LLaMA-3.1-70B · Groq LPU · pgvector RAG · v3.2

Indian Legal Research
Grounded in Sources

AI co-pilot for Indian law — citations verified against retrieved documents, hallucination risk scored, structured output on every response. Not a law firm. A research engine.

Free plan available · No credit card · Upgrade anytime

50K+

Legal chunks indexed

4 Cr+

Judgments available

< 12s

Typical response time

Free

To start today

Built for Serious Legal Research

Every component designed to minimize hallucination risk and maximize citation trust.

Anti-hallucination

Source-Gated Citations

Every precedent verified against retrieved documents. Hallucinated references flagged with source_verified: false — not silently presented as facts.

Fast

Groq LPU Inference

LLaMA-3.1-70B running on Groq's Language Processing Unit. Typical response in 5–12 seconds — not 60+ seconds of a GPU cluster.

Accurate

Hybrid RAG Retrieval

FAISS dense + BM25 sparse retrieval fused 70/30. Captures both semantic meaning and exact legal citation strings.

Secure

Enterprise Security

Input sanitization, prompt injection detection, per-user rate limiting, SHA-256 hashed audit logs. GDPR-compliant by design.

New

Document Analysis

Upload PDFs or images of judgments, FIRs, contracts. AI extracts parties, sections cited, key holdings, and runs full legal analysis.

Transparent

Confidence Scoring

Every response scored: citation_validation_score, hallucination_risk, confidence_level. Know exactly how much to trust each answer.

How It Works

From question to validated answer in seconds.

01

Ask in plain English

Type your legal question — anticipatory bail test, cheque dishonour procedure, landmark judgments on Article 21.

02

Hybrid retrieval

NyayaVeda searches 50,000+ legal chunks from Supreme Court, High Courts, and Indian statutes using dense + sparse retrieval.

03

LLM synthesizes + validates

LLaMA-3.1-70B synthesizes an answer grounded in retrieved sources. Each citation is verified against the retrieval context.

04

Structured, scored output

You get statutes, precedents, analysis, and confidence scores — all in a structured format you can act on.

Validated output

Structured Output, Every Time

Every response validated against a Pydantic schema. Disclaimer enforced by the system prompt — cannot be overridden by the LLM.

  • citation_validation_score0.0 – 1.0 float
  • hallucination_riskLow | Medium | High
  • source_verifiedtrue only if found in context
  • disclaimermandatory on every response
response.json
{
  "issue": "Test for anticipatory bail
    under Section 438 CrPC",
  "confidence_level": "High",
  "citation_validation_score": 0.95,
  "hallucination_risk": "Low",
  "relevant_law": [{
    "act_name": "Code of Criminal Procedure",
    "section": "438",
    "summary": "Anticipatory bail..."
  }],
  "precedents": [{
    "case_name": "Gurbaksh Singh Sibbia
      v. State of Punjab",
    "year": "1980",
    "court": "Supreme Court",
    "source_verified": true
  }],
  "disclaimer": "Educational purposes only..."
}

Legal Corpus

50,000+ chunks from authoritative Indian legal sources. Scrapes Indian Kanoon weekly for fresh judgments.

Supreme Court of India
High Courts (24 states)
IPC / BNS 2023
CrPC / BNSS 2023
Indian Evidence Act
Constitution of India
POCSO Act
Prevention of Corruption Act

Document Analysis

New in v3.1

Upload PDFs or images. AI extracts structured metadata and runs full legal analysis.

Judgment PDFs → parties, holdings, sections cited
FIR images → charge details, sections invoked
Contracts → parties, obligations, risk clauses
Legal notices → deadline extraction, compliance check

🇮🇳 Hindi support

Set EMBEDDING_LANG=multi in your .env to enable Hindi query support via multilingual embeddings. No GPU required. Zero additional cost.

Designed for Indian Legal Practice

Every feature built around how Indian lawyers actually work.

Advocates

Research bail applications, precedents, and statutory interpretations in minutes instead of hours.

Law Students

Understand landmark judgments, trace doctrinal evolution, and cite correctly for moots.

Legal Teams

Team plan with high daily quota, custom corpus, and API access for integration.

Start Researching Today

Free plan available, no credit card required. Upgrade for more queries. Enterprise plan with API access and custom corpus.