Back to browse
GitHub Repository

Witness-Centered Filter Layer for LLM Applications

5 starsPython

Viveka: filter LLM output against a Lean-verified Advaita Vedanta model

by adyashakti·Jun 2, 2026·7 points·0 comments

AI Analysis

●●SolidBig BrainNiche Gem

Lean-verified philosophical axioms as LLM guardrails — genuinely novel architecture.

Strengths
  • Formal verification via Lean 4 for Śaṅkara's Advaita Vedānta axioms, not just heuristics
  • Honest LIMITS.md acknowledging claim extraction is heuristic while axioms are verified
  • Non-mutating evaluate() returns Verdict with reframes instead of blocking outright
Weaknesses
  • Extremely niche audience — who needs Advaita-based LLM filtering in production?
  • Zero stars and forks suggests early stage with unproven real-world utility
Category
Target Audience

AI developers building ethical guardrails, philosophy-focused ML researchers

Similar Projects