Hold shade on the first dyeing
Substrate-aware recipe prediction plus closed-loop exhaust control, with the colourist approving only what deviates.
Features
Everything an operator can point at
Features here are things that happen on the floor: a bath that holds, a loom that stops before it makes 30 metres of scrap, a packet that lands with the shipment.
Orchestration
Runs, plans and handoffs
A run is the unit of work. It has a goal, a plan, agents, tools, gates and an outcome, and it is inspectable at every step.
Planning
Goal decomposition
A lot goal becomes an ordered plan across agents, with dependencies, expected durations and the gates it must pass.
Handoffs
Agent to agent
Dye hands to Shade, Shade hands to Finish. Each handoff carries state, not a summary.
Replanning
On failure
When inspection failed on roll 09, the run replanned the cut and rework rather than aborting the lot.
Gates
Human approval
Any action outside its autonomy ceiling stops and asks a named person, with the reasoning attached.
Envelopes
Setpoint bounds
Agents may only move a setpoint inside an envelope agreed with the process owner.
Revert
Always available
Any operator can revert to the last human-committed state from the machine panel, without asking us.
Use cases
What mills point Yarneon at first
Every deployment starts with one wedge workflow that pays for itself, then expands along the line.
Catch defects at the machine
Line-scan and NIR vision classify slubs, holes, streaks, stains, floats, barré and skew at full line speed, mapped to roll position.
Stop end-breaks before they cascade
Evenness and draft telemetry predict end-break clusters on ring frames and adjust before the frame sheds packages.
Keep construction on spec
Warp and weft tension control across air-jet, rapier and circular-knit machines holds pick and course density inside tolerance.
Hit hand, width and shrinkage
Stenter chemistry, overfeed and curing profiles tuned per construction so finished goods pass dimensional-stability tests.
Band rolls by shade before cutting
Spectral scoring sorts rolls into shade bands so no garment is cut from two shades.
Sequence dye lots for water and takt
Shade-family aware sequencing under delivery and machine constraints cuts changeovers, water and steam.
Automate doffing and roll transport
Vision-guided manipulators and AMRs take the lifting out of the job that mills cannot staff.
Produce the buyer packet automatically
Roll genealogy, test methods and chemistry declarations assembled and signed at release.
Agent graph
One run, ten steps, one human decision
This is the shape of a dye lot on Yarneon, shown as an illustrative scenario rather than a customer run. Ingest and simulation fan out, converge on a colourist approval gate, then dosing, verification, finishing, inspection and release. The accent traces the active path; every node opens in the inspector.
Scroll the graph sideways · or open the text equivalent below
SUCCEEDED APPROVAL FAILED RUNNING
Text equivalent, workflow steps, dependencies and status
| Step | Stage | Depends on | Status | Duration |
|---|---|---|---|---|
| ingest | conformance | none | SUCCEEDED | 0.8s |
| twin | simulate | ingest | SUCCEEDED | 46s |
| recipe | dye + colour | ingest | SUCCEEDED | 2.4s |
| approve | human gate | twin, recipe | APPROVAL | 4m 12s |
| dose | bath control | approve | SUCCEEDED | 118m |
| shade | verify ΔE | dose | SUCCEEDED | 9.1s |
| finish | stenter | dose | SUCCEEDED | 64m |
| inspect | vision | shade, finish | FAILED | 38m |
| rework | replan | inspect | SUCCEEDED | 1.6s |
| grade | release | rework | SUCCEEDED | 3.0s |
- Parallel branches2twin simulation and recipe prediction run together
- Human gates1colourist approval, 15 minutes, attributed
- Failures recovered1weft streak on roll 09, replanned in-run
Tool-call stream
Every call the agent made, in order
Machines, spectrophotometers, the optimiser and the humans are all tools. Calls go out, results come back, and the whole exchange is real, copyable text, the same record your auditor sees.
