The tools form five complementary families: retrieval & reading, hadith verification, exegesis & commentaries, the knowledge graph, and bibliography. All share one covenant: no fabrication in absence — when the corpus holds no text, it says so plainly; when it holds a near match, it declares the degree of similarity.
Targeted scoping — optional across the retrieval tools: the semantic search, hadith lookup, verification, exegesis, and scientific-explanation tools all accept — only when needed — narrowing parameters a specialist may employ; otherwise the search spans the whole corpus:
books restricts to specific titles, author to a specific author, and death_before_ah / death_after_ah confine results to an author death-year layer (AH) — e.g. only early authorities who died before 460 AH, or only later ones. These are levers in the researcher's hand, not query prerequisites; your AI assistant infers them from how you phrase the question — say "only from al-Kāfī" or "among the early authorities" and it sets them for you.1 · Retrieval & Reading
moheet_search_corpus — Semantic corpus search
Purpose: the main gate — "what do the sources say about X?" Retrieves the most relevant passages across all ingested books, even when your wording shares no vocabulary with the text.
Methodology: hybrid dual-engine search — multilingual semantic embeddings (Cohere embed-v4.0 over Qdrant) plus normalised Arabic lexical matching — fused by reciprocal rank (RRF), then re-ordered by a reranking model. Every result carries full citation with the editors' footnotes in place; for scanned books a vision agent reads the actual page image. Optional parameters tune result count, alternative phrasings, source diversity, category restriction — plus scoping to specific books, a specific author, or an author death-year range (AH): search only Wasāʾil al-Shīʿa, or only authors who died before 460 AH.
Researcher example: "What did al-Zahrāʾ (a) say when her inheritance was withheld?" — passages return from their loci with book and page; a question asked in Persian reaches Arabic sources and vice versa.
moheet_read_range — Sequential reading
Purpose: read consecutive pages of a specific book (up to 15 per request) — for context around a citation, or to browse an entire chapter.
Methodology: reassembles the book's passages in page order into continuous text, with heading chains and editors' footnotes in their positions — a faithful reading of the print, not a paraphrase.
Editor example: found your witness at "Fadak fī al-Tārīkh p. 116"? Request 115–117 to see the full argument before citing.
moheet_view_page — Original page image
Purpose: for scanned books: returns time-limited signed links (24 h) to the page image in three resolutions — display, thumbnail, archival original — with its machine-read excerpt.
Methodology: page images are archived at original resolution; the image itself is always the authority, the machine reading is auxiliary — stated explicitly in every output.
Editor example: citing a scanned source? Attach the page image beside your citation — visual documentation beyond dispute.
2 · Hadith Scholarship
moheet_hadith_lookup — Narration sourcing (takhrīj)
Purpose: give it a narration's text (or a distinctive fragment) and it returns the canonical text with full sourcing: book, volume, page, hadith number, chapter, edition — terse output built for critical-edition footnoting.
Methodology: search across the whole corpus with hadith books boosted (×1.15) — a narration may surface in history or theology works, and any result from outside the hadith genre is explicitly tagged with its genre. Lexical-match weighting and match-centred excerpts. A dual honesty threshold guards it: if the match score is weak or your distinctive words are absent from the candidate, it declares "no strong match" rather than dressing weak candidates as takhrīj. Lookup can also be scoped to specific books or a specific author. A
source_preference parameter serves editorial practice: earliest fronts the sources whose authors died earliest (classical takhrīj convention), latest the most recent — chronological re-ordering applies strictly within the strong-match band, so a weak match never outranks a strong one by age alone.Editor example: "Source: whoever dies on Friday is spared the torment of the grave" — returns Biḥār al-Anwār vol. 6 p. 230 no. 35, with chapter and edition.
moheet_hadith_verification — Verifying attributed texts
Purpose: present a text circulating as hadith and receive an explicit verdict: found / a similar text exists / not found in this wording — with the two closest matches and their scores — each match fully cited in MLA style: author, title, editor, edition, publisher, publication year, volume, page, hadith number and chapter where present — ready for a footnote.
