← All repos

doofzoff/SIMURG

View on GitHub ↗ ★ 106 Python

Zero-leak online detection of LLM decoding corruption and Free Web Search for your AI agents! Catch repetition loops, language drift and garbage mid-stream, before the user sees a bad token. Works with any OpenAI-compatible API.

3 mentions on Hacker News by 2 people: 3 stories, 0 comments · first 2026-08-24 · latest 2026-08-30