AnalAIser turns PostgreSQL-stored IoT data into styled, interactive HTML reports, published into per-customer Nextcloud folders. Design a template in chat, graduate it into a reproducible package, then run it against any workspace with the same sensors.
Three steps from sensor data to a published report. Start at Step 1 for any new report; jump straight to a later step for an existing one.
Pick a customer, sensor types and date range, then chat with the LLM to design a styled HTML template. Iterate on the live preview until the design is right — no numbers are baked in, so the template works for any compatible workspace later.
Open Builder →Convert a dev session preview into a reproducible 5-file package — manifest, Jinja template, SQL transforms, Plotly chart specs and residual narrative prompts. The LLM draws the structure once; real aggregates do the work at run time.
Browse packages →Run a graduated package against any workspace with the same sensor shape, choosing date range, bucket size and step at run time. Outputs go to Nextcloud, email and the audit log. (Automated scheduling UI is coming — manual Run for now.)
Browse & run →| App | running |
|---|---|
| Metadata DB | connected |
| IoT DB (read-only) | connected |
| AWS Bedrock | configured |