Indexer External Calls

External Calls (Effect API)

Handlers run twice: parallel preload pass (warms caches), then sequential pass (state changes). All external I/O (fetch, RPC, APIs) MUST go through createEffect + context.effect() — otherwise it double-executes and blocks parallelization.

Define and call

import { S, createEffect } from "envio";

const getOwner = createEffect(
  {
    name: "getOwner",
    input: { tokenId: S.bigint },
    output: S.union([S.string, null]),
    cache: true,
    rateLimit: false,
  },
  async ({ input }) => {
    const res = await fetch(`https://api.example.com/owner/${input.tokenId}`);
    return res.json();
  }
);

indexer.onEvent(
  { contract: "Token", event: "Transfer" },
  async ({ event, context }) => {
    const owner = await context.effect(getOwner, { tokenId: event.params.tokenId });
  }
);

Pass minimum input

input carries only what varies per call. Bake static config (URLs, tokens, channel IDs, env vars) into the effect body. Build payloads/strings inside the effect.

// ❌ input: { url, chatId, text }   — config and pre-built strings leak into the call site
// ✅ input: { usd, blockNumber }    — only the values that vary per call

Dedup is keyed by hash of input; leaner inputs dedupe better, validate faster, and let one effect serve many call sites.

Schema (S)

S.string, S.number, S.bigint, S.boolean, S.schema({ ... }), S.array(...), S.union([..., null]), S.optional(...). Full ref: https://raw.githubusercontent.com/DZakh/sury/refs/tags/v9.3.0/docs/js-usage.md

RPC pattern (viem)

Always pass { batch: true } to the transport. The preload pass fires effects for the whole batch concurrently, and batching collapses them into a few JSON-RPC requests instead of one request per read.

const client = createPublicClient({
  transport: http(process.env.ENVIO_RPC_URL, { batch: true }),
});

const getTokenMetadata = createEffect(
  {
    name: "getTokenMetadata",
    input: S.string,
    output: { name: S.string, symbol: S.string, decimals: S.number },
    cache: true,
    rateLimit: false,
  },
  async ({ input: address }) => {
    const args = { address: address as `0x${string}`, abi: ERC20_ABI };
    const [name, symbol, decimals] = await Promise.all([
      client.readContract({ ...args, functionName: "name" }),
      client.readContract({ ...args, functionName: "symbol" }),
      client.readContract({ ...args, functionName: "decimals" }),
    ]);
    return { name, symbol, decimals: Number(decimals) };
  }
);

Options

crossChain (per-chain caching)

crossChain decides whether an input is cached and rate-limited once globally or once per chain. Default true, or false when config.yaml sets disable_default_cross_chain: true — see the indexer-multichain skill.

• crossChain: true (default): one shared cache. Use for chain-agnostic calls — the result is the same whatever chain asked (token metadata, a price by symbol).
• crossChain: false: a separate cache and rate-limit budget per chain, and context.chain.id becomes available in the handler. Use when the result depends on the chain — an on-chain read, a per-chain RPC. The same input is fetched once per chain instead of once globally.

// Result depends on the chain -> crossChain: false, read context.chain.id
const getBalance = createEffect(
  { name: "getBalance", input: S.string, output: S.bigint, rateLimit: false, crossChain: false },
  async ({ input: account, context }) => rpcFor(context.chain.id).getBalance(account)
);

A cross-chain effect can only call other cross-chain effects (it has no single chain to resolve a chain-scoped call against); a chain-scoped effect can call either. Reading context.chain on a cross-chain effect throws. Changing an effect's crossChain or name starts a fresh cache.

If something is unclear, use the envio-docs skill to search and read the latest documentation.