high-performance analytics
for quantitative finance

An analytical database for time-series workloads, designed for AI agents and written in Rust. DataFusion SQL · native ASOF joins · immutable Parquet segments.

cargo install h5i-db  ·  pip install h5i-db
agent shell
time travel
version log · trades
v0 create0 rows
v1 write4 rows
v2 append5 rows
v3 delete_range3 rows · planned
every version stays readable
read v1O(1), one manifest GET
as-of tsO(log V) binary search
previewable mutations · plan / apply

disk-backed, 20M rows · vs the best of Polars · DuckDB · pandas · PyArrow

time-range scan 0.01% h5i-db 7.3ms pyarrow 10.1ms
1-min OHLCV + VWAP 164ms duckdb 774ms
ASOF join 20M × 5M 439ms polars 408ms
full aggregation 47ms duckdb 47ms

manifest pruning: narrow queries never open irrelevant segments, plus time travel, snapshots, and crash-safe commits none of the baselines have

quant-grade speed.
agent-grade safety.

h5i-db init market.db · h5i-db ui market.db · pip install h5i-db