A repair-knowledge database for home battery systems, tested before launch

Situation

Knowledge of what fails in a home battery system and what fixed it was spread across thousands of forum threads, repair videos and installers' memories. Manufacturers published little of this information, including what each alert code means. Owners whose systems were out of warranty, and the independent electricians they called, had nowhere to look it up.

Solution

The database records how home battery systems fail and what fixed them. AI read forum threads, manufacturer manuals and notices, a federal recall database and repair-video transcripts. It extracted the repair cases and organized them into failure modes and fixes.

Every statement carries its source, the date it was retrieved and the kind of claim it is. Each claim is marked as stated by the manufacturer, reported by owners or confirmed by several sources. A confidence score for each statement is calculated by fixed rules.

A public site, provenfixes.com, shows failure names, symptoms, alert codes and manufacturer documents. Repair partners receive the full entries and an AI lookup that ties each answer to its entry.

Design decisions

Conflicting reports are both kept and flagged. Jeff Hunt wrote down and dated the pass mark for each test before it ran. The test cases were kept apart from the material the database was built from.

Outcome

Owners and independent electricians can look up failures, symptoms and alert codes on the public site. Repair partners can see the documented fixes, each with its source. The design was settled by a test with a margin set in advance. An AI model reading the database answered more repair cases correctly than the same model working alone, by more than that margin.