How a result is formed
1. Safety comes first
Conditions involving sewage, active overflow, prior chemical use, electricity, or rapid water spread take priority over every ordinary rule. A hard stop removes incompatible DIY guidance.
2. Scope matters
One fixture, known separate levels, and direct cross-fixture reactions provide different kinds of evidence. Unknown levels are excluded from multi-level evidence, and a basement fixture is not assumed to use gravity drainage when pump topology is unknown.
3. Patterns are explicit and testable
The published checker uses deterministic rules stored with the site. The same valid answers produce the same result. A runtime language model does not invent or reorder the result.
4. Uncertainty stays visible
Results use “likely,” “possible,” or “not enough information.” DrainPath does not display a percentage unless that number comes from a relevant, cited dataset—which diagnostic outputs do not.
5. One urgency label is not enough
Every result keeps three decisions independent: whether there is a life-safety reason to leave, whether drain discharge should stop or be minimized, and when professional service should be arranged. A same-day service need does not automatically mean evacuation, and a life-safety alert does not erase secondary drain-control instructions.
What the tool uses
- Each observed fixture instance, its level, symptoms, and returned-water appearance
- Which fixture or appliance caused a different fixture to react
- Whether rain, an automatic drain device, a septic alarm, or a public signal was observed
- Odor type and strength, multiple exposure symptoms, and the kind of drain product used
- Water supply, pumped-drainage uncertainty, and shared-building context
What the tool does not use
- Camera or microphone access
- Exact location or a contractor bidding system
- A hidden confidence score
- Free-form AI diagnosis
- Stored personal symptom histories
Review status
How community cases are used
Public forum and field-service cases are used to discover missing combinations and build regression scenarios. They do not become safety authority by repetition. Safety actions and prohibited actions must still map to the official evidence register, while the rule wording retains jurisdiction and inspection limits.
Change control
Rules, sources, and content live in version control. Material changes require automated scenario tests, a source check, and a visible last-reviewed date. Safety rules take precedence over general classification rules.
Review the public evidence register to see each supported claim, original source, usage location, review date, and evidence limit.