back

blockweaver

Feature-selected EVM block datasets with verifiable provenance.

Immutable, verifiable EVM block dataset acquisition.

// overview

Blockweaver downloads feature-selected EVM block ranges through JSON-RPC or Google BigQuery into immutable Parquet or CSV datasets. Chains, sources, and providers are configuration rather than code, and each successful request publishes one data file with one canonical manifest.

The tool is built for reproducible research inputs: it resolves block or time ranges against finalized chain state, resumes only exact request bindings, and keeps source-specific acquisition behind one artifact contract.

// how it works

A Typer CLI loads strict TOML configuration. aiohttp batches EVM JSON-RPC requests; the optional BigQuery source discovers compatible schemas, enforces a dry-run byte cap, and streams bounded pages; Polars builds the same typed tables for either source.

Complete chunks are digest-bound before reuse. The assembled candidate is validated, synced, and atomically published without replacement; its manifest records schema, provenance, finality, verification samples, and a SHA-256 digest without secrets.

// capabilities

Configurable acquisition

Uses named chain, source, and provider profiles to acquire finalized EVM history through JSON-RPC or optional Google BigQuery.

Feature-selected datasets

Coalesces requested block-header and fee-history features into typed Parquet or canonical CSV output.

Layered verification

Validates artifact integrity offline; RPC verification refreshes finality and checks ancestry and deterministic row samples, and independently verifies BigQuery acquisitions before publication.

Recoverable publication

Resumes validated checkpoints and atomically publishes immutable two-file artifacts without overwriting destinations.