Nika – Intent as Code for AI Workflows (github.com)

Nika has introduced a new workflow language focused on transforming repetitive AI tasks into manageable, verifiable files that can be run, shared, and audited, much like SQL with databases or Docker with container definitions. This innovative approach allows users to convert AI operations into readable YAML files without requiring cloud services, enabling local execution on any chosen large language model (LLM). The open Apache-2.0 specification makes it accessible and adaptable, while a single Rust binary serves as the reference engine for performance. The significance of Nika lies in its ability to provide a structured framework for AI workflows, ensuring safety, reproducibility, and portability. Its unique "Intent as Code" philosophy means that workflows can be audited before execution, checking for dependencies, cost implications, and security boundaries, thereby minimizing risks associated with running AI tasks. The system allows for parallel task execution while maintaining a clear execution trace—making it easier for developers to manage AI processes efficiently. This robust and open-source tool could redefine how developers interact with AIs, transitioning from vague chat interfaces to precise, actionable code that is both human-readable and machine-executable.