Finyuus

Finyuus is a code-first platform designed for the creation, execution, and governance of AI workflows. It caters to developers and teams looking to build robust and manageable AI applications by providing a structured approach to AI logic. The problem Finyuus addresses is the entanglement of AI logic, such as prompts, tool calls, and retries, within application code. This makes it difficult to manage, update, and govern AI behaviors independently. Traditional methods often lead to lengthy system prompts embedded directly in application code, requiring full application rebuilds for minor AI adjustments. Finyuus introduces a small, indentation-based Domain Specific Language (DSL) for composing AI workflows. This DSL allows for the seamless integration of agents, tools, guards, and human approval steps. Workflows can also be nested for complex orchestration. The platform leverages Temporal for durable execution, ensuring that workflows can handle retries, cancellations, and replayability, which is crucial for reliable AI operations. Unlike visual builders, Finyuus stores workflows as plain text. This enables Git-based review, diffs, and version control, making it easier for teams to track changes, collaborate, and maintain a clear history of their AI logic. This approach contrasts with visual tools where workflow changes can become opaque and difficult to manage at scale. Compared to simple prompt chains, Finyuus offers durable, versioned, and observable runs. Observability is provided through integrations like Langfuse, allowing for detailed tracing and cost reporting. Governance is enhanced through built-in guards and human approval mechanisms, ensuring that AI actions adhere to predefined rules and policies. The platform's unique approach lies in its code-first DSL and its reliance on Temporal for robust workflow execution. This combination provides a clear separation between AI logic and application code, offering static analyzability of workflows. The DSL is constrained, allowing for the extraction of workflow graphs, which aids in governance and auditing. Benefits for users include enhanced manageability of AI logic, improved collaboration through Git integration, and greater control over AI behavior via governance features. The durable execution ensures that AI workflows are reliable and can recover from failures. Concrete use cases include building complex AI agents that require multiple steps, integrating human oversight into AI processes, and managing AI-driven tasks that need to be version-controlled and auditable. Developers can use Finyuus to orchestrate sequences of LLM calls, tool executions, and conditional logic. Finyuus is open-source and can be run locally with Docker, including Temporal, ClickHouse, MinIO, Langfuse, and Redis. The platform is designed for developers and teams working on AI applications, offering a more structured and governable alternative to existing solutions. In summary, Finyuus provides a code-first, durable, and governable solution for building and managing complex AI workflows, emphasizing version control and observability for enhanced reliability and team collaboration.