Dike logoDocs

Welcome to the Dike API Documentation. Dike is a bilingual (Arabic & English) legal intelligence API built for AI agents and enterprise LegalTech applications that must cite real, verifiable law, not hallucinations.

Rendering diagram...

Core Pillars


Why Dike?

Traditional LLMs hallucinate statutes, invent article numbers, and cite repealed legislation when asked complex legal questions in Arabic. Dike solves this with three structural guarantees:

  1. Deterministic Hallucination Guard: When you call POST /v1/reason, the model is forced to cite retrieved corpus documents using strict internal tokens ([[REF:...]]). Any token that does not resolve to an indexed, verified legal source is stripped before returning the answer to you.
  2. Hybrid Semantic + Lexical Retrieval: Legal queries require matching both conceptual phrasing (semantic vectors) and exact statute numbers or legal jargon (sparse lexical vectors). Dike fuses both with server-side Reciprocal Rank Fusion (RRF) and cross-encoder reranking.
  3. Canonical Legal Graph: Legal documents exist in a dynamic relationship web. Dike's graph engine (POST /v1/graph/traverse) tracks how laws amend, repeal, interpret, and implement one another over time.

Endpoints Overview

MethodEndpointDescriptionAccess
POST/v1/searchHybrid dense & sparse search with RRF fusion and rerankingFree
POST/v1/reasonGrounded legal Q&A with verified citations and streaming SSEPaid
POST/v1/citeArabic citation normalization and resolutionFree
GET/v1/document/{id}Full legal text, article structure, and metadataFree
POST/v1/graph/traverseMulti-hop legal graph traversal (amends, repeals, cites)Free
GET/v1/jurisdictionsDiscovery catalog for supported countries and courtsFree

Interactive Developer Console

Prefer testing live queries in the browser? Visit the Dike Developer Dashboard to test requests against your API keys with full parameter inspection.

On this page