Wickra Feature StoreOne spec. A feature matrix.

The build is JSON, not code ​

A build is a FeatureSpec: a universe, an ordered list of features, and an optional list of labels. Feature columns are price fields, indicators or microstructure statistics; label columns are forward-looking targets joined to each row.

{
  "universe": ["AAA", "BBB", "CCC"],
  "features": [
    { "kind": "indicator", "name": "Rsi", "params": [14] },
    { "kind": "indicator", "name": "Ema", "params": [20] },
    { "kind": "price", "field": "close" }
  ],
  "labels": [
    { "kind": "forward_return", "horizon": 5 },
    { "kind": "triple_barrier", "horizon": 20, "upper": 2.0, "lower": 1.0 }
  ],
  "scaling": "z_score"
}

Install ​

The same feature build from every language — native Rust, Python, Node.js and WASM, plus a C ABI for C, C++, C#, Go, Java and R.

pip install wickra-feature-store

Run it from any language ​

Construct a FeatureStore from the JSON spec, then drive it with command(json) -> json. Every binding returns the same bytes.

import json
from wickra_feature_store import FeatureStore

spec = json.dumps({
    "universe": ["AAA"],
    "features": [
        {"kind": "indicator", "name": "Sma", "params": [2]},
        {"kind": "price", "field": "close"},
    ],
    "labels": [{"kind": "forward_return", "horizon": 1}],
})

store = FeatureStore(spec)
data = {"AAA": [{"ts": 0, "open": 100, "high": 100, "low": 100, "close": 100, "volume": 1}]}
response = store.command(json.dumps({"cmd": "build_batch", "data": data}))
matrix = json.loads(response)
print(matrix["columns"])

Built on the Wickra core ​

Wickra Feature Store is part of the Wickra ecosystem. Every feature column draws on the 514 O(1) streaming indicators of wickra-core, so a matrix sees exactly the same numbers a backtest or a live chart would.

Wickra Feature Store is a software library, not a trading system, and comes with no warranty — use at your own risk.