dataforge-labs/solana-dex-execution

Solana DEX execution costs and routing

What a swap was quoted at on Solana, and which venues the aggregator split it across to get there.

A quote is computed per request against pool state that has already changed by the time anyone asks again. The fill that follows may land on chain; the quote never does, and neither does the route the aggregator considered and rejected.

Contents

name	one row is
e24_solana_quotes	one pair at one size at one moment: output, price impact, dollar value, leg count
e24_solana_routes	one leg of the chosen route: venue, amounts, and the percentage of the order it carried

Reading it

Join the two on (round_ts, input_symbol, output_symbol, in_amount). A quote with n_legs above one was split, and percent on each leg says how the order was divided.

The size ladder is fixed at 1, 50 and 1,000 SOL so that impact is comparable over time rather than against a moving notional. Impact rises with size on every pair, which is the sanity check this table has to pass: one observation ran 0.0000%, 0.0013% and 0.065% on a thin pair against 0.0000%, 0.000018% and 0.000068% on the deepest one.

Routing is the part that is hard to get elsewhere. The same pair goes through one venue at small size and four or five at large size, and the venue mix changes between rounds, which maps where routable liquidity actually sits rather than where volume was reported.

A column that is raw on purpose

amm_report_raw is the aggregator's own report on the venues it considered, stored verbatim as JSON rather than parsed into a comparison.

It is kept raw because it is NOT restricted to the pair being quoted. A request for SOL into a memecoin returns values on the scale of a stablecoin quote, and at least one entry has been the string "Pair insufficient liquidity" rather than a number. An earlier version derived a "chosen route versus best single venue" spread from this field and produced 340 million basis points, which is what exposed the mismatch. The raw object keeps the information without publishing a derived number that means nothing.

Before you build on this

• One aggregator. It is the dominant one on Solana, but a quote is what it would route, not the best price obtainable anywhere.
• Quotes are indicative. A real swap adds transaction landing risk, priority fees and slippage against a moving pool, none of which a quote reflects.
• price_impact_pct is the aggregator's own estimate, not something recomputed here.
• Five pairs, all with SOL as the input. That is a slice of Solana, not a census.
• A failed request is written as an explicit error row with the measurements null, so an unavailable aggregator is distinguishable from a pair with no route.

Partitions are parquet, one file per collection window, under dataset/YYYY/MM/. Every dataset here carries a FIXED 7-day sample WINDOW starting at its own first day of collection, together spanning 2026-08-30 to 2026-09-05, so you can check schema, coverage and quality before asking for more. It does not advance, so there is nothing to gain by re-downloading it. The full history is held privately, available on request.

from huggingface_hub import snapshot_download
import pandas as pd, glob

path = snapshot_download("dataforge-labs/solana-dex-execution", repo_type="dataset",
                         allow_patterns="e24_solana_quotes/**")
df = pd.concat(map(pd.read_parquet,
                   glob.glob(f"{path}/e24_solana_quotes/**/*.parquet", recursive=True)))

Coverage

e0_run_manifest lists every collection window with its poll counts and failure counts, and is published in full rather than windowed. Gaps between windows are real, cannot be filled in afterwards, and nothing here is interpolated.

License and contact

ODC-BY: use it freely, credit "DataForge (dataforge-labs)". Questions and requests for the full history via the discussions tab.