JSON Array to Unique Record Stream

Keep the first complete value from each exact typed equality group and export ordered JSON Lines. Existing JSON Records to CSV deduplicates exported rows; existing Clean & Organize Text deduplicates text lines. This uses typed whole-JSON equality and first numeric-token spelling before serializing records as NDJSON, preserving nested values and record order.

• Each downstream CSV/JSON input remains limited to 5 MiB; a larger completed intermediate stays a checkpoint and final success is refused. Originals/checkpoints retain omitted or replaced values. Root JSON array with at most50,000 values;200,000 parsed/identity values,40levels,10,000-unit number tokens,20MiB identities/combined outputs. Objects compare without member-order significance; arrays/string values remain exact and typed; mathematical decimal equivalents deduplicate without float rounding. First retained numeric spelling/order survives. The primary-only connection deliberately omits standalone deduplication-report.json, still available in that unchanged standalone UI. Output is UTF-8 NDJSON source (.ndjson), carried by the existing plain-text artifact contract: one compact JSON value per LF record and one terminal LF; empty array gives empty stream. No by-key deduplication, event replay or streaming network connection.
• No native other-browser/physical-device/heap-erasure acceptance is established.