Every dataset behind the MadeOnSol Solana & Robinhood Chain API — what it contains, how far back it goes, and whether a gap in it could ever be recovered. Coverage windows below are measured nightly against the live database, not asserted from memory; the same measurements are available at GET /api/v1/manifests.
Buys and sells of launchpad tokens from the pump.fun pipeline (pump.fun, LaunchLab/bonk, bags) while each mint is inside its capture window, on the curve and the pools it trades in after graduation. Not every Solana DEX trade. Buys under 0.05 SOL (under $3.50 when paid in USDC/USDT) are not stored; sells that receive SOL are kept at any size. MEV-stripped, with executed price vs. pool price at the slot.
The same trade-tape treatment as token_trades, for Robinhood Chain (chain ID 4663) — Uniswap v2/v3/v4 and Pons.
Price, market cap, liquidity, and volume per token, bucketed to the minute.
The same candle treatment as token_ohlc_1m, for Robinhood Chain tokens.
Current price, market cap, and liquidity state per token, updated continuously as trades arrive.
Fired events when a token's price/volume crosses a surge or dormant-token-revival threshold, with the +1h outcome attached.
Every token a tracked deployer wallet has launched, with bonding outcome.
The indexed deployer-wallet roster with current reputation tier and bonding-rate statistics.
One row per deployer per day their reputation state changed — the record behind our as-of (no look-ahead) reputation queries.
Every trade from our tracked roster of influencer and smart-money wallets, with entry market cap and PnL context.
Write-on-change history of wallet flags — deployer, alpha, dump-cluster, KOL, sniper, and bundler tags — per wallet.
Streamflow, Jupiter Lock, and Bonfida vesting/lock accounts subscribed per token.
Pump.fun fee-sharing and social-handle claim events per token.
It depends on the dataset — trade tape and candles go back to when we started capturing that specific stream (see the coverage window on each entry below, measured nightly, not estimated). Deployer and KOL reputation data is continuously enriched, so its usefulness grows with time rather than having a fixed start date.
For most datasets, no — trade tape and candle data are either kept permanently in Postgres or exported to a dual-provider Parquet archive under Object Lock before being relocated, so nothing is deleted. A few datasets have genuine, disclosed exceptions: 1-minute candles (token_ohlc_1m / rhc_ohlc_1m) are built incrementally and have no rebuild path, so a gap from an outage is permanent unless corrected by hand. See each entry's backfill note below.
A few datasets in this dictionary are marked as rolling windows rather than permanent (e.g. token_prices, token_surge_events) — those reflect current state or a recent window by design, not full history. Where a permanent-history equivalent exists (e.g. token_ohlc_1m for price history vs token_prices for live state), we say so.
A nightly job measures rows written in the last 24 hours, the earliest and latest event timestamp, and a row-count estimate for each dataset, and writes one row per dataset per night. The same numbers are available programmatically at GET /api/v1/manifests (any valid API key).