JSON Schema Auto-Generator

Provide one or more JSON samples (inline or from URLs) and get an inferred JSON Schema (Draft 7 / 2020-12) describing their shape. Bootstrap API validators, Apify input schemas, BigQuery / DuckDB schemas. Powered by genson. $0.01 per inference.

---

Why this exists

You hit an API that returns JSON. You want to validate downstream payloads, store them in a typed table, or auto-generate TypeScript types — but writing a JSON Schema by hand from 30 nested fields is tedious.

This actor takes one or more sample payloads and infers the schema. The result is a real, RFC-compliant JSON Schema you can drop into Ajv, BigQuery, OpenAPI, Apify input_schema.json, etc.

---

What you get

{
  "_type": "schema",
  "samples_used": 3,
  "schema_uri": "https://json-schema.org/draft/2020-12/schema",
  "title": "User",
  "inferred_schema": {
    "$schema": "https://json-schema.org/draft/2020-12/schema",
    "type": "object",
    "title": "User",
    "additionalProperties": false,
    "properties": {
      "id": {"type": "integer"},
      "email": {"type": "string"},
      "tags": {"type": "array", "items": {"type": "string"}},
      "meta": {
        "type": "object",
        "properties": {
          "created": {"type": "string"},
          "verified": {"type": "boolean"}
        },
        "required": ["created"]
      }
    },
    "required": ["id", "email"]
  },
  "schema_str": "<pretty-printed schema as text>",
  "top_level_keys": ["id", "email", "tags", "meta"],
  "top_level_required": ["id", "email"]
}

The full schema is also saved as inferred_schema.json in the run's KeyValueStore — easy to download.

---

Quick start

Single sample

{
  "sample": {
    "id": 1,
    "email": "test@example.com",
    "tags": ["a", "b"],
    "meta": {"created": "2024-01-01", "verified": true}
  }
}

Merge multiple samples (recommended — better union types)

{
  "samples": [
    {"id": 1, "name": "foo"},
    {"id": 2, "name": "bar", "deleted_at": "2024-01-01"},
    {"id": 3, "name": "baz", "deleted_at": null}
  ]
}

Fetch samples from URLs

{
  "sampleUrls": [
    "https://api.github.com/users/torvalds",
    "https://api.github.com/users/octocat"
  ],
  "schemaTitle": "GitHub User",
  "schemaUri": "https://json-schema.org/draft/2020-12/schema"
}

---

Pricing

Pay-Per-Event: $0.01 per schema inference.

Cheap, fixed-cost. Run as many times as you want during API evolution.

---

Use cases

1. Bootstrap Apify input_schema.json — Use it on a sample input, drop the schema into your actor's .actor/input_schema.json
2. REST API validators — Generate Ajv-compatible schemas from real production responses
3. BigQuery / DuckDB tables — Convert JSON Schema → DDL with a follow-up tool
4. OpenAPI — Drop schemas under components.schemas for type-safe SDK generation
5. Comparison — Run on prod vs staging payloads; diff the schemas to spot drift

---

Output details

• inferred_schema — the schema as a JS object (use this in code)
• schema_str — pretty-printed text (paste into a file)
• top_level_keys — convenience for renaming / sorting fields
• additionalProperties: false is set by default for objects (strict mode). To allow extras, remove it before using the schema.

---

Limitations / gotchas

• More samples = better schema. A single sample can't tell required vs optional. Pass 5-10 samples covering all common cases.
• null handling — When a field is sometimes null, the inferred type becomes ["string", "null"] (or similar). This is correct JSON Schema but some tools (e.g. old Avro) don't like unions.
• Format detection — genson doesn't infer format: date-time etc. by default. For format-aware generation, post-process manually.

---

Engine

• genson v1.2+ — the reference Python implementation. Well-maintained, used in many data pipelines.

---

Related actors (same author)

• Web Page → Markdown Converter
• HTML Metadata Extractor
• PDF Text Extractor
• Sitemap URL Discovery

---

Feedback

A short review helps developers find it: Leave a review on Apify Store