AXIOM

AI-assisted monitoring, scheduling, analytics, and control for an entire Klipper fleet—from the first layer to the final invoice.

Running a print farm looks like buying more printers. It is not.

Once you move beyond two or three machines, the hard part stops being the printing itself. The real work becomes everything around it: camera feeds, failed jobs, maintenance schedules, filament inventory, production queues, operating costs, customer deadlines, and the constant question of which machine needs attention next.

A print fails at 2 a.m. and turns a spool of filament and six hours of machine time into a bird's nest. You find out at breakfast. You become the scheduler, the quality inspector, the maintenance coordinator, and the accountant—usually across a pile of disconnected tools that do not share data.

AXIOM is a desktop operations platform that sits above a Klipper fleet and brings the whole operation into one interface. It watches prints through cameras, combines visual evidence with printer telemetry, schedules production, tracks maintenance, calculates costs, measures performance, and tells the operator what needs attention.

It talks to printers through Moonraker, making the platform brand-agnostic by design. The AxisBreak fleet is deliberately mixed. AXIOM does not care which company made the machine. If it runs Klipper and exposes Moonraker, AXIOM can work with it.

Most AI failure detection relies on one model and a confidence slider. That creates a dangerous cycle: the system cries wolf, users lose trust, they disable it, and then it misses the failure that mattered.

AXIOM approaches detection as a layered decision system. Multiple signals have to support the same conclusion before it acts.

Designed to Reduce False Positives

• Moving-head filtering recognizes the toolhead crossing the camera view.
• Telemetry nozzle masking maps nozzle position through guided calibration.
• First-layer adhesion checks compare commanded extrusion with visible material on the bed.
• Camera-health gating rejects blocked, fogged, dark, or unusable frames.
• Bed-warp detection separates growing lift from static lens distortion.

Bounded Corrections, Not Uncontrolled Retuning

When AXIOM identifies a correctable issue, it can make bounded changes to nozzle temperature, flow, speed, pressure advance, fan, or acceleration. A correction must be supported across multiple frames, is attempted only within defined limits, and is automatically reverted when the quality trend does not improve.

For a genuine catastrophic failure, AXIOM auto-pauses. Pausing is reversible. The system stops the damage first, then asks the operator to resume or cancel. If nobody responds overnight, a safety timeout can bring the printer down to a safe cold state.

AXIOM is built around fleet behavior. Polling fans out across an automatically sized worker pool, offline machines retry on backoff timers, and one disconnected printer does not stall the rest of the operation.

Each printer keeps an independent tuning and maintenance profile. A direct-drive machine and a Bowden machine have no business inheriting the same assumptions.

A queue of thirty jobs is not a plan. AXIOM can score each queued job using business value, expected profit per print hour, deadline pressure, and a manual priority adjustment for cases where the operator knows something the software does not.

The software does not become the manager. It explains its recommendation in plain language, while the operator remains in control of the decision.

This is where AXIOM stops looking like a printer utility and starts looking like an operations platform.

It tracks the true cost of a print using filament weight and price, electricity, machine time, maintenance allowances, and operator-defined margin targets. Pricing suggestions are just that—suggestions. The operator chooses the assumptions and desired profit.

The Morning After

Because the cameras, telemetry, costs, maintenance, and production queue share one data spine, AXIOM can surface relationships that single-purpose tools cannot see.

Each morning, the platform can produce a digest covering what ran overnight, what failed, what requires attention, which jobs should run next, and how much capacity remains.

Telegram notifications can deliver print completion, confirmed defects, catastrophic failures, low-filament warnings, suspected clogs, and the scheduled morning digest—with a camera snapshot when visual context matters.

AXIOM already operates as an agentic system within defined limits. It perceives the farm through cameras and Klipper telemetry, evaluates what is happening, recommends or takes bounded actions, and reports the result.

There is one part of the cycle it does not yet possess: hands.

The same design principles carry into the physical system. Every action must remain bounded, confirmed, and recoverable. A machine with hands has to earn even more trust than a machine with only eyes.

The operations layer runs on the AxisBreak fleet today. The physical automation comes next, followed by deeper robotic management and the larger print shop and community makerspace built around it.