`spot_meta_and_asset_ctxs` doesn't inject dex into the payload

[satwikkansal]

    def spot_meta_and_asset_ctxs(self) -> SpotMetaAndAssetCtxs:
        """Retrieve exchange spot asset contexts
        POST /info
        Returns:
            [
                {
                    universe: [
                        {
                            tokens: [int, int],
                            name: str,
                            index: int,
                            isCanonical: bool
                        },
                        ...
                    ],
                    tokens: [
                        {
                            name: str,
                            szDecimals: int,
                            weiDecimals: int,
                            index: int,
                            tokenId: str,
                            isCanonical: bool
                        },
                        ...
                    ]
                },
                [
                    {
                        dayNtlVlm: float string,
                        markPx: float string,
                        midPx: Optional(float string),
                        prevDayPx: float string,
                        circulatingSupply: float string,
                        coin: str
                    }
                    ...
                ]
            ]
        """
        return cast(SpotMetaAndAssetCtxs, self.post("/info", {"type": "spotMetaAndAssetCtxs"}))

While the Info class supports perp_dexs constructor argument, when I fetch spot_meta_and_asset_ctxs the dex is not passed in the payload, nor the method supports a dex argument

[massimiliano1991]

I looked into this and measured it live today (2026-09-05, mainnet). The short version: the API ignores dex on spotMetaAndAssetCtxs, so injecting it into the payload would be a silent no-op rather than a fix. The reason is that HIP-3 partitions perp markets, not spot.

Spot is not dex-partitioned — verified. The response is structurally identical whether dex is absent, empty, a made-up string, or a real HIP-3 dex name:

import json, urllib.request
def post(p):
    r = urllib.request.Request("https://api.hyperliquid.xyz/info",
        data=json.dumps(p).encode(), headers={"Content-Type": "application/json"})
    return json.load(urllib.request.urlopen(r))

def structure(r):  # drop volatile prices, keep the shape
    meta, ctxs = r
    return ([(u["name"], u["index"]) for u in meta["universe"]],
            [(t["name"], t["index"]) for t in meta["tokens"]],
            sorted(c["coin"] for c in ctxs))

base = structure(post({"type": "spotMetaAndAssetCtxs"}))
for dex in ("test", "vntl"):
    same = structure(post({"type": "spotMetaAndAssetCtxs", "dex": dex})) == base
    print(f"dex={dex!r:8} identical structure: {same}")
# dex='test'   identical structure: True
# dex='vntl'   identical structure: True
# universe=326  tokens=499  coins=718 in all cases

Contrast — the perp endpoints do respond to dex, which is why the request feels natural but the analogy doesn't hold for spot:

def uni(p): return [u["name"] for u in post(p)["universe"]]
print(len(uni({"type": "meta"})))                       # 233  (BTC, ETH, ...)
print(len(uni({"type": "meta", "dex": "xyz"})))         # 119  (xyz:XYZ100, xyz:TSLA, ...)
print(len(uni({"type": "meta", "dex": "flx"})))         # 16   (flx:TSLA, ...)

Each HIP-3 perp DEX carries its own perp universe; the global spot universe (spotMetaAndAssetCtxs) is shared and has no per-dex view to select. So dex is meaningful on the perp meta/metaAndAssetCtxs/user_state/open_orders family (that gap is being tracked in #278 and documented in #286/#320), but on spotMetaAndAssetCtxs there is nothing for it to switch.

My suggestion: rather than add a dex argument that the API silently drops (which would mislead callers into thinking a per-dex spot view exists), it's more accurate to leave the signature as-is and, if anything, add a one-line docstring note that spot contexts are global and not dex-scoped. Happy to open that docs PR if a maintainer agrees.

(Observed: the identical-structure and count results above, from live mainnet calls. Deduced: the HIP-3 perp-vs-spot partitioning as the cause — consistent with the perp-meta results but I haven't seen it stated in the docs.)