Stateflow Labs releases two open-source Python SDKs for building resilient AI systems with state persistence and automatic recovery

Stateflow Labs published ALGOgent Runtime and Adaptive Runtime, two MIT-licensed SDKs designed to handle failures in automation scripts and AI systems. ALGOgent Runtime is a lightweight synchronous SDK for simple AI pipelines with retry logic, state persistence, checkpoint recovery, and confidence scoring. Adaptive Runtime is a full async runtime layer for production AI systems with context analysis, confidence scoring, decision-making, SQLite persistence, and automatic crash recovery.

Both SDKs run without GPU, external cloud services, or heavy ML frameworks—pure Python on Python 3.10+ (ALGOgent) or Python 3.11+ (Adaptive). ALGOgent uses JSON file state storage and synchronous execution; Adaptive Runtime uses SQLite async state persistence and full asyncio-based execution.

The SDKs are built on the premise that production AI problems are primarily runtime problems, not model problems. They provide memory that survives crashes, resilience through checkpoints and retry logic, contextual behavior that adapts to real conditions, and confidence awareness—how certain a decision is.