When triggering runs in Inferable, you can specify a schema to ensure the output is structured in a way that’s easy to process programmatically.

​
Defining Output Schema

You can define the expected structure of your run’s output using Zod schemas:

import { z } from "zod";

const run = await client.run({
  initialPrompt: "List all products with low inventory",
  resultSchema: z.object({
    products: z.array(
      z.object({
        name: z.string(),
        quantity: z.number(),
        needsRestock: z.boolean(),
      }),
    ),
  }),
});

​
Output Format

When a run completes successfully, the result will be structured according to your schema:

{
  "id": "01J8GNNHRRZA9E605BEM57A3HT",
  "status": "done",
  "result": {
    "products": [
      {
        "name": "Widget X",
        "quantity": 5,
        "needsRestock": true
      },
      {
        "name": "Gadget Y",
        "quantity": 12,
        "needsRestock": false
      }
    ]
  }
}

If a run cannot produce a structured result matching your schema, the result field will be null and a message explaining why will be provided.

​
Using with Agents

When working with Agents, you can define a default resultSchema that will be used for all runs created with that agent. This is particularly useful when you want to standardize output across multiple runs:

const run = await client.run({
  agentId: "AGENT_ID",
  initialPrompt: "Analyze customer sentiment",
  // This will override any resultSchema defined in the agent
  resultSchema: z.object({
    sentiment: z.enum(["positive", "neutral", "negative"]),
    confidence: z.number(),
    keywords: z.array(z.string()),
  }),
});

​
Integration Examples

​
Zapier

When using Inferable with Zapier, structured output allows you to use the results in subsequent steps of your Zap. The output will be automatically converted to a format Zapier can understand.

​
API Endpoints

When building API endpoints that use Inferable, structured output ensures your API responses are consistent and typed:

app.post("/analyze", async (req, res) => {
  const run = await client.run({
    initialPrompt: req.body.text,
    resultSchema: z.object({
      categories: z.array(z.string()),
      summary: z.string(),
      confidence: z.number(),
    }),
  });

  const result = await run.poll();
  res.json(result);
});

Runs

Learn about run execution and lifecycle

Agents

Configure reusable run settings

Zapier

Use structured output with Zapier

Functions

Work with function results

Was this page helpful?

Suggest edits
Self hosting InferableAdopting Inferable