feat: image extraction via in-Space vision model (MiniCPM-V-4.6)

feat: V1 text PII detection and reversible redaction