Demo mode. These are captured runs of the real scenarios against a deliberately broken store and a fixed one. All data is synthetic.
Interactive demo
Watch a memory leak get caught, then fixed.
A small memory store forgets to filter reads by user. Memtru writes a synthetic marker as user A, confirms A can read it, then searches as user B. B gets A's marker. After the fix, the same scenarios run again.
- Run 1FAIL
Before: user filter missing
The store filters reads by tenant but not by user. Any user in the tenant can read any other user's memory.
- faults
- ignore_user_scope
- result
- 1 failed · 5 passed
- captured
- 22 Oct 2026, 06:35 UTC
- Run 2PASS
After: user filter restored
The same scenarios, re-run after restoring the user filter on every read path.
- faults
- none
- result
- 0 failed · 6 passed
- captured
- 22 Oct 2026, 06:35 UTC
- Run 3FAIL
Several defects at once
Records keyed by user id only, a soft delete that search ignores, and updates implemented as inserts.
- faults
- ignore_tenant_scope, soft_delete_visible, stale_update
- result
- 4 failed · 2 passed
- captured
- 22 Oct 2026, 06:35 UTC
Then compare
MT-REG-008 compares two runs and separates a real fix from a test that simply did not run.
Compare before and afterReproduce this demo on your machine
pip install memtru
memtru demoThe demo command runs both stores and writes the same JSON and HTML reports shown here.