JSON Repair, Normalize & Schema Checker

LLM Output Validator: JSON Schema + Repair Hints

codeclouds/llm-output-validator

Validate LLM/tool-call JSON output against a JSON Schema (draft-07 or 2020-12) — get machine-readable error paths, concrete repair hints per error, and optional deterministic normalization/repair.

Validate a stream of JSONL/NDJSON records against a JSON Schema (draft-07 subset) one line at a time. Per-record error rows, drop-or-tag policy, summary by error type. Pairs with json-schema-validator + jsonl-to-csv.

Function Call JSON Repair

junipr/function-call-json-repair

Repair malformed JSON function-call arguments and tool outputs using safe deterministic fixes, validate against schemas, and emit repair diagnostics with before/after records.

JSONL Lint and Repair

junipr/jsonl-lint-and-repair

Lint JSONL files line by line, detect malformed records and schema issues, optionally repair safe common problems, and output cleaned JSONL plus detailed line-level diagnostics.

JSON Schema Batch Validator

junipr/json-schema-batch-validator

Validate JSON/JSONL batches against JSON Schema and return row-level validation errors and summaries.

Extract auto repair shops in any US state: name, address, phone, website, hours & GPS. Clean B2B lead list from OpenStreetMap. No key.

Clean, deduplicate, normalize, and schema-check JSON datasets with a simple structured input.

JSON Schema Validator for AI Agents | validate_json_schema

scryptex_labs/validate-json-schema

Agent-ready: JSON string + optional schema in → valid/repair/errors with agentHints out. USE WHEN: pipeline validation, LLM output repair, tool chaining. MCP twin at scryptexlabs.com.

Validate, format, and analyze JSON data. Check JSON syntax, validate against a JSON schema, count objects/arrays, identify data types, find duplicate keys, and pretty-print minified JSON. Essential for developers, API integration testing, and data quality assurance.

LLM Output Guard | JSON Schema Validation for Agent Output

broomwagon/llm-output-guard

Check that your model's JSON matches the schema you asked for. Every failure comes back with its path, the keyword it broke, and the repair that would fix it. Shape repairs are opt-in and never invent content, so a repair can never hide a real failure.