🧠 AI Chain-of-Thought Parser

Reasoning token extractor for o1, o3, DeepSeek-R1, and Claude Extended Thinking. Separates internal scratchpad reasoning from client-facing output and detects cognitive self-corrections.

THINKING RATIO
73%
Internal compute
SELF-CORRECTIONS
2 Points
Heuristic backtracking
OUTPUT SANITIZED
100% Clean
Zero thought leaks