Skip to content

roadmap.md

Latest commit

 

History

History
138 lines (93 loc) · 3.86 KB

roadmap.md

File metadata and controls

138 lines (93 loc) · 3.86 KB
Raw

QWED Roadmap — Historical Archive

⚠️ This document is a historical archive. The QWED roadmap has moved to docs.qwedai.com. The information below is preserved for reference only and may not reflect the current (v5.2.0+) ecosystem.

Vision: The First Open Source Neurosymbolic AI Guardrail

What is Neurosymbolic AI?

QWED combines:

  • Neural Networks (LLMs like GPT-4, Claude, Llama) for natural language understanding
  • Symbolic Reasoning (SymPy, Z3, AST) for deterministic verification

This hybrid approach gives you the best of both worlds: LLMs translate problems into formal logic, and symbolic engines provide mathematical proof.

Completed ✅

v2.1.0 - "QWEDLocal" (January 2026)

  • Client-side verification (no backend needed)
  • Math verification (SymPy)
  • Logic verification (Z3)
  • Code security (AST)
  • Smart caching (50-80% cost savings)
  • Beautiful CLI tool
  • Works with ANY LLM (Ollama, OpenAI, Anthropic, Gemini)

Next Up 🚀

v2.2.0 - "The Privacy Release" (Late January 2026)

Theme: Enterprise-Ready Security

Features:

  • PII Masking 🔒

    • Detect and mask sensitive data before sending to LLMs
    • Credit cards, SSNs, emails, phone numbers
    • Uses Microsoft Presidio
    • mask_pii=True parameter

  • Enhanced Confidence Scores 📊

    • Granular scoring (not just 0% or 100%)
    • Visual confidence bars in CLI
    • Reasoning explanations

  • Neurosymbolic Branding 🧠

    • Updated positioning as neurosymbolic guardrail
    • Technical documentation

Why These Features?

  • PII masking unlocks enterprise/healthcare/banking customers
  • Enhanced confidence provides nuanced feedback
  • Neurosymbolic positioning differentiates from competitors

v2.3.0 - "The Intelligence Release" (Mid February 2026)

Theme: Developer Experience

Features:

  • Cost Tracker 💸

    • Track API spending per query
    • Show cache savings
    • Budget alerts
    • Weekly/monthly reports

  • Streaming Verification ⚡

    • Live progress updates
    • Better UX for slow queries
    • Generator-based API

Why These Features?

  • Cost tracking helps developers optimize spend
  • Streaming improves perceived performance

Future Considerations 💭

Potential v2.4.0+ Features:

  • Multi-model consensus (3 LLMs vote, majority wins)
  • More programming languages (JavaScript, Go, Rust)
  • RAG verification (validate RAG pipeline outputs)
  • Enhanced code verification (more languages)

Note: Features will be prioritized based on user feedback and adoption.

What We Won't Build ❌

Fact Verification (Dropped)

Why: QWED's strength is deterministic verification (Math, Logic, Code = 100% or 0%). Adding probabilistic fact-checking (Wikipedia, news = 60-90% accuracy) would dilute our core value proposition and risk damaging user trust in our deterministic engines.

Alternative: Recommend users use specialized fact-checking tools for this use case.

How to Contribute

We welcome contributions! Here's how you can help:

  1. Try QWEDLocal - Use it, find bugs, suggest improvements
  2. Documentation - Help improve guides and examples
  3. Code - Pick an issue labeled good first issue or help wanted
  4. Feedback - Tell us what features YOU need

See CONTRIBUTING.md for details.

Timeline

Version Theme Target Date
v2.1.0 QWEDLocal ✅ January 2026
v2.2.0 Privacy Late January 2026
v2.3.0 Intelligence Mid February 2026
v2.4.0+ TBD Spring 2026

Dates are estimates and may change based on development progress.

Questions?

  • GitHub Discussions: Ask questions, propose features
  • Issues: Report bugs, request features
  • Twitter: @rahuldass29 for updates

Built with 💜 for a deterministic future.