Skip to main content

Documentation Index

Fetch the complete documentation index at: https://superwire.dev/llms.txt

Use this file to discover all available pages before exploring further.

Superwire builds a dependency graph from references. Nodes with no dependency relationship can run concurrently.

​
Independent agents

agent extract_risks {
    model: model.fast
    instruction: "List risks for {{ input.topic }}."

    output {
        risks: [string]
    }
}

agent extract_actions {
    model: model.fast
    instruction: "List action items for {{ input.topic }}."

    output {
        actions: [string]
    }
}

agent create_plan {
    model: model.smart
    instruction: "Create a plan from risks {{ agent.extract_risks.risks }} and actions {{ agent.extract_actions.actions }}."

    output {
        plan: string
    }
}
extract_risks and extract_actions can run at the same time. create_plan waits for both.

​
Looped agents

agent summarize_each for task_id in input.task_ids {
    model: model.fast
    instruction: "Summarize task {{ task_id }}."

    output {
        task_id: number
        summary: string
    }
}
Each loop iteration is independent unless it references shared upstream work that must run first.

​
Practical guidance

  • Split unrelated work into separate agents.
  • Keep dependencies explicit through references.
  • Use looped agents for batch work.
  • Add aggregation agents only after fan-out work has completed.