~/projects/istidlalREADME.md

AI knowledge system

Istidlal

A production research system that combines multilingual retrieval, source verification, and structured AI answers.

status
live
category
AI knowledge system
stack
Next.js · TypeScript · PostgreSQL · pgvector · hybrid retrieval

cat problem.md

The problem

Research across classical Arabic and Urdu sources is slow, fragmented, and difficult to verify. A useful AI layer must retrieve the right passages and keep every answer anchored to inspectable source material.

cat system.md

The system

Istidlal ingests, structures, indexes, and retrieves a large multilingual corpus. It combines semantic and full-text retrieval, reranking, citation assembly, and answer generation inside a production application with identity, subscriptions, and source-reading workflows.

~/projects/istidlalcapabilities.map

inspect --capabilities

What the system handles

  • Citation-grounded answers with inspectable sources
  • Arabic and Urdu semantic, full-text, and hybrid retrieval
  • PDF processing, OCR, transcription, and document ingestion
  • RAG pipelines, reranking, and structured citation output
  • Authentication, subscriptions, and production deployment

cat evidence.log

Evidence in production.

Hundreds of classical booksHundreds of thousands of indexed passagesMultilingual source retrievalLive production system