What's inside an hourly Pump.fun file

Each file holds one UTC hour of one venue: the Pump.fun bonding curve or PumpSwap, where coins trade after they graduate. Nothing is filtered out: every swap, every launch, every graduation and every liquidity change is there, including the trades of bots.

Rows

• swap: a buy or a sell, with the amounts and the pool's reserves right after it.
• create: a new token, with its name, symbol and metadata link.
• bonding_complete: the curve filled up and the coin graduated.
• liquidity: someone added to or removed from a PumpSwap pool.

Order

Rows are sorted by slot, then by the transaction's position in the block, then by the instruction inside the transaction. That is the order the chain executed them, so two trades in the same block are never ambiguous. signature + outer_ix_index + inner_ix_index identifies a row uniquely.

Loading a file

import pandas as pd

df = pd.read_parquet("pump_fun_2026-10-28_12.parquet")
swaps = df[df.event_type == "swap"]
print(swaps.groupby("action").lamports_amount.sum() / 1e9)  # SOL volume by side