Show HN: Smithers - Declarative AI Orchestration with React (github.com)

Smithers, a newly introduced AI workflow engine, revolutionizes the way AI tasks are orchestrated by allowing users to define workflows as JSX component trees. This innovative approach leverages React to create deterministic and durable workflows, which are executed as directed acyclic graphs (DAGs) while persisting all results in SQLite. Smithers addresses common issues in AI task management, such as brittleness in prompt chains and loss of state during failures, ensuring that tasks can resume seamlessly from the last completed point. This tool empowers developers to create complex, multi-step workflows that are both typed and inspectable. Key technical features include task validation against Zod schemas, error handling that retries tasks upon failures, and the ability to execute tasks in sequences, parallel, or through loops. The workflow re-renders after each step, enabling dynamic adjustments and contributing to its robustness for scenarios like human-in-the-loop reviews or AI pipeline automation. By combining the flexibility of React with the reliability of SQLite, Smithers positions itself as a significant advancement in AI orchestration, ideal for developers seeking to construct resilient and maintainable AI systems.