fabro-create-workflow

To generate a DOT graph from natural language for an agentic pipeline, you can define your multi-step process requirements in plain text to produce a valid Fabro digraph with correct node shapes, start and exit nodes, and conditional routing.

Conditional routing in a Fabro workflow works by using diamond gates that evaluate routing conditions, allowing the pipeline to branch into parallel fan-out topologies or trigger verify and fix loops without getting stuck in infinite loops.

You can assign different LLM models to specific roles in a workflow DOT graph by generating a model_stylesheet block, which handles role-based model selection for coding, review, and verification instead of scattering model IDs across individual nodes.

The best way to create a TOML run configuration for a multi-step agent process is to generate a minimal runnable file containing the version and graph references, extending it with sandbox settings as needed for your agentic pipeline.

You do not need to manually write retry policies for verify and fix loops in agentic pipelines, as the workflow generation automatically sets retry and goal-gate behaviors to ensure conditional routing resolves correctly without infinite loops.

An agentic pipeline topology requires specific node types because LLM prompts must be assigned only to LLM nodes, while routing and conditions are strictly handled by diamond gates, ensuring correct node semantics throughout the DOT graph.