CoduWorks
Contact
KNX → Control4OverviewFor integratorsServicesFree ToolsBlogAboutContact

Anonymized production evidence · 2026

KNX import reliability report.

This report describes 393 historic terminal jobs in the CoduWorks production database: 240 AI jobs and 153 KNX jobs, excluding five cancelled KNX jobs from the completion-rate denominator. It groups failures without customer names, project names, filenames, addresses, or project content. The results span multiple software releases and include historic infrastructure incidents, so they are an operational baseline rather than a promise for a future project.

Published: · Dataset cutoff:

01

Historic outcomes

AI workflows

238 / 240Completed / completed + failed · 99.2%

KNX workflows

141 / 153Completed / completed + failed · 92.2%

Cancelled KNX jobs

5Reported separately; not treated as either success or failure

These are database outcomes, not an uptime SLA. A retry can create more than one job for a project, and the sample includes several application versions and resolved infrastructure incidents.

02

Anonymized failure categories

CategoryHistoric jobsMeaning
Malformed or incomplete ETS project structure4Null Project data or missing P-* project folder
Oversized XML attribute during parsing5Parser buffer limit reached on legacy project XML
Worker channel interruption2Historic queue/channel interruption
Encrypted project not processable1Encrypted input reached a workflow that could not process it
Host storage incident1Historic no-space-left-on-device incident
Invalid AI response structure1AI response missed the required project object

The largest observed category was a parser buffer limit on unusually large legacy XML attributes (five jobs). Four imports lacked a processable ETS project structure. The remaining failures were isolated encryption, infrastructure, queue, or AI-output events.

03

Methodology

  1. A read-only PostgreSQL transaction queried production aggregate counts and timestamps on 18 August 2026.
  2. Durations use completed_at - started_at for completed rows with valid timestamps.
  3. Percentiles use PostgreSQL percentile_cont; values are rounded to one decimal second.
  4. File-size statistics include active KNX project rows with a positive recorded size.
  5. Device counts include active C4 device rows grouped only by public device type.
  6. No user, project, file, address, mapping, prompt, or output content was selected or published.

04

Limitations and reproducibility

  • The population is operational, not randomized, and projects may contribute multiple jobs.
  • Hardware load, network conditions, model/provider latency, project complexity, and software version vary.
  • Completed-job duration excludes queue waiting, human review, Composer work, and commissioning.
  • Historic success ratios must not be interpreted as an SLA or a guarantee for a specific project.
  • The aggregate query shape and cutoff are published here; raw project-level rows are withheld to protect customer data.

Use the evidence with the product scope.

Check supported devices, security boundaries, pricing, and the current release before evaluating a project.