Skip to content

Release v1.1.0 - TypeScript SDK + Comprehensive Documentation

Choose a tag to compare

View all tags
@rahuldass19 rahuldass19 released this 31 Dec 21:27
v1.1.0
ae34d66

QWED v1.1.0 - TypeScript SDK + Technical Deep Dive

Released: January 1, 2026

🚀 Major Updates

1. TypeScript SDK Published to npm ✅

npm install @qwed-ai/sdk

Available now on npmjs.com!

  • Full TypeScript typing support
  • Works with Node.js and browser (ESM/CJS)
  • Same API as Python SDK
  • 216 downloads on Zenodo whitepaper (high interest!)

Example:

import { QWEDClient } from '@qwed-ai/sdk';

const client = new QWEDClient({ apiKey: 'your_key' });
const result = await client.verifyMath('2+2', '4');
console.log(result.verified); // true

NPM Package: @qwed-ai/sdk

2. Comprehensive Technical Documentation (566 Lines Added)

New Whitepaper Sections:

✅ Section 3.4: Output Format Enforcement (197 lines)

  • Function calling mechanism (OpenAI Tools/Anthropic Tool Use)
  • Pydantic schema validation
  • Pre-verification schema checks
  • Result: Zero ambiguity in LLM outputs

✅ Section 4.2.1: DSL-Based Logic Security (170 lines)

  • CRITICAL: Complete DSL documentation (was missing!)
  • Whitelist-based S-expression parsing
  • No eval(), exec(), or __import__() - guaranteed
  • 40+ whitelisted operators
  • Secure compilation to Z3

✅ Section 4.3.1: Language Detection (91 lines)

  • Explicit language declaration (Python, JS, Go, SQL, etc.)
  • Language-specific AST parsers
  • No misidentification possible

✅ Section 4.4.1: SQL AST-Based Obfuscation Detection (90 lines)

  • AST parsing (not regex!)
  • Catches obfuscated injections: WHERE /**/1/**/=/**/1
  • Tautology detection

✅ Section 3.3.1: Handling Ambiguous Outputs

  • Function calling prevents multiple values
  • Schema rejects List[float] when expecting float

✅ Section 6.4: Case Study Methodology Note

  • Clarifies $12,889 bug is illustrative
  • References actual benchmark data

Whitepaper DOI: 10.5281/zenodo.18110785

3. PDF-Friendly Documentation

  • Replaced all Mermaid diagrams with ASCII art
  • Works in GitHub, PDF exports, and offline docs
  • 5 diagrams updated across all repos

📦 What's Included

SDKs Available

Language Package Install Status
Python qwed pip install qwed ✅ Live on PyPI
TypeScript @qwed-ai/sdk npm install @qwed-ai/sdk ✅ NEW: Live on npm!
Go In repo See sdk-go/README.md 🔶 Source available
Rust In repo See sdk-rust/README.md 🔶 Source available

🔧 Integrations

  • ✅ LangChain (439 lines, production-ready)
  • ✅ CrewAI (455 lines, production-ready)
  • ✅ LlamaIndex (full integration)

📊 Benchmarks

Against Claude Opus 4.5:

  • Financial calculations: 73% LLM accuracy → 100% QWED detection
  • Mathematical reasoning: 81% → 100%
  • Adversarial prompts: 85% → 100%
  • Code security: 78% → 100%

215 test cases, 100% error detection rate

🐛 Bug Fixes

  • Fixed Mermaid diagram rendering on GitHub (chunk 2291 error)
  • Updated DOI references across all documentation
  • Corrected Zenodo version field formatting

📖 Documentation

🔗 Links

🙏 Contributors

📜 License

Apache 2.0

🎯 What's Next?

  • JOSS peer review submission
  • Go SDK official release (pkg.go.dev)
  • Rust SDK on crates.io
  • Additional language support
  • More verification engines

Full Changelog: v2.0.0...v1.1.0

Assets 2
Loading