Deterministic A2A Broken-JSON Repair & Schema Validator

Choose this endpoint when your AI agent pipeline receives JSON output from an LLM that fails JSON.parse() due to common LLM formatting errors like trailing commas, markdown code fences, single quotes, or mid-generation truncation. It is purpose-built for deterministic, rule-based repair rather than probabilistic rewriting, making it reliable for production multi-agent pipelines where consistency matters. Prefer this over manually writing repair logic or prompting the LLM to retry, especially when you need low-latency, per-call fixes without burning additional LLM tokens.

• Input so severely malformed that deterministic repair cannot recover a valid structure — returns an error or best-effort partial result
• Extremely long JSON strings may hit query parameter length limits when passed via GET
• Ambiguous truncation points may result in incorrect inferred structure closing
• Empty or whitespace-only raw_json input returns a validation error