Dealer positioning / Gamma Exposure — perp analog (Pro).

Built from market-maker accounts (the perp "dealers"): per coin we report mm_net_delta (dealer inventory long/short), gamma_profile (the MM liquidation-density by price = where MM forced unwinds cluster, the short-gamma acceleration zones) and gamma_flip (the price where MM long-liq-below crosses short-liq-above — the inflection).

Note: perps have no literal options gamma; this is a behavioral analog from market-maker inventory + liquidation-driven forced flow, not options dealer gamma. The MM set is a heuristic classification (turnover + breadth + known desks), so treat it as a positioning signal, not an exact dealer book.

Each coin also carries the composite regime flag — amplify (short-gamma analog: crowded book + dense clusters near mark → forced flow follows through) vs dampen (balanced, mean-reverting) vs transitional (on the flip line) — joining the cluster profile with funding skew, OI rate-of-change, the realized-liquidation cascade signal and the dist_to_flip_pct. Every sub-signal is returned in regime.inputs so the call is reproducible. meta.maintenance_margin_tiers gives the per-pair MM / max-leverage ladder for context.

distribution_context (per coin) puts today's readings in distributional context: trailing-30-day percentile ranks of the near-mark cluster density, |distance to flip|, normalized MM skew (mm_net_delta/mm_gross), regime score and funding rate — e.g. "BTC near-mark fuel density is the 92nd percentile of 30d". The history is forward-only (hourly samples, started 2026-07); history_days / sample_count report how much exists and every *_pctile is null while status is warming (< 2 days). meta.distribution_legend documents each field. The same block is archived in the gamma_exposure snapshot type.

2026-09-09 additions. regime.coverage (full | funding_only | none) + regime.insufficient_mm_coverage say how much evidence the state rests on (118/231 coins carry < 10 MM accounts — a funding_only state is a funding read, not a positioning read); per-coin coverage {bands, mm_n_accounts, has_flip, history_days}. regime.rules_version = "2": the skew leg now scores from distribution_context.mm_skew_pctile (≥90/≤10 one-sided, 25–75 balanced) with the raw market_skew rule as the fallback while a coin's history is warming (regime.inputs.skew_source says which). regime.since = when the current state was first committed. regime.inputs.funding_pinned flags a funding series glued to HL's 1.25e-5 interest baseline (then funding_rate_pctile is null — the rank would be meaningless). distribution_context.mm_skew_z / sample_ts. levels = the 3 nearest liquidation clusters above/below mark across EVERY account class. positions_as_of (unix ms) top-level and per coin. meta.regime_rules publishes every threshold and weight; meta.classifier_version the tagging rule-set. symbol accepts a comma list (BTC,ETH), alias coin, matched case-insensitively against Hyperliquid's names (kPEPE). An unrecognised query parameter is a 400.

meta.maintenance_margin_tiers is Hyperliquid's live margin table per coin.

Requires a Pro or Pro Plus key; full-universe (all perps) on both.