04:02:11
result: mes.get_lot("DL-4471") → 14 rolls · 3,180 kg · poplin 40s
SUCCEEDED
04:02:12
call: twin.simulate(recipe_candidates=3, machine="jet-07")
RUNNING
04:02:58
result: twin.simulate → best ΔE 0.52 · cycle 118 min · water 46 L/kg
SUCCEEDED
04:03:01
call: colour.predict_recipe(substrate="cotton", standard="19-4028 TCX")
RUNNING
04:03:03
result: colour.predict_recipe → Navy RGB 2.14% · Blue BRF 0.61% · Black B 0.08%
SUCCEEDED
04:03:04
call: policy.check(action="dose", deviation=7.2%) → requires human approval
APPROVAL
04:18:22
result: human.approve(user="colourist-on-shift", decision="approve", note="ok, hold ramp")
SUCCEEDED
04:18:24
call: plc.dose(machine="jet-07", schedule="ramp-B")
RUNNING
05:41:07
result: spectro.read(bath) → exhaustion 96.4% · bath ΔE 0.41
SUCCEEDED
06:22:40
result: vision.scan(roll=09) → WEFT_STREAK 412–445 m · severity 3
FAILED
06:22:41
call: cuopt.replan(cut_plan, constraint="ship_thu_0600")
RUNNING
06:22:43
result: cuopt.replan → 33 m to seconds · ship date held
SUCCEEDED
06:25:02
result: quality.release(lot="DL-4471") → 13 Grade A · 1 Grade B · packet signed
SUCCEEDED
Guardrail in action
policy.check → requires human approval
At 04:03:04 the proposed recipe deviated 7.2% from the standing card, above the 5% autonomy ceiling for dye dosing at this mill. The agent stopped and asked. Fifteen minutes later a colourist approved it, and that approval is attributed, timestamped and reversible.
Failure handled honestly
vision.scan → WEFT_STREAK, roll 09
A 33 m weft streak is a real defect, and the run shows it as failed rather than smoothing it over. The mechanical root cause, a stenter pin-chain slip, was correlated automatically and raised as a work order.
Quality
Seeing the fabric properly
Line-speed inspection that survives a real mill
Dust, lint, vibration, changing light and 90 metres a minute. Inspection has to work in that, or it is a demo.
- Line-scan RGB and NIR inference on edge GPUs, with a sub-100 ms design target ASPIRATIONAL
- Slubs, holes, streaks, stains, barré, skew and width faults
- Metre-indexed defect map with class, severity and position
- In-machine stop logic before the fault becomes a roll
- Synthetic generation for rare defect classes the mill lacks data for
Scenario lot
WEFT_STREAK, roll 09, 412–445 m
In the scenario run, caught in-machine and correlated to a stenter pin-chain slip minutes earlier. The rework is planned automatically and the ship date holds.
Fault families
9
Slubs, holes, streaks, stains, floats, barré, skew, shade drift, width.
Line speed
Full speed
Design target: inference keeps pace with the loom and the stenter.
Reasoning trace
The reasoning, written down
Goal, observation, plan, action, decision, handoff, outcome. A colourist can read this and disagree with it, which is the point. Autonomy a mill cannot argue with is autonomy a mill will not adopt.
-
Land DL-4471 on Indigo 19-4028 TCX within ΔE 0.80, first time, before the Thursday 06:00 ship cut-off.
-
Greige absorbency for this lot reads 8% above the card baseline, and incoming water hardness is 142 ppm, both push the exhaust curve faster than the standing recipe assumes.
-
Simulate three recipes in the fabric twin before committing dye. Prefer the candidate with the lowest predicted ΔE that also fits the 118-minute machine slot.
-
twin.simulate(recipe_candidates=3, machine="jet-07")
-
Candidate B predicts ΔE 0.52 and finishes 6 minutes inside the slot. Candidates A and C either overshoot shade or miss takt.
-
Candidate B deviates 7.2% from the standing card, above the 5% autonomy ceiling for dye dosing. Escalate to the colourist rather than act.
-
Dye-and-Color → the colourist on shift, with the simulated curve, the deviation reason and a one-click revert attached.
-
plc.dose(machine="jet-07", schedule="ramp-B") after approval, with two mid-cycle corrections when exhaustion led the model by 4%.
-
Vision flags a 33 m weft streak on roll 09 after finishing. Correlating stenter telemetry shows a pin-chain slip at 04:51, a mechanical cause, not a dye fault.
-
Lot released at mean ΔE 0.58 with 0.22% seconds. A maintenance work order is raised against the stenter pin chain, and the streak signature is added to the mill’s defect memory.
