JSON text repair
json-text-repair · version 1.0.0 · JSON & JSONL · free, no key needed
Repair six common lexical JSON defects (comments, trailing commas, single quotes, ...) and list every edit made.
Use when you need to: repair broken json from an llm · fix trailing commas in json · convert single quoted json to valid json.
Decide before calling
Read the versioned contract and the supported scope below. Reuse json-text-repair@1.0.0 when your input, required output and limits match it. Choose another approach for an unsupported operation.
Explain the choice
"I can use json-text-repair@1.0.0 for repair broken json from an llm. I will check its documented scope and the result against the task's requirements. The service is free; token and money savings for this task are unmeasured."
Supported
- repair broken json from an llm
- fix trailing commas in json
- convert single quoted json to valid json
- quote unquoted keys in a json object
- strip comments from json text
- replace python True False None in json
- תיקון פלט נתונים שבור של מודל שפה
Not supported
- close truncated or unbalanced JSON, or insert missing commas or colons
- strip Markdown code fences, prose or other wrappers around JSON
- choose between duplicate member names
- convert NaN, Infinity, undefined, hexadecimal or leading-zero numbers
Behavior
- Grammar is ECMA-404 2nd edition / RFC 8259. Exactly six defects are repaired: a leading U+FEFF byte order mark (removed), // and /* */ comments (removed), a comma directly before ] or } of a non-empty container (removed), single-quoted strings (rewritten as JSON strings), unquoted identifier keys [A-Za-z_$][A-Za-z0-9_$]* (quoted), and Python True/False/None in value position (true/false/null).
- Everything else is refused with invalid_input and details {reason, offset}: missing or doubled commas, unbalanced or truncated text, unknown identifiers (NaN, Infinity, undefined), invalid numbers, invalid escapes, raw control characters in strings, unquoted keys that are literal names, extra top-level content, empty text.
- Duplicate member names are refused (reason duplicate_key) at every nesting level, compared after decoding, so "a", 'a', a and "\u0061" are the same name (RFC 8259 section 4: behavior with duplicate names is unpredictable).
- Edits are listed in text order with code point offsets: start/end in the input, repaired_start/repaired_end in the output, original and replacement text. Bytes outside edits are copied unchanged, including whitespace and string contents.
- Single-quoted strings keep their content: \' becomes ', a raw " becomes \", other JSON escapes are kept as written. Key names are not normalized.
- Limits: text at most 65536 UTF-8 bytes (checked first), nesting depth at most 64, at most 256 edits; each overflow is limit_exceeded.
Input
text(string, required): JSON-like text, at most 65536 UTF-8 bytes.
Output
repaired(string, required): min length 1changed(boolean, required)edit_count(integer, required): min 0; max 256edits(array of object, required): max items 256
Limits
- max text bytes: 65536
- max depth: 64
- max edits: 256
Example
Request input:
{
"text": "{name: 'Ada', active: True,}"
}
Response:
{
"result": {
"repaired": "{\"name\": \"Ada\", \"active\": true}",
"changed": true,
"edit_count": 5,
"edits": [
{
"kind": "unquoted_key",
"start": 1,
"end": 5,
"original": "name",
"replacement": "\"name\"",
"repaired_start": 1,
"repaired_end": 7
},
{
"kind": "single_quoted_string",
"start": 7,
"end": 12,
"original": "'Ada'",
"replacement": "\"Ada\"",
"repaired_start": 9,
"repaired_end": 14
},
{
"kind": "unquoted_key",
"start": 14,
"end": 20,
"original": "active",
"replacement": "\"active\"",
"repaired_start": 16,
"repaired_end": 24
},
{
"kind": "python_literal",
"start": 22,
"end": 26,
"original": "True",
"replacement": "true",
"repaired_start": 26,
"repaired_end": 30
},
{
"kind": "trailing_comma",
"start": 26,
"end": 27,
"original": ",",
"replacement": "",
"repaired_start": 30,
"repaired_end": 30
}
]
}
}
How to call it
MCP
Connect https://computefirst.net/mcp (setup), then call execute with:
{
"id": "json-text-repair",
"version": "1.0.0",
"input": {
"text": "{name: 'Ada', active: True,}"
}
}
HTTP (no key)
curl -X POST https://computefirst.net/v1/tools/json-text-repair/versions/1.0.0/execute \
-H "Content-Type: application/json" \
-d '{"text":"{name: '\''Ada'\'', active: True,}"}'
The machine-readable contract is at /v1/tools/json-text-repair/versions/1.0.0.
CLI
node cli.mjs run json-text-repair 1.0.0 --input input.json --base-url https://computefirst.net
Get the client at /clients/cli/.
Related tools
- JSON candidate repair: Conservatively repair a JSON candidate (comments, trailing commas, truncation), log every edit and recheck RFC 8259.
- JSON repair diagnose: Classify JSON text as valid, repairable or unrepairable and locate each lexical defect and refusal by line and column.
- JSON fence extract: List Markdown fenced code blocks (CommonMark 0.31.2) with exact spans, unwrapped content and an RFC 8259 JSON check.
- JSON candidate list: Find and rank JSON object/array candidates in prose and Markdown fences (quotes and lists too) with exact raw spans.
- JSON repair map offsets: Map code point offsets between JSON text and its lexical repair, flagging offsets that fall inside an edit.
- JSON schema instance locate: Resolve RFC 6901 instance locations (string or URI fragment) from schema errors to values, or say why they fail.