Fewer claims.
Things to open.

From ticket
to job sheet.

A local program turns fictional tickets into job sheets, then into a table. The columns can be changed; the CSV table and the settings can be downloaded. Missing information stays visible.

• OpenThe transformation rules · A sample job sheet
• What works hereThe text is read, the fields are checked and the table can be exported, without calling an external service.
• What this example does not demonstrateA diagnosis, the quality of an AI model or a connection to a client’s tool.

Adapt a model.
Export it for local use.

The lab carried out a LoRA fine-tuning of an open model with 0.5 billion parameters, then exported it to run locally.

• The available recordsTraining measurements, export report and a check of where the model’s parameters come from.
• The data for this trial73 internal examples. A small technical experiment, not a representative business dataset.
• What remains to be verifiedThe quality gain and the generalization are not established. A robust independent comparison is missing.

Prepare the switch
from one model to another.

Code prepares the switch from an external AI service to a local model, depending on the error encountered. This fallback code was tested and loaded by the system.

• What existsAn internal record dated 14 August 2026: schemas, tests and loading.
• The important limitThe validation did not call any external model. It does not prove a real end-to-end switch to the cloud.
• The next useful testReplay the task and its interruptions with the selected models.