"The Workflows4s Web UI, originally developed during Google Summer of Code by Atharva, was further polished, merged, and officially released. It is now a usage-ready, full-stack application that enables inspection and interaction with running workflows. Feel free to check the dedicated announcement for more details. Giovanni built llm4s-tripper, a travel planning agent combining llm4s with workflows4s. It serves both as an end-user application and a concrete example of how workflows4s can orchestrate complex, long-running, AI-driven processes."
"A number of important features landed during this period: Workflow instance signal inspectionAdded the ability to inspect expected signals for a specific workflow instance (contributed by Watson), improving introspectability and operations. WorkflowInstanceEngineA new abstraction enabling fine-grained and extensible customization of workflows4s runtime behavior - conceptually similar to how STTP allows backend customization. Drafting support polished and expandedDrafting now covers all available operations and comes with proper documentation, thanks to substantial work by Watson."
Workflows4s advanced across multiple areas, becoming more usable and robust. The Web UI was polished, merged, and released as a usage-ready, full-stack application enabling inspection and interaction with running workflows. A real-world application, llm4s-tripper, demonstrates orchestration of complex, long-running AI-driven workflows for travel planning. Engine and introspection features were added: instance signal inspection, a WorkflowInstanceEngine abstraction for customization, expanded drafting support with documentation, and stateful durable retries supporting configurable strategies like exponential backoff across days to months. Work began on effect-polymorphism to support multiple Scala effect systems. A linter project started to enable static analysis of workflows.
Read at Medium
Unable to calculate read time
Collection
[
|
...
]