Parse logs and NDJSON before formatting
Use single-payload mode for a CloudWatch, Lambda, Docker, or API line that wraps one escaped object. Keep prefix extraction enabled when timestamps or log levels appear before JSON.
Use line-by-line mode for NDJSON or batches with one record per line. Formatted JSON returns a readable array, while JSON Lines preserves one compact record per output line.
If lines 2 and 5 are truncated or plain text, valid records remain copy-ready and the parser reports only those source line numbers. It never adds failed content to analytics.
Format the extracted payload for review
Pretty formatting exposes objects, arrays, and nested fields so extracted log payloads and API responses are easier to inspect.
Formatting does not repair invalid JSON. A failed record still needs complete quotes, braces, and brackets from the original log source.
Validate before generating types
Run validation before creating TypeScript interfaces so comma, quote, bracket, and trailing-character issues are caught early.
Use a representative, reviewed sample with optional and nullable fields when generating types from an API response or normalized log record.
Minify only after review
Minify JSON when embedding config, sending compact payloads, or storing fixtures where whitespace is unnecessary.
Keep a formatted version during review to make diffs and troubleshooting easier.