Methodology
Every narration in the corpus passes through four stages before it appears. Each stage is recorded separately, so what the platform asserts can be distinguished from what its sources assert.
The matn is transcribed from a printed edition, normalised, and linked to its place in the source book — its chapter, and where the edition records it, its page. The transcription preserves the vocalisation of the printed text; searching does not require it, because queries and stored text are both folded to a common form before matching.
The isnad is parsed into named transmitters and rebuilt as a directed graph rather than kept as prose. Where a narration reaches us by several routes, each route is retained as a separate chain, so branching is visible rather than flattened into a single line.
Each transmitter is a record in their own right, carrying kunya, nasab, and where known their dates and the region they worked in. The biographical literature’s judgement on their reliability is recorded as given, along with flags for the specific weaknesses the tradition tracks — tadlis, and loss of precision in later life.
Scholarly assessments are recorded in full and attributed to whoever made them. The platform does not collapse disagreement into a single verdict, and does not issue rulings of its own: where scholars differ, the difference is the finding.
This layer is the least complete part of the corpus. Assessments are still being gathered and attributed, and until that work is further along, hadith-level gradings are not displayed.
It does not adjudicate. It does not assign its own grade to a narration, and where a printed source is ambiguous the ambiguity is preserved rather than resolved silently. Corrections are welcome and are treated as data problems, not as disputes.