Exchange Feed Architecture
Trading venues publish trades, order book deltas, snapshots, funding, liquidations, and instrument metadata through exchange-specific WebSocket and REST feeds. DeepTick normalizes those feeds into typed records.
Trading venues publish trades, order book deltas, snapshots, funding, liquidations, and instrument metadata through exchange-specific WebSocket and REST feeds. DeepTick normalizes those feeds into typed records.
Collector records keep exchange timestamps when provided and local receipt timestamps from the collector. This lets users inspect latency, ordering, and collector health.
L1 is best bid/offer, L2 is price-level depth, and L3 is order-level lifecycle data where venues expose it. DeepTick should only publish L3 where the exchange feed supports it.
Current infrastructure work has focused on Hyperliquid, Lighter, Pacifica, and Bitstamp. More connectors exist in the codebase and should be promoted only after collector and quality validation.
Bitstamp belongs in Frankfurt. Current APAC collectors run in Tokyo for lower latency to the active Asian venues.
The intended serving model is daily compacted Parquet files backed by quality metadata. Hourly or 15-minute raw batches can remain internal collector artifacts.
Production downloads should use authenticated API requests that return signed object URLs. Public sample data can be served separately through CloudFront.
The product direction is a local sidecar that downloads partitions, reconstructs books, and streams historical events to strategy code with deterministic timing.
Python, TypeScript, and Rust clients are repo-level goals. Public package installation commands should be treated as launch copy until package publishing is complete.