PakJuris

Purpose

PakJuris is an agentic RAG system for Pakistan’s legal corpus. It brings together a React + TypeScript web app, a Node/Express API, and scraping/vectorization scripts that ingest legal documents into a searchable retrieval pipeline.

Features

1. Legal document search and retrieval for Pakistan-focused legal material.
2. Scrapers for downloading, processing, and organizing source documents.
3. Vectorization workflows for semantic search over the legal corpus.
4. Agentic API layer using LangGraph/LangChain-style tooling and document-management endpoints.
5. A user-facing web interface for exploring and querying legal content.

Lesson Learnt

1. Data Layer Matters - Building a legal RAG pipeline starts with the data layer: scraping, cleaning, naming, storing, and updating documents matters as much as the model call.
2. Chunking Strategies - Large legal corpora need different chunking strategies for different document types, especially when page boundaries, sections, and citations carry meaning.
3. Agentic Guardrails - Agentic workflows are useful for document management, but they need guardrails so search, retrieval, and summarization stay grounded in the source material.