Evidence before confidence.

Source order

We prefer manufacturer materials, public institutions and original parts documents. Specialist collections and repair references can add context. Commercial listings, community posts and seller statements may help discover a lead, but they do not automatically confirm a fitment or model merge.

What we publish

A documented part row is tied to a specific source page and locator. A shared OEM number requires separate supporting model documents. Source type, document edition, variant conditions and review state are retained when available. We do not treat a market label as proof of a shared mechanical platform.

What we do not infer

We do not upgrade a seller’s “fits” statement to confirmed compatibility, or combine similar model names without evidence. Electrical parts, gears, precision or tolerance-sensitive parts, and modified components require a separate safety review before any compatibility claim can be confirmed.

Uncertainty labels

When evidence is incomplete, the record remains candidate, unknown or conflicted. These labels are a feature of the research record, not placeholders for a guess. A missing part row does not mean the part does not exist; it may mean the available evidence has not established it.

Review and corrections

Research is checked against original pages where possible, with rule-based validation and independent review for the published evidence slice. When a source, transcription, model boundary or evidence interpretation proves inaccurate, the record should be corrected, narrowed, removed or returned to an uncertainty state rather than silently overstated.

Copyright and source access

The atlas does not publicly rehost source PDFs unless redistribution rights have been positively reviewed. It links to sources and stores structured research facts so readers can inspect provenance without treating the site as a document archive.