بسم الله الرحمن الرحيم
٦١٧٬٩١٢ حديث · ٢٦٬٩٢٣ راوٍ · ١٢٣ مصنفاً · كل إسناد مُحلَّل
A Computationally Graded Hadith Corpus — Every Grade Traceable to Its Inputs
617,912 narrations from 123 classical works (IPSC v4.0.0) — every chain of transmission parsed, narrators resolved where determinable, every grade derived from documented inputs.
Built in public · External review cycle & quality practice
The corpus layer for TheoAI™ — reasoned over by Eve-Theology™ F5/reasoner.
Provenance & Audit Practice, Published
The only AI hadith corpus that ships an authoritative _provenanceDisclosure block in the manifest with every record — AI-involvement scope, regression suite, NAQD adversarial framework, and the recurring external review cycle, all visible. Read the disclosure →
Chain Authentication
Every hadith's chain parsed into narrator positions, identified against 28,586 NRS entries, graded by weakest link.
Content Analysis
First-ever computational matn criticism. Quran contradictions, fabrication patterns, anachronistic vocabulary, prophetic linguistic baseline.
Cross-Collection Linking
8,170 canonical parallels (≥2-member matn clusters) and 52,938 near-duplicate (bag-of-words) clusters, plus semantic clusters at thresholds 0.85–0.92 (matn-core re-embedding rebuild), linking the same hadith across all 123 collections with attestation levels.
Universal Narrator Identity
26,923 narrators indexed; 21,617 with a reliability grading. Recurrences resolve to one ID where the evidence determines it; ambiguous homonyms are held rather than merged.
Transmission Network
889,913 directed edges mapping 1,400 years of who-learned-from-whom.
Grade Provenance
Every grade traces to a specific narrator tier, a Taqrib entry, and a documented grading rule.
Taqwiyah Independence
Supporting chains verified for genuine independence. T10+ narrators cannot be elevated.
Fabrication Detection
NAQD-3 matn-core re-embedding re-run: 1,851 findings — 300 critical, 638 high. 6,822 records from fabrication-focused reference collections evaluated. Source-collection caps applied to 2,168 records in fabrication anthologies (Mawḍūʿāt, Tanzīh, Fawāʿid).
Search 617,912 narrations. Open any record at a permanent, citable URL.
Six JSONL files, or a tiered HTTP API with provenance on every response.
License under Research or Scholar terms, against a versioned corpus of record.
Tier separation is the central design decision of the IPSC API. Each tier is a physically separate index — the Public surface does not contain Scholar-tier fields at all, so field-exposure cannot be accidentally widened by a query parameter or a misconfigured client.
Public
Attribution-only license, throttled rate limit
Research
Paid commercial subscription
Scholar
Identity-verified reviewer agreement, no redistribution