Seeds
Defined data. Compact changes. Shared understanding.
1,000,000 Events · Same Synthetic Dataset
Expanded JSON → Seed Packs
Expanded JSON Array
428.60 MBSeed Packs Implementation of PIL PACs
44.78 MBSeed packs as stored in Pods. Search indexes, summaries, and other datastore files bring the full Trie-Trunk fixture to 168.41 MB.
Measurement Details
One million synthetic events across five resources over 30 days, with repeated messages and dictionary values. The JSON array retains expanded field names and routing context. Its size is derived from the verified JSONL export by replacing separators and adding brackets: one extra byte.
Pods includes its container overhead. Write seed packs remain uncompressed; other seed packs use gzip. Whole-file gzip of the JSONL export is 25.44 MB, without the indexes or independently retrievable seed packs. This is a storage comparison, not a throughput benchmark.
Read the Measurements ↗Define Once. Stop Repeating.
Field Positions Replace Keys
The data definition carries field names and types. Each data seed stores values in the agreed positions, instead of repeating the same keys in every object.
Store the Value at Its Defined Width
Numbers use binary types rather than decimal text. Declared string dictionaries replace repeated values with compact IDs. Optional adaptive integer encoding reduces widths when values fit.
Send Only the Change
A data seed can hold a complete object or a partial change. Absent and null remain distinct, so an omitted field does not silently erase a value.
Share Values, Then Compress
Seed packs group data seeds, factor common values, and compress the remaining body. The codec owns decoding, so the same data can travel into local storage and analysis.