Predict outcomes.
Entirely off the grid.

Writ is designed to predict potential outcomes of your case or legal issue using previous judgments and applicable Central and State acts. A fine-tuned language model and hybrid retrieval ground each prediction in an encrypted on-device knowledge base. No account. No cloud. Queries never leave your machine.

Download for Free View on GitHub
100%Offline
~2.0 GBModel size
3 PlatformsWin / Mac / Linux
0 AccountsRequired

What's inside the framework.

Every architectural decision prioritises privacy, depth, and precision over convenience and cloud dependency.

Fully On-Device

The complete inference pipeline — model loading, retrieval, generation, and re-ranking — runs on your local hardware via Ollama. Nothing is transmitted to any external server. Your research stays yours.

Encrypted Knowledge Base

Judgments, acts, and curated legal notes stored encrypted on-device. No plaintext exposure at rest.

Hybrid Retrieval

BM25 keyword search combined with neural re-ranking surfaces the most relevant passages, not just the most frequent matches.

Vidhi — Fine-tuned on Indian Law

Writ runs Vidhi (Llama 3.2 3B fine-tune), a model trained specifically on Indian legal texts. It understands the structure of judgments, identifies precedents, and reasons about statutory provisions in context.

Case Synopsis → Predicted Outcome

Paste a case synopsis and get a predicted judgement grounded in previous High Court / Supreme Court judgments and State / Central acts — predicted outcome first, then the governing ratio, application to your facts, and limits.

Clean, Distraction-free UI

A focused interface designed for long-form legal reading. Light and dark modes. No telemetry, no banners, no feature clutter.


How a query becomes an answer.

Four steps from your prompt to a grounded, cited legal response — all on your machine.

01

Query Input

Describe your case or legal issue in plain language to explore potential outcomes based on previous judgments and applicable acts. No query syntax required.

02

Hybrid Retrieval

BM25 keyword search across the encrypted knowledge base is fused with a neural embedding pass and candidates are scored.

03

Neural Re-rank

An offline embedding model re-scores candidates by semantic similarity and fuses that signal with keyword scores — no topic-specific rules.

04

Grounded Generation

Vidhi receives the top passages as context and generates a prediction with a sources list — judgments and acts first — entirely on your hardware.

"The right to access justice should not require trusting a third-party server with the most sensitive details of a legal matter." — Writ, built for privacy-first legal research

Download Writ.

Free for personal, non-commercial use. No account required. Windows, macOS, and Linux.

Windows
Installer — .exe · 64-bit
  1. Download Writ-Setup-1.0.0-Windows.exe
  2. Run the installer and follow the wizard
  3. Install Ollama if not already present
  4. Run ollama run prawinin/vidhi in your terminal
Download for Windows
macOS
Disk Image — .dmg · Universal
  1. Download Writ-1.0.0-Mac.dmg
  2. Open the disk image, drag Writ.app to Applications
  3. Install Ollama if not already present
  4. Run ollama run prawinin/vidhi in your terminal
Download for macOS
Linux
Archive — .tar.gz · x86_64
  1. Download Writ-1.0.0-Linux.tar.gz
  2. Extract and run the installer script
  3. Install Ollama if not already present
  4. Run ollama run prawinin/vidhi in your terminal
$ tar -xzf Writ-1.0.0-Linux.tar.gz
$ cd Writ-1.0.0-Linux && ./install.sh
Installation complete.
Download for Linux

Note: These installers are provided directly and are currently unsigned. You may need to bypass OS-level security warnings (such as Windows SmartScreen or macOS Gatekeeper) to run them.

Prerequisites & On-Device Setup: Writ requires Ollama running locally for model inference. You will need to manually download the Vidhi model (2.0 GB) by running ollama run prawinin/vidhi in your terminal. See the Vidhi on Ollama. The encrypted knowledge base (1.7 GB — 1,718,935,552 bytes) ships with the install.



Personal use, free of charge.

Writ is licensed for personal, non-commercial use. You may install and run the software on your own device for personal legal research. Redistribution, modification, or commercial use requires explicit written permission from the author.

Licensing inquiries →

Need a custom deployment?

VyapAI configures Writ for law firms, enterprises, and agencies — with your private corpora, custom LoRA fine-tuning, strict citation guardrails, and models scaled to your dedicated hardware. Reach out to discuss an on-premise deployment.

Send an enquiry

Contact for scaling