The plant, addressable
Treadum exposes the tire, the batch, the component, the cure trace, the finding and the action as first-class API objects — with a factory-edge SDK for custom models and connectors, and a control-envelope schema you can review line by line.
from treadum import Site, Envelope
site = Site("plant-nl-02", mode="shadow")
# stream findings as they are produced at the station
for finding in site.findings.stream(station="xray-2"):
if finding.confidence > 0.9:
tire = site.tires.get(finding.tire_id)
print(tire.batch_id, tire.builder, finding.family)
# propose a bounded action (blocked in shadow mode)
site.actions.propose(
id="build.drum_tension", target="drum-2", delta_pct=1.2,
evidence=[finding.id],
)Protocols and runtimes we speak
Four ways to build on Treadum
REST and streaming API
Query tires, batches, components, cure traces, findings and actions. Subscribe to station streams over gRPC or websockets with at-least-once delivery.
Factory-edge SDK
Python and C++ SDK for deploying your own models into the edge runtime, with the same telemetry, versioning and rollback machinery as ours.
Connector framework
Write a connector for an unusual machine in a manifest plus a small adapter. Connectors are sandboxed and cannot write outside the envelope.
Webhooks and exports
Push holds, escalations, NCRs and shift reports into your MES, QMS or ticketing system, with signed payloads and replay.
What the integration actually looks like
Trace a tire in one call
Every entity is linked, so a trace is a traversal rather than a join across four systems.
tire = site.tires.get("8841236")
print(tire.batch_id) # 4468
print(tire.builder, tire.drum)
print(tire.cure_trace.press, tire.cure_trace.state)
for f in tire.findings:
print(f.family, f.confidence, f.localized_to)
scope = site.trace.forward(batch="4468")
print(len(scope.tires)) # 612Bring your own detector
Site data science teams can deploy their own models into the edge runtime and have them gated by the same evaluation pipeline as ours.
Talk to the machine nobody supports
A manifest plus a small adapter is enough for most legacy assets. Connectors are sandboxed and read-only unless the envelope grants otherwise.
name: "legacy-uniformity-machine"
kind: file-drop
watch: "/mnt/uni/out/*.csv"
parse:
tire_id: "col:0"
rfv_kgf: "col:4"
conicity: "col:7"
emit: "uniformity.measurement"
write_access: noneThe safety contract is a file
The control envelope is versioned, signed and reviewable in a pull request like any other critical config.
{
"site": "plant-nl-02",
"version": 7,
"signed_by": ["process.eng", "safety.ehs", "ops.dir"],
"actions": [
{ "id": "cure.dwell", "delta_s": [-12, 12], "auto": true },
{ "id": "build.drum_tension", "delta_pct": [-2, 2], "auto": true }
],
"deny": ["safety.*", "interlock.*"]
}Core resources
| Resource | Methods | Streaming | Notes |
|---|---|---|---|
/v1/tires | GET, list, trace | Yes | Genealogy traversal from any linked entity |
/v1/batches | GET, list | Yes | Mix cycle telemetry and derived rheology estimate |
/v1/components | GET, list | Yes | Extrusion and calendering rolls with gauge history |
/v1/cure-traces | GET, list | Yes | Press and mold telemetry plus cure-state estimate |
/v1/findings | GET, list | Yes | Defect findings with per-modality confidence |
/v1/actions | GET, propose, approve | Yes | Envelope-checked; propose is blocked in shadow mode |
/v1/envelopes | GET, list, diff | No | Read-only via API; changes require signed review |
/v1/reports | GET, export | No | Shift, quality and audit exports |
Develop without touching production
A plant on your laptop
The Treadum simulator replays recorded or synthetic production so you can build connectors, models and integrations against realistic streams without any line access — including rare defect families you would otherwise wait months to see.
- Replay recorded site data under your access controls
- Synthetic rare-defect streams from the same generator used in training
- Deterministic seeds for reproducible tests
- Same API surface as production
Engineering teams on building with Treadum
“We had three people watching X-ray images and still shipped uniformity rejects. Treadum flagged the cord shift at the building drum, not four hours later at final inspection.”
“The cure agent found 40 seconds of margin on a construction we had run the same way for eleven years. It proved it in shadow mode before it touched a press.”
“Genealogy is the part I did not expect to care about. Every tire now has a linked record from batch to grade, and audit prep went from weeks to an afternoon.”
[PLACEHOLDER] Design-partner quotes are illustrative until pilot references are published.
What the API guarantees
Building on the platform
Yes. The edge SDK accepts your models, and they are gated by the same golden-dataset evaluation and rollback machinery as ours. Several sites run a mix.
Yes — the simulator replays recorded or synthetic production with the same API surface as a live site, so integration work never needs line time.
The v1 API carries a 12-month deprecation notice for breaking changes, and site deployments pin an API version explicitly so an upgrade is a decision rather than an event.
Only by proposing actions, which are then evaluated against the signed control envelope. There is no API path that bypasses the envelope — that is deliberate and permanent.
Put one cell on autonomy in 90 days
Pick one wedge — splice inspection, X-ray defect detection, cure-state optimization or uniformity prediction. We baseline it, run shadow mode, then graduate to bounded autonomy under a signed control envelope.
Pilots start in shadow mode. No line changes until accuracy and safety gates pass.