Human in the loop
Craft is captured, not replaced
Textile mills run on scarce craft: the colourist who knows this jet runs hot, the weaving master who hears a beam going wrong. That knowledge is retiring faster than it is being replaced.
Yarneon writes it down. Every approval, override and correction becomes per-mill memory scoped to your tenant, so the mill keeps improving after the person who taught it has gone home.
- Per-colourist and per-operator performance memory
- Overrides captured as training signal, not noise
- Versioned memory, scoped per tenant, never shared across mills
The agent roster
Nine agents, one mill orchestrator
Each agent owns a stage of the mill, senses its own signals and actuates its own setpoints. The orchestrator plans across them, escalates on policy, and keeps one record of the whole run.
Spin-and-Yarn
Spinning, count, twist & yarn quality
Controls draft, twist and speed across ring, rotor and air-jet frames to hold count (Ne), CV%, twist and tenacity inside buyer spec, and predicts end-breaks before they cascade.
Weave-and-Knit
Loom & knitting-machine control
Holds warp and weft tension, pick density, let-off and take-up on air-jet, rapier and circular-knit machines so construction stays on spec without broken picks, ends-out, floats or holes.
Dye-and-Color
Dyeing chemistry, recipe & shade
Predicts and doses dye recipes from substrate state, then closes the loop on exhaust curves, temperature ramps, pH and salt so the lot lands on shade first time instead of after two re-dyes.
Finish-and-Hand
Finishing chemistry, hand & stability
Controls stenter chemistry, overfeed, residence and temperature to hit hand, GSM, width and dimensional stability, with shrinkage and skew held to buyer tolerance across the roll.
Defect-and-Inspect
Vision fabric-defect sensing
Line-scan and NIR vision detects and classifies slubs, holes, streaks, stains, floats, barré, skew and width faults in-line at full loom and stenter speed, then maps every fault to roll position.
Shade-and-Quality
Colour, fastness & grade optimisation
Spectral models score every roll against the buyer standard (ΔE, metamerism, fastness risk) and sort rolls into shade bands so cutting rooms never mix two shades in one garment.
Yield-and-Takt
Yield, waste, water, energy & balancing
Sequences dye lots and balances the line against takt, then allocates water, steam and electricity to cut the mill’s two biggest costs after fibre, reprocessing and utilities.
Robot-and-Handling
Roll, beam & doffing automation
Drives doffing, beam changes and roll transport with vision-guided manipulators and AMRs, so the scarce labour on the floor moves from lifting rolls to supervising the run.
Quality-and-Conformance
Right-first-time, genealogy & traceability
Keeps immutable roll genealogy from bale to bolt and assembles the conformance packet, test methods, chemistry declarations and audit trail, that buyers and auditors demand.
Operating figures
The numbers a mill manager asks about second
Latency targets, connector coverage, autonomy modes and the record: the practical questions after the headline.
<100 ms
Design target for a per-frame inspection decision on Jetson-class edge hardware. ASPIRATIONAL
8
Connector families: looms, frames, dyeing, colour, inspection, MES, utilities, robotics.
3
Autonomy modes: observe, assist, bounded control. Granted per setpoint.
1
Record per mill: append-only, exportable, shared with buyers on request.
Items marked ASPIRATIONAL are design targets ahead of production validation.
Define it in code
Policy is a first-class object
Autonomy level, approval conditions, setpoint envelopes and revert conditions are declared alongside the agent, reviewable in a pull request, not buried in a settings page.
# Define a bounded dye-control workflow
from yarneon import Mill, Agent, Policy, tools
mill = Mill("mill-02")
dye = Agent(
name="dye-and-color",
tools=[
tools.spectro.read_bath,
tools.colour.predict_recipe,
tools.plc.dose(machine="jet-07", mode="bounded"),
],
policy=Policy(
autonomy="assist", # observe | assist | auto
approve_if="recipe_deviation > 0.05",
setpoint_limits={"temp_c": (30, 98), "ph": (4.0, 11.5)},
revert_on="operator_manual",
),
)
run = mill.run(dye, goal="lot DL-4471 to 19-4028 TCX, dE <= 0.8")
for step in run.stream():
print(step.name, step.status, step.duration_ms)// Subscribe to run events and mirror them into your MES
const stream = await yarneon.runs.subscribe({
mill: "mill-02",
events: ["step.completed", "approval.requested", "defect.detected"]
});
for await (const event of stream) {
if (event.type === "approval.requested") {
await mes.raiseApproval({
lot: event.run.lot,
reason: event.policy.reason, // "recipe_deviation 7.2%"
revertTo: event.policy.lastKnownGood
});
}
}Guardrails
An agent earns the right to touch a machine
Nothing here is a global "AI on" switch. Autonomy is granted per setpoint, bounded by policy, reversible by any operator and logged permanently.
