How Lovguiden's data sources repair themselves
Public websites change without warning. Lovguiden's data collection is built to notice when that happens, let an AI agent diagnose the fault, and test the fix against the live source before it is applied.
Lovguiden collects Danish legislation, case law, parliamentary work and publications from more than 2,000 public sources. None of them are designed to be read by software. Courts, ministries and agencies redesign their websites, move documents and change formats, and every change can stop a data source from working.
A fixed schedule, day and night
Collection is run by a purpose-built system on local hardware. Around 40 jobs run on a fixed schedule, and before each run the system checks its connections, the database and storage. Jobs run in parallel within set limits and recover on their own after interruptions.
Noticing that something is wrong
The hardest faults are the quiet ones. A source that changes its layout rarely returns an error. It returns nothing, or the wrong thing. Every run is therefore logged, and a job that stops returning data is flagged, even when it technically succeeded.
Repair, then verification
When a source is flagged, an AI agent examines the failing job and the source, diagnoses the fault and prepares a fix. The fix is only applied after it has been tested against the live source. A fix that does not pass that test is not applied.
The numbers
Measured over the 30 days up to 28 September 2026:
- 2,093 runs
- 2,067 of them successful (98.8 %)
- 2.84 million items processed
- 41 active data sources
None of this is specific to legal data. Any process that depends on collecting information from sources outside one's own control, such as tenders, prices or regulatory publications, can be run the same way. More on tender monitoring.