AGENTIC & AUTONOMOUS SYSTEMS

Theme Submission for Hackathon 2026

---

Lex Agent

its an ai powered legal aid for anyone(for now only includes some indian laws regarding land disputes)

---

What is Lex Agent??

its a simple website where a user can explain there legal matter in plain english and agent help ordinary people understand their legal rights and explains it in plain english , for free...

How it works(for the user):

1. You explain your problem in english

2. The Agent classifies the legal matter(Case type, Legal Doamin, and severity)

3. Provides a simple summary

4. finds the law chunks from teh database

5. Provides a case Radiness Score and... and gives a evidence check list ... to imporve the case rediness score

6. Provides a recommended action plan

7. Provides some ai genrated notice and letters. which might come in handy

Scalable: if a the admin wants to add more laws the person needs to put the law documents in the backend/law_docs and run scripts/ingest_laws.py (currently only laws regarding land disputes are present in teh database)

How it works under the hood

The 4-Phase Pipeline(i know it sounds cool)

Your Description
       │
       ▼
┌─────────────────────┐
│  Phase 1: Classify   │  What type of case? Is the input clear enough?
│  (2 parallel LLMs)  │  If vague → returns "Insufficient data to proceed"
└─────────┬───────────┘
          │
          ▼
┌─────────────────────┐
│  Phase 2: RAG        │  Search law acts for relevant sections
│  Vector + BM25       │  using semantic search + keyword matching
└─────────┬───────────┘
          │
          ▼
┌─────────────────────┐
│  Phase 3: Generate   │  Generate evidence checklist, legal notice,
│  (4 parallel LLMs)   │  supporting docs, and action plan
└─────────┬───────────┘
          │
          ▼
┌─────────────────────┐
│  Phase 4: Assemble   │  Parse results, score readiness, build response
└─────────┬───────────┘
          │
          ▼
    Complete Case Analysis

[This was ai genrated(the flow chanrt i mean)]

Total Time: ~60-90 sec

Tech Stack

Layer	Technology
Frontend	Flask, TailwindCSS, Vanilla JS
Backend	FastAPI, Python 3.11
LLM	Google Gemini 2.0 Flash (via OpenRouter)
Embeddings	OpenAI text-embedding-3-small (via GitHub Models)
Database	Supabase (PostgreSQL + pgvector)
PDF Generation	ReportLab

Limitations

• Rate limits on free apis sometimes proivide delays

• Sometimes the backend server takes a long time wake up(cause i am using the free tier) (in the deployment)

• sometimes the application breaks (i have fixed it a bunch of times and henvent seen anything break but it might)

Future scope

• Multi languge support for indian languges

• voice transcibtion

• Conversational chat interface(like claude or other ai platforms)

• Case management and tracking(by adding auth and storiung rach report in the database if the user is logged in)

Ai usage

i did use ai to help me make this project but i know everthign happening in the project and can explain it if needed in future..... i designed the project on my on and i am positive that it is not considered as ai slop as there was actual thiniking that i neede to do i thought of the designed and stuff and know why a certain piece of code was implemented

License

This project is for educational and demonstration purposes.