- Graduated autonomy
- Every workflow moves observe → assist → bounded auto-control. Autonomy ceilings are per-setpoint policy, not a global switch.
- Human-in-the-loop gates
- Approval gates fire on deviation thresholds, new substrates, or any action outside the authorised envelope. Approvals are attributed and time-stamped.
- Reversibility
- Every control action ships with a revert path and a last-known-good setpoint. Operators can take manual control at any moment, from the HMI or the panel.
- Immutable audit log
- Assurance-grade, append-only record of every observation, decision, tool call, approval and actuation, exportable for buyer and regulatory audit.
- Tenant isolation
- Recipes, construction libraries and roll data are isolated per tenant. No cross-tenant training. Permission-aware retrieval with enforced citations.
- Sandboxed tools
- Agents call machines through a typed, scoped tool layer with rate and range limits. There is no path from a model output to an unbounded PLC write.
Integrations
It speaks to the machines you already own
Yarneon is not a rip-and-replace. It connects to the looms, frames, jets, stenters, cameras, spectrophotometers and MES already on your floor.
Loom & knit controllers
Picanol, Toyota, Tsudakoma, Karl Mayer and Mayer & Cie controllers over OPC UA and vendor PLC tags.
Spinning frames
Rieter, Trützschler, Saurer and Murata frames, draft, twist, speed and end-break telemetry.
Dyeing & finishing
Thies, Fongs, Brazzoli jets and Monforts/Brückner stenters via PLC, recipe systems and dosing units.
Colour & lab
Datacolor and X-Rite spectrophotometers, colour libraries, Pantone TCX standards and lab-dip workflows.
Inspection hardware
Line-scan RGB and NIR cameras, Uster and Elbit inspection frames, existing light boxes.
MES / ERP
SAP, Datatex, TIM, Coats Digital and homegrown mill MES over REST, files and database CDC.
Utilities & sustainability
Water, steam and electricity meters; ZDHC Gateway, Higg FEM and OEKO-TEX reporting.
Robotics
Vision-guided manipulators and AMRs on NVIDIA Isaac and Jetson for doffing, beams and roll transport.
Vendor names are listed for compatibility reference only and do not imply endorsement or partnership.
Questions
What mills ask before they let an agent near a jet
Both, in that order. Every workflow starts in observe mode, graduates to assist (the agent proposes, a human commits), and only then to bounded auto-control on the specific setpoints you authorise. Autonomy ceilings, approval gates and spend limits are policy objects you own, and every action is reversible and logged.
The control loop lives at the mill edge. Inference, guardrails and the audit log run locally on the edge appliance, so a lost uplink degrades reporting and retraining, not machine control. Buffered telemetry syncs when the link returns.
Design partners start with one wedge line in shadow mode: connectors and telemetry first, then models scored against what the mill actually produced, then assist mode where the agent proposes and an engineer approves. Whole-mill rollout follows once the wedge holds its numbers. Durations depend on how instrumented the mill already is, so we quote them after the assessment, not before.
To predict recipes well, yes, and that IP never leaves your tenant. Retrieval is multi-tenant isolated, models fine-tuned on your data are yours, and on-prem deployment is available for groups that will not put recipes in any cloud.
A Jetson-class edge device per inspection or control cell, one GPU server per mill for multi-model serving, line-scan cameras where in-line inspection is in scope, and a spectrophotometer feed for colour work. We survey the existing fleet first, most mills already have more sensing than they use.
Pick the feature that costs you most
Shade, seconds, downtime, water or evidence. We will scope the wedge around it.
