About fihrist
fihrist — فهرست — is where the sources are found. It is one part of IlmFlow, an effort to make sourced answers about the Shia tradition, faithful to their sources, available to people who currently get neither. The word means the index, the finding aid: fihrist finds the page in a library of Arabic and Persian books so that a person can open it and read it. It does not, and will not, produce text you can quote.
The problem it exists for
A citation is a claim that a page exists and says a thing. Most citations in circulation were never checked against the page: they were copied from another citation, or remembered, or produced by a machine that had learned what a citation looks like. The library behind fihrist is some nineteen gigabytes of the tradition's own books — hadith collections, tafsir, fiqh, history, the words of the Imams — and finding the page in them used to mean opening volumes one at a time. Now it takes a moment. What it does not do is the other half of the work, on purpose.
What this is not
These are constraints built into the tool, not disclaimers added afterwards.
- It does not give you text to quote.
- Everything a machine read out of a page is marked 🔴 not quotable, on every result, in every payload, to every assistant. Only a person who has opened the rendered page may quote it — and the tool records who that person was.
- It does not judge a hadith.
- It finds where a text is; it says nothing about its chain, its grading or its authority. That is scholarship, and it belongs to the scholar reading the page.
- It does not issue rulings.
- For any binding ruling it points you, as daftar does, to your own marjaʿ.
- It does not replace reading the book.
- The snippet under a hit is a signpost to recognise the page by. It is often slightly wrong — a letter here, a word there — and the page is the source.
- It does not send your work anywhere.
- The library, the index, the searches and the pages live on our own machines. No text leaves them to be read by a third party's model; the estate's own models run inside it.
- It does not track you.
- No analytics, no telemetry, no third-party scripts, no advertising. A session cookie, so that you stay signed in, and an audit log of who was admitted — nothing else.
What one result is
A locus: which book, which volume, which page — with the page number as it is printed in the edition — and the words that found it. The page renders as an image, exactly as printed, for you to read. A result is a signpost to a page, never a quotation of it.
Why a person always opens the page
If a citation is produced and nobody opens the book, what is learned is not a wrong page number. What is learned is that a confident-looking citation is what an answer is — and that habit spreads to every book never seen. So the tool is built so that the only way to get quotable text out of it is for a named person to have read the page. That is the whole product, and it is the part most easily undone by a shortcut.
What we cannot do ourselves
The technical work is ours. Naming the books — which edition, which work, which volume — deciding what belongs in the library, and reading a page to vouch for what it says: none of that can be substituted with engineering. It needs readers who know the tradition and its editions, and it goes at their pace.
Taking part
fihrist is for the IlmFlow team. There is no sign-up: you sign in with your daftar account, and a fihrist admin adds you to its list. If you have a daftar account and this is your first time here, or you think this is for you, tell us who you are.