Experimental prototypes, proof-of-concepts, and sandbox projects.

A scratchpad for building and testing ideas quickly before they graduate into dedicated repositories — or get quietly retired.

This repository houses small-scale experiments, early-stage prototypes, and one-off tools that don't yet warrant their own repository. Projects here range from functional tools shared with collaborators to quick explorations of an idea or technical approach.

Not everything here is polished. That's the point.

An AI-powered tool for translating Hindi Supreme Court and High Court petitions into formal legal English.

Directory: court-translator/ Serverless function: netlify/functions/court-translate.js Stage: Prototype

How it works:

1. Hindi text is split into chunks and translated via Sarvam AI (sarvam-translate:v1, formal mode)
2. The raw translation is refined by Llama 3.3 70B (via Groq) with a legal-terminology-aware prompt that corrects terms like याचिकाकर्ता → Petitioner, माननीय → Hon'ble, अनुच्छेद → Article, etc.

Environment variables required:

• SARVAM_API_KEY — from Sarvam AI dashboard
• GROQ_API_KEY — from Groq console (free tier)

Usage: Open court-translator/index.html in the browser (via Netlify deploy). Paste Hindi petition text and click Translate.

A static HTML signature-collection page for an AOR Supreme Court petition, migrated from ImpactMojo. Submissions go to Netlify Forms; the latest snapshot lives in signatures-backup-2026-04-16.json (66 verified signatures after removing one test record).

Directory: chamber-petition/ Stage: Active

Methodology, data, and grading outputs for evaluating AICC-appointed District Observers on their preparation for District Congress Committee (DCC) membership recommendations. Six components scored: district report submission, proposed names, daily activity reports, supporting documents, potential leaders, non-political influencers.

Directory: aicc_analysis/ Stage: Congress-facing deliverable

Files:

• grading_system.py — scoring engine (weights, thresholds, joins documented in module docstring + commit history)
• extract_candidates.py — pulls candidate-level detail from the AICC portal (requires AICC_USERNAME/AICC_PASSWORD env vars)
• consolidated_district_grading.csv — 592-district A/B/C/D grading
• candidates_detailed.csv — 2,413 candidate-level profiles
• all_timestamps.csv — 9,104 per-record submission timestamps (authoritative count source)
• *.json — raw API extracts (district_reports, proposed_names, attachments, daily_reports, potential_leaders, political_influencers)

A small Node + Playwright tool that captures view-only Google Drive PDFs (the kind you can read but not download) and writes them out as a normal PDF.

Directory: tools/drive-pdf-extractor/ Stage: Internal tool

A Right to Care: Kashmiri Muslim and Other Tribes report (PDF), kept here for direct linking from project pages.

File: docs/a-right-to-care-kmut-report.pdf

Projects in this repo follow a simple lifecycle:

Experiments/
├── index.html                          # Experiments index page
├── court-translator/
│   └── index.html                      # Court Petition Translator UI
├── chamber-petition/
│   ├── index.html                      # AOR Supreme Court petition form
│   ├── README.md
│   └── signatures-backup-2026-04-16.json
├── aicc_analysis/
│   ├── grading_system.py               # DCC-readiness scoring engine
│   ├── extract_candidates.py           # AICC portal extractor (needs env vars)
│   ├── consolidated_district_grading.csv
│   ├── candidates_detailed.csv
│   ├── all_timestamps.csv
│   └── *.json                          # raw source extracts
├── tools/
│   └── drive-pdf-extractor/            # view-only Drive PDF capture
├── netlify/
│   └── functions/
│       └── court-translate.js          # Hindi → English translation API
├── docs/
│   └── a-right-to-care-kmut-report.pdf # KMUT report
├── netlify.toml                        # Netlify build & functions config
├── .env.example                        # Required environment variables
├── LICENSE                             # MIT License
└── README.md                           # This file

All experiments default to the simplest possible stack unless a specific tool requires otherwise:

git clone https://github.com/Varnasr/Experiments.git
cd Experiments
open index.html

This repository is part of the OpenStacks for Change ecosystem — an open-source toolkit for development research, evaluation, and program design.

MIT License — see LICENSE for details.