change. A Command Center can take the AI away at any time.

Path 1: watch the live demos#

1. Open the live demos. Nine simulated products run there: a pacemaker, a thermostat, a dog collar, an EV charger, a greenhouse and more.
2. Pick one. Its page shows its Pulse, its alerts and what its HEART noticed, as it happens.
3. Watch a fault being staged, noticed and resolved. Each demo says which faults it stages, and when.
4. Look inside a demo to see the HEART's own replies: its Card, its Pulse and its alerts, exactly as the standard defines them.

Every demo is also one Python file you can download and run on your own computer: see Demo simulators.

Path 2: a HEART in Python, on HEART Hub#

You need Python 3.8 or later, and an email address.

1. Get the template. Download the Python template and unzip it. It is one file, heart.py.

2. Make it yours. Open heart.py and change the four lines marked CHANGE:

SYSTEM_ID = "org.example.my-product"     # a reverse-DNS name under a domain you own
NAME = "My Product"
VERSION = "0.1.0"
INSTANCE = "unit-1"                      # must not identify a person

Then write check_main_function(): return (True, "") while your program does its main job, and (False, "what is wrong, in plain words") when it does not. Leave it as it is for now to try the demo.

3. Turn reporting on. Set MONITOR = True. Reporting is off by default, because watching is the owner's choice.

4. Run it.

python3 heart.py

At first start it makes a device ID and a claim code, and prints them once:

HEART Hub: register this device at https://heartcortex.ai/register?id=3f9c0a71b2d4e865 with claim code K7Q2M-...

It keeps them in heart-monitor.json. Keep that file, and keep it private.

5. Register it. Open the address it printed. Sign in with your email address (a 6-digit code is emailed to you), enter the claim code, give the device a name, and press Register. The HEART Hub owner's manual shows every step with screenshots.

6. Watch it. Its page shows Healthy within 30 seconds.

7. Break it. Create a file named fault.flag next to heart.py. Within 30 seconds the page shows Failing, with the alert system.main-function-lost explained in plain words. Delete the file, and the alert is resolved.

8. Silence it. Stop the program. After three missed reports (at least 90 seconds) the page shows Not heard from, and HEART Hub emails you once.

Next: the Templates manual explains every option of the template. To ask the owner before a change, see ask_owner() in the owner's manual.

Path 3: choose a template for your product#

1. Read Choosing a template: 17 templates, from Python and Go to C on bare metal, three assembly languages, WebAssembly and the ESP32.
   • Server templates are for anything with an operating system.
   • Device templates are for microcontrollers.
2. For a microcontroller with sensors and a button, start from the ESP32 Full starter. It serves a signal table, explains alerts with a cloud model through a gateway, and carries out a physical command only after a press of the button on the device.
3. For a product that must approve changes, act on its own or use an AI brain, build on HEART Core, the reference runtime in Rust, and its daemon heartd (configuration).
4. See a whole product worked through in the product guides, such as the thermostat or the EV charger.
5. Measure how closely your product follows the standard with HEART-Rating, a score from 0 to 10.

Path 4: pick a reading track#

You do not have to read the whole standard. Start here asks four questions and gives you one of three tracks:

Each track lists the exact sections to read, in order, and where to stop.

Path 5: Docker with a HEART#

Docker Engine was given a HEART in two ways: rebuilt from source with the HEART inside its daemon, and as a sidecar beside the unchanged daemon. Both notice a restart loop, an out-of-memory kill and a failing health check within about 3 seconds.

• Docker with a HEART: what it adds, and how to run it.
• Replicate it: every step, with the exact versions and commands.
• Screenshots of a live run: Docker on heartcortex.ai, with Claude recommending more CPU and the owner changing it from the site.

Where next#

• All documentation
• The standard
• Downloads