FAQ

Questions buyers actually ask

Short answers, no hedging. If yours isn't here, ask us directly — replies come from the team that builds the pipelines.

Buying

What exactly am I buying?: A license to the selected coverage window of one dataset, delivered as Parquet files in the verified 15-column shape, with manifests and schema documentation. You download and keep the files.
Does one purchase cover every table in a dataset?: Yes. A dataset is a program; its tables are the instruction and event types inside it. Any window you buy includes all of them.
Do prices differ per dataset?: No — pricing is uniform across the catalog: $200 latest month, $500 six months, $1,000 twelve months, $2,000 full history, per dataset.
Can I get an invoice or pay by wire?: Yes. Card checkout runs on Stripe; for invoicing, wire transfer, or procurement paperwork, email orders@datastore.sh.

How is the data delivered?: After payment we deliver signed download links to your checkout email. Enterprise buyers can receive drops directly into their own S3 or GCS buckets.
What format do the files come in?: Parquet only — one stable 15-column physical schema across every file, with the decoded payload in the json column. It reads natively in Spark, DuckDB, ClickHouse, pandas, and Polars.
How do I verify a delivery?: Every delivery ships with a manifest carrying per-file sha256 checksums. Verify the bytes before they enter your pipeline; if a file doesn't match, we correct it with a new version at no cost.

What is a stream, exactly?: One dataset delivered live into one sink you control — Kafka, ClickHouse, NATS JetStream, or S3 — for a flat monthly price per pipeline. Same decoded tables and schema versions as the archive, arriving continuously instead of as files.
How do I connect my sink?: Subscribe on the dataset page, then follow the secure link we email after payment. We verify your credentials with a live test write before the stream starts — typically within a few hours, always within one business day.
Is stream delivery exactly-once?: Delivery is at-least-once with deterministic dedup keys per message, plus continuous gap tracking and a public delivery-health record per pipeline. That combination lets your sink enforce exactly-once — honestly.
Can a stream continue where my archive ends?: Yes — that's Backfill + Tail. When your archive is fulfilled we record its final block, and your stream's cursor starts at the next one. Zero gap, zero overlap, verifiable in your sink.
How do I cancel a stream?: Cancel any month — the subscription ends at the close of the billing period and the pipeline stops. Everything already delivered stays in your sink; it's yours.

What period does the archive cover?: Observed row bounds span April 2024 through June 2026 (Solana slots 260,496,001–427,679,999). Earlier history back to genesis is available on request as a backfill.
Are failed transactions included?: Yes — failed transactions are included and flagged via the success column, so you can include or exclude them per query.
What do the observed row counts on dataset pages mean?: They are exact counts from scanning sampled epochs of the real archive — evidence of activity, not an estimate of the full archive. Full-archive counts are confirmed at delivery.
What happens when a dataset is corrected?: Published versions are immutable. Corrections ship as new versions with change logs, free to buyers of an affected window.

What can I do with the data?: Anything internal: research, backtesting, model training, analytics, and powering your own products. Models and signals you build are yours, commercially.
What can't I do?: Resell, redistribute, or publicly host the raw files, or sublicense the data outside your organization. The full terms are in the data license.

Which chains are covered?: Solana today, deeply. Hyperliquid is next. We add networks when we can hold them to the same standard, not before.
The program I need isn't in the catalog. Now what?: Tell us. Coverage requests directly shape what we decode next, and observed-only programs can usually be documented quickly.

Still stuck on something?

Schemas, completeness, procurement — whatever it is, ask.