Switch model to openai/gpt-oss-120b on Groq

updated readme for btter visiblity

Fix pool disconnect issues: recreate Supabase client session per request

Fix false positives on prompt injection checks: add word boundaries to DAN and STAN

Support ES256 algorithm for new Supabase JWT signing keys

Secure backend JWT verification using Supabase JWKS signature validation

Sanitize env.example: replace credentials with placeholders

Security fixes: JWT auth, rate limiting, and atomic usage increments

Increase free limit to 8

Update to Razorpay Orders checkout

Phase 6 Step 2: Implement Razorpay subscription integration and webhooks

Configure Human-Grade Writing constraints and force enhanced mode

Update Dockerfile to copy usage.py to container

Phase 6 Step 1: Add Supabase Auth, usage limits checking and tracking

Add input sanitization against prompt injection

Improved few-shot examples and natural burstiness in Enhanced mode

Fix hyphenated compound words in Enhanced prompt

Apply Walters research anti-AI prompts with perplexity and burstiness rules

Remove hyphens and dashes from all output prompts

Add automatic double pass for Enhanced mode

Integrate few-shot writing examples for enhanced prompt

Upgrade humanizer model to Llama 3.3 70B

Add readability and purpose controls to humanizer

Remove detector and evaluator from Dockerfile

Remove detector and evaluator, simplify API

Swap to ChatGPT-specific detector model

Update CORS for Vercel

Switch to Groq API

Deploy humanizer backend