JSON Generation API

The .generate.json() method creates structured JSON content based on prompts, messages, or schemas. Use this for generating API responses, structured records, or documents with optional real-time streaming.

Usage

Example 1 - Prompt with Plain Schema

Define a schema and generate structured JSON:

unbody.generate.json("Provide user details:", {
  schema: {
    type: "object",
    properties: {
      name: { type: "string" },
      age: { type: "number" },
      address: { type: "string" },
    },
  },
});

Example 2 - Prompt with Zod Schema

Use Zod for schema validation:

import { z } from "zod";
 
unbody.generate.json("Provide user details:", {
  schema: z.object({
    name: z.string(),
    age: z.number(),
    address: z.string(),
  }),
});

Example 3 - Message-Based Input with Schema

Use messages and a schema to guide the generation:

import { z } from "zod";
 
unbody.generate.json(
  [
    { role: "user", content: "What is the structure of a business entity?" },
    { role: "system", content: "Explain in detail with examples." },
  ],
  {
    schema: z.object({
      companyName: z.string(),
      employees: z.number(),
      headquarters: z.string(),
    }),
  }
);

Example 4 - Streaming JSON Generation

Generate structured JSON with real-time streaming:

import { z } from "zod";
 
const stream = await unbody.generate.json(
  "Generate a detailed product catalog entry:",
  {
    schema: z.object({
      productName: z.string(),
      price: z.number(),
      description: z.string(),
      features: z.array(z.string()),
    }),
    stream: true,
  }
);
 
for await (const chunk of stream) {
  if (chunk.finished) {
    console.log('Stream finished');
    console.log('Final response:', chunk.content);
    break;
  } else {
    console.log(chunk.content);
  }
}

Options

schema: Defines the structure of the output (plain JSON schema or Zod schema).
model: AI model to use (e.g., gpt-4).
topP: Diversity control via nucleus sampling (0 to 1).
maxTokens: Maximum length of the response.
temperature: Creativity adjustment (0 to 1).
presencePenalty: Penalize new tokens based on input presence (-2.0 to 2.0).
frequencyPenalty: Penalize tokens based on frequency of occurrence (-2.0 to 2.0).
stream: Enable streaming for real-time content generation (boolean).

Text Generation Overview