VIDHAAN AI

Legal AI Platform

Advanced Hybrid RAG Pipeline

Empowering Indian Statutory & Constitutional Research

An institutional-grade legal workbench mapping the entire codification landscape of Indian statutory law. Grounded directly in verified legislative drafts with zero disclaimers or AI preamble fluff.

View Database Ingestion Scope

Real-Time Index Status

113Acts Fully Ingested
1,133Legal Sections
5,233Vector Chunks
4Key Departments
Core AI Model & Engine Architecture
LLM Inference Nodellama-3.3-70b-versatile (Gemini Flash fallback)
Dense Embedding Modelgemini-embedding-2 (768 Dimensions)
Retrieval PipelineReciprocal Rank Fusion (Dense RAG + ts_rank)
Database Sync: 100% Complete Local Latency: < 85ms

Architected for Legal Professional Precision

Explore key systems driving Vidhaan AI's search speed, grounded citations, and sovereign statutory safety.

SYSTEM FEATURE 1

113 Acts Ingested & Streamed

Vidhaan AI maps 113 key statutory acts and legal archives from the Legislative Department, running with 1,133 pages and 5,233 chunks in active memory.

  • Constitution of India active indexing
  • Indian Contract Act & statutory extensions
  • Modernized criminal law acts and sections
STATUTECHUNKS
Constitution of India1,822 Chunks
Indian Contract Act, 1872912 Chunks
Bharatiya Nyaya Sanhita, 2023832 Chunks
SYSTEM FEATURE 2

Authoritative RAG Grounding

Our proprietary Reciprocal Rank Fusion combines sparse keyword scanning and semantic dense embeddings to retrieve sections with laser accuracy.

  • No conversational preamble pre-processing
  • Displays primary PDF source hypertexts in bubbles
  • Links section detail drawer dynamically
Retrieved Parent ContextSection 124, Indian Contract Act

"A contract by which one party promises to save the other from loss caused to him..."

Dense Embedding Match: 98%RFF Rank: #1
SYSTEM FEATURE 3

Sovereign Digital Standards

Designed using premium saffron and Navy Blue tricolor styling, reminiscent of official digital platform guidelines of India.

  • Light-mode high contrast reading canvas
  • Elegant Scales emblem layout
  • Fully compliant MVP architecture
Scales of JusticeNational Legal Workbench

Covered Departments & Legislative Scope

We actively ingest, tag, and structure statutory documents across key constitutional and judicial bodies in India.

Constitution of India

Complete articles covering parts, schedules, fundamental rights, directive principles, and critical constitutional amendments.

Department of Justice

Rules, legal directives, organizational structures, judicial appointments, and legal administration statutes of India.

Dept of Legal Affairs

Ingested treaties, litigation reports, contracts guidance, and statutory frameworks governing international & domestic arbitration.

Legislative Department

113 active acts, statutory rules, regulations, codifications, and state-wise gazette adjustments kept up-to-date.

Technical Architecture

Sovereign RAG Search & Notebook Pipeline

How statutory legal datasets flow from ingestion to dense-sparse retrieval, AI synthesis, and case briefing.

1. Ingest & Index

Statutory Scraping & Extraction

Authoritative Indian legislative acts and constitution articles are parsed, cleaned of OCR noise, and indexed.

Raw Statutory Acts

Vector Space Embeddings

Documents are split into semantic chunks and embedded into 768-dimensional space using gemini-embedding models.

768-Dim Dense Vectors

Sovereign Database Storage

Saves text chunks, vector values, and user session profiles securely to the primary PostgreSQL (pgvector) database node.

PostgreSQL DB Node
2. Query & Retrieve

System Prompting & Routing

Directs inquiries through strict legal-focused system instructions, routing to RAG lookup or Direct LLM modes.

Strict System Prompts

Hybrid Dense-Sparse RAG

Combines cosine vector similarity (dense) and BM25 text match (sparse) using Reciprocal Rank Fusion (RRF) algorithms.

RRF Rank Merger

Statutory Verification

Extracted statutory chunks are cross-referenced to verify relevance and ground truth before sending to LLM.

Anti-Hallucination Guard
3. Synthesize & Edit

Grounded LLM Synthesis

Llama-3.3-70b-versatile processes the verified context, generating detailed legal insights containing exact footnoted acts.

Llama 70B Engine

Streaming Interface Response

Token streams are rendered in real-time on the Next.js frontend chat workspace with clickable citation reference drawers.

Next.js SSE Stream

Case pad & My Notebook

Researchers pin cited statutory references to “My Notebook” to write drafts, add custom notes, and export brief files.

Brief Compiler PAD
Evolutionary Roadmap

Release Status & Journey

Detailing the key technical milestones driving Vidhaan AI's legal intelligence evolution.

v0.1 ReleaseINITIAL MVP

Core Retrieval & Search

Established the foundational statutory RAG pipeline and vector structures.

  • Indian Constitution & Acts Ingestion
  • Dense-Sparse Hybrid Vector Indexing
  • Dual Chat Interface (Direct LLM vs RAG)
  • Interactive Dark-Mode Dashboard UI
Status: Deployed & Stable
ACTIVE RELEASE
v0.2 ReleaseDEPLOYED

Mobilization & Workspace

Enhanced user collaboration, layout efficiency, and PWA packaging.

  • Sovereign Fullscreen Mobile PWA
  • “My Notebook” Pinned Citations Panel
  • Widescreen Layout Expansion (max-w-5xl)
  • Refined 3D App Icon & UI Layout Cleanup
Status: Deployed & Active