Malformed JSON, repaired

Response Healing

Automatically validate and repair malformed JSON responses from models when you ask for structured output.

Get startedRead the docs
response_format: json_schema

Model output

Sure! Here is the JSON:
```json
{name: 'Ada',
  "tags": ["math",

Valid JSON

{
  "name": "Ada",
  "tags": ["math"]
}
01

One plugin

Add response-healing, done.

02

Common fixes

Fences, commas, truncation.

03

Healing metadata

Know when a fix happened.

04

JSON modes

json_object and json_schema.

Built for

  • Extraction

    Reliable structured data

  • Agents

    Outputs that always parse

  • Cheaper models

    Use them for JSON safely

Turn it on
terminal
curl https://api.llmgateway.io/v1/chat/completions \  -H "Authorization: Bearer $LLM_GATEWAY_API_KEY" \  -H "Content-Type: application/json" \  -d '{    "model": "gpt-4o",    "messages": [      {        "role": "user",        "content": "Return name and age as JSON"      }    ],    "response_format": {      "type": "json_object"    },    "plugins": [      {        "id": "response-healing"      }    ]  }'
Read docs

More in models & modalities

All features

Vision

Send images, get answers

Document Reading

Send PDFs straight to the model

OCR

Documents and images to markdown

Start using Response Healing today

One API key. Every model. Live in minutes.

Start buildingBrowse models