HausPost

The next visit, from the tenant’s words.

A real Apertus 1.5 8B CPU run, followed by source review, a missing-unit question, and a coordinator’s repair handoff.

Actual browser recording with synthetic tenant correspondence and scripted coordinator actions. The CPU waiting period is shortened. The evaluation segment uses labelled, recorded model responses. English synthesized narration and subtitles.

Keep the four updates separate.

The kitchen tap is repaired. The doorbell still needs attention. Wednesday’s visit is cancelled. Friday is a tenant request awaiting confirmation.

Click each field to read its source, correct the model’s interpretation, and add missing details from your own information. The resulting brief carries those review and source labels into the handoff.

Explore the raw results and development checks

A small installation pilot for your office.

CHF 149

One office, one existing compatible host, six agreed short sample threads, and a reviewed handoff batch. Your team supplies the infrastructure and checks the maintenance decisions.

Blucca’s AI-led studio handles setup and the acceptance batch. Share your host setup and target date to agree scope and delivery before payment.

Discuss the installation pilot

Run the full self-managed stack.

make run provisions the pinned 5.06 GB model and starts the Python application with llama.cpp. Runtime inference stays between the organisation-managed containers. Deployment guide and host requirements.

Read the film transcript

Plain-text transcript