JusticeGuide

IPC-focused legal RAG assistant with citation-backed answers, Next.js UI, and Ragas evaluation.

JusticeGuide helps users explore Indian Penal Code (IPC) topics through grounded Q&A. It retrieves from a legal corpus, generates answers with citations, and includes evaluation and observability hooks for answer quality.

Disclaimer: Educational / research assistant. Not legal advice.

---

Highlights

• RAG pipeline: query enhancement → FAISS retrieval (LlamaIndex) → cited answer generation
• Models: Gemini primary, Groq fallback
• Frontend: Next.js app in apps/web
• Quality: Ragas eval pipeline + optional Arize Phoenix tracing
• Ingestion: PDF / Markdown / TXT corpus import CLI
• Deployable backend: Docker + uv workflow

---

Architecture

Next.js (apps/web)
      │
      ▼
POST /api/ask  ──►  FastAPI backend (src/)
                         │
                         ├─ query enhancement (LCEL)
                         ├─ FAISS retriever (dataset/corpus)
                         └─ answer generation + citations

---

Tech stack

Layer	Tools
Backend	Python, FastAPI, LangChain, LlamaIndex, FAISS, sentence-transformers
Frontend	Next.js, React, TypeScript
LLMs	Gemini, Groq
Eval / observability	Ragas, Arize Phoenix
Tooling	uv, Docker, pytest

---

Quick start

Prerequisites

• Python 3.11–3.13
• uv
• Node.js 20+
• GEMINI_API_KEY (and optional GROQ_API_KEY)

Backend

git clone https://github.com/Znaxh/Justice-Guide.git
cd Justice-Guide
cp .env.example .env   # add API keys
uv sync --all-groups
uv run uvicorn src.main:app --host 0.0.0.0 --port 8000

Health check: http://localhost:8000/api/health

Frontend

cd apps/web
npm install
npm run dev

Set NEXT_PUBLIC_API_BASE_URL=http://localhost:8000 if needed.

Default UI: http://localhost:3000

---

Corpus ingestion

Supported formats under dataset/ (recursive): .pdf, .md, .txt

# Import a folder of legal docs
uv run python -m src.ingest --from-dir /path/to/legal-docs

# Import specific files and rebuild index
uv run python -m src.ingest ./notes/section302.md --rebuild-index

---

Core API endpoints

Method	Endpoint	Description
POST	/api/ask	Ask a legal question
GET	/api/health	Health check
POST	/api/admin/rebuild-index	Rebuild FAISS index (admin key)
POST	/api/admin/run-evals	Run Ragas evaluation job (admin key)

---

Project structure

Justice-Guide/
├── apps/web/          # Next.js frontend
├── src/               # FastAPI + RAG backend
├── dataset/corpus/    # Seed legal corpus
├── data/              # Eval datasets and results
├── tests/
├── Dockerfile
└── docker-compose.yml

---

Evaluations & tests

uv run python -m src.evals.eval_pipeline
SKIP_PHOENIX=1 uv run pytest

---

Docker

docker compose up --build

API: http://localhost:8000

---

Links

• Live demo: https://lawhelpai.netlify.app/
• Author: Znaxh

License

MIT — see LICENSE.