prose

You orchestrate multi-agent workflows by writing structured programs that define agent interactions, sequences, parallel tasks, and loops. The AI session acts as a virtual machine interpreter, executing the declarative logic and managing data flow for complex AI-driven tasks.

Automating parallel execution and state persistence is achieved by defining deterministic workflows in a structured program. This approach manages data flow, handles errors, and maintains execution state across sequences and loops of AI agent sessions.

To get started with agent orchestration, you can run the example program 'examples/01-hello-world.prose'. This demonstrates how to define and execute reusable, deterministic workflows for tasks like research synthesis or content generation.

Yes, you can use declarative programs to manage error handling and data flow in AI agent sessions. The execution model supports defining execution logic that automatically handles errors and routes data between orchestrated agents.

Yes, this approach supports building reusable workflows for complex AI-driven tasks. By defining agent interactions and execution logic in a structured programming language, you can create deterministic workflows that automate tasks like code review and content generation.

The virtual machine model executes structured programs by having the AI session embody the interpreter. This allows the declarative programming language to directly manage agent orchestration, state persistence, and parallel execution within the session.