Methodology: semantic-lexical matching with calibrated verdict thresholds, guarded by a numbers guard: if meaning agrees but numerals differ (a thousand prayers becomes a hundred), the verdict is downgraded with an explicit warning. Absence never condemns a narration — coverage is bounded by ingested books, and every answer says so.
Researcher example: received a viral attributed text? Submit it verbatim — if it distorts an original, you see the original and the exact point of distortion.
3 · Exegesis & Commentaries
moheet_quran_interpreter — Exegetical material for a verse
Purpose: give it a verse or fragment (surah optional) and it returns documented exegetical passages from ingested tafsīr works.
Methodology: search across the whole corpus with tafsīr works boosted (×1.1) — boost, not isolation: a narration in the hadith collections or a philological note may well interpret the verse, and any result from outside the tafsīr genre is explicitly tagged with its genre.
Example: "What is narrated on: This day I have perfected your religion" — returns exegetical loci by book, volume, page.
moheet_scientific_explanation — Scholarly commentary on questions
Purpose: give it a technical question or phrase and it returns explanatory passages from classified commentary works, with a facet line showing which books discuss it most.
Methodology: whole-corpus search with commentary (shurūḥ) works boosted (×1.1) and out-of-genre results tagged, plus a small-to-big expansion tail: every result carries a ready reference for reading its neighbourhood sequentially when needed.
Advanced student example: "Explain this passage of Kifāyat al-Uṣūl" — returns commentary loci grouped by work.
4 · Knowledge Graph
moheet_graph_explore — Relational exploration
Purpose: for relational questions text search cannot answer: how do two concepts relate? who critiqued whom? what does a position rest upon? Returns documented relation chains between concepts and figures.
Methodology: a knowledge graph (Neo4j) whose nodes and edges are machine-extracted from book passages, under one strict rule: no relation is shown without its verbatim evidence, source, and page.
Researcher example: "How does Fadak relate to the succession conflict?" — returns the chain Fadak → al-Zahrāʾ → opposes → the Caliphate, each edge with its literal evidence and page.
moheet_get_communities — Panoramic thematic survey
Purpose: before diving deep: what are the corpus's major themes around a field? Returns summaries of the thematic clusters connected to your question.
Methodology: automated community detection over the graph (GDS algorithms) groups densely connected concepts into clusters, each summarised with documentation and refreshed as the corpus grows.
Example — opening a study: "What themes surround the barzakh in the corpus?" — returns the major clusters; you choose where to zoom in.
5 · Bibliography
moheet_book_card — Bibliographic card
Purpose: the full card before you cite: title, author and death year, volume, editor, edition, publication year, publisher, classification, and indexed-passage count.
Methodology: source-file identity data, completed where absent by an AI cataloguing agent that reads the book's opening pages — so you know exactly which edition the corpus relies on.
Editor example: before referencing "Faḍl al-Kūfa wa-Masājidihā", request its card for your footnote: ed. al-Ṭurayḥī, Dār al-Murtaḍā, Beirut.
moheet_about — About the corpus
Purpose: answers "what is this service, its scope, its tools?" — a concise statement of content, method, and coverage bounds.
Composite Scenarios from Real Research
The critical editor footnoting: an unsourced narration in the manuscript →
hadith_lookup for takhrīj → read_range for context → book_card for edition data in the footnote → and if the source is scanned: view_page for an image filed with the edition.The theology researcher building a chapter:
get_communities to survey the field → search_corpus to gather texts from their loci → graph_explore to discover who relies on whom and who critiqued whom → then sequential reading of the pivotal passages.Verifying a circulating text: an attributed saying making the rounds →
hadith_verification returns the explicit verdict with its score, the nearest original if any, and the point of divergence if distorted.All tools are read-only — no writing, no deletion — accessed over the MCP standard with OAuth 2.1 authorisation. Get started with Google sign-in.