<?xml version="1.0" encoding="UTF-8"?><oembed><type>video</type><version>1.0</version><html>&lt;iframe src=&quot;https://www.loom.com/embed/0aa2652adedb4315ab12c244be16f515&quot; frameborder=&quot;0&quot; width=&quot;1662&quot; height=&quot;1246&quot; webkitallowfullscreen mozallowfullscreen allowfullscreen&gt;&lt;/iframe&gt;</html><height>1246</height><width>1662</width><provider_name>Loom</provider_name><provider_url>https://www.loom.com</provider_url><thumbnail_height>1246</thumbnail_height><thumbnail_width>1662</thumbnail_width><thumbnail_url>https://cdn.loom.com/sessions/thumbnails/0aa2652adedb4315ab12c244be16f515-d2c0650fd86157a0.gif</thumbnail_url><duration>604.032</duration><title>Hyperliquid Trade Settlement Pipeline Overview</title><description>This Loom presents the architecture for a Hyperliquid settlement pipeline that ingests live trade data, batches it, and submits batches to a mock chain with dead-letter handling for failures. It describes four services: FeedAdapter (WebSocket to Trades), Ingestor (normalize and validate), BatchPublisher (flush by size of 50 trades or time), and Chain submitter (submit batch data). The system uses Kafka via Redpanda, at-least-once delivery with mockchain deduplication via shared objects and batch and readup IDs, plus exponential backoff for WebSocket reconnects. It includes demos showing successful chain submissions, a dead-letter queue test by forcing a chain failure rate of 1, and an end-to-end test verifying processed trade counts match input after duplication checks to avoid data loss.</description></oembed>