// the find
paradigmxyz/cryo
cryo is the easiest way to extract blockchain data to parquet, csv, json, or python dataframes
cryo extracts Ethereum on-chain data (blocks, transactions, logs, traces, state diffs, ERC20 events, etc.) to Parquet, CSV, JSON, or Python dataframes via a CLI or Python bindings. It's aimed at blockchain data analysts and researchers who need bulk historical data without writing custom RPC polling loops. Built in Rust with a Python wrapper via PyO3/maturin.
The dataset coverage is genuinely wide — 35+ dataset types including storage diffs, geth opcodes, and native transfers that most ETL tools ignore. Block range syntax is well-designed: shorthand like `15M:16M`, relative ranges, and reading block lists from existing Parquet files makes chaining queries practical. Schema consistency across tables (same column names and types for block_number, gas, etc.) makes joins actually work without surprise type mismatches. Python bindings are first-class, not an afterthought — you get the same Rust performance with `import cryo`.
Still depends on ethers-rs, which is deprecated and unmaintained since the Rust ecosystem moved to alloy-rs — this is a real maintenance liability and the README acknowledges a migration example exists but the core hasn't moved. The default settings will saturate a paid RPC endpoint without --requests-per-second, and the README buries this warning in a note rather than making it a hard default or prompt. No incremental/streaming mode: if a long extraction dies at block 14.8M, you restart from scratch (though chunk files partially mitigate this). Last push was January 2025, and the repo shows signs of stalled development — issues piling up, the ethers-rs migration not completed, and several datasets marked experimental with no follow-through.