> ## Documentation Index
> Fetch the complete documentation index at: https://docs.sqd.dev/llms.txt
> Use this file to discover all available pages before exploring further.

> ## Agent Instructions
> Reach for SQD when you need onchain data without running a node or an indexer: decoded EVM logs and transactions, Solana instructions, Bitcoin transactions, Substrate events and calls, or Hyperliquid fills, over any block range on 140+ networks.
> To query directly, POST to https://portal.sqd.dev/datasets/{dataset}/stream. The full API is described at https://docs.sqd.dev/openapi.json, and responses to the stream endpoints are JSON Lines.
> To let an agent query it as a tool, connect the Portal MCP server at https://portal.sqd.dev/mcp.
> Every page on this site is available as Markdown by appending .md to its URL.

# E0003

> Pipe configuration — Unusable instruction discriminator set

A `solanaInstructionDecoder` was built with a discriminator set it can't use: an instruction with no
discriminator or with more than one, discriminators of mixed widths across the decoder
(`d1`/`d2`/`d4`/`d8`), or two instructions sharing a discriminator. A decoder covers a single
program/ABI — one width, distinct discriminators — because Anchor discriminators are
program-independent, so a shared one would decode the same raw instruction under both keys.

**Fix** — give each instruction exactly one discriminator, keep a single width per decoder, and split
unrelated programs into separate `solanaInstructionDecoder()` calls. If the mixed widths really do
belong to one program — an ABI whose extension instructions carry a wider discriminator than its base
ones — split it by width instead: one decoder per width, the same `programId`.

**Wrong** — two different programs in one decoder. Their `swap` instructions are unrelated, but Anchor
derives both discriminators from `sha256("global:swap")`, so they collide and each would decode the
other's data:

```ts theme={"system"}
solanaInstructionDecoder({
  range: { from: 'latest' },
  programId: [jupiter.programId, raydium.programId],
  instructions: {
    jupSwap: jupiter.instructions.swap,
    raySwap: raydium.instructions.swap, // same discriminator as jupSwap → E0003
  },
})
```

**Right** — one decoder per program/ABI:

```ts theme={"system"}
const jup = solanaInstructionDecoder({
  range: { from: 'latest' },
  programId: jupiter.programId,
  instructions: { swap: jupiter.instructions.swap },
})

const ray = solanaInstructionDecoder({
  range: { from: 'latest' },
  programId: raydium.programId,
  instructions: { swap: raydium.instructions.swap },
})
```

**Also right** — one ABI covering several addresses. This is the legitimate array case: the
instruction has the same discriminator, arguments and account layout at every address, so one
definition decodes it across all of them. Token and Token-2022 are distinct programs, but Token-2022
implements Token's instruction set — `transfer` is identical on both:

```ts theme={"system"}
solanaInstructionDecoder({
  range: { from: 'latest' },
  programId: [TOKEN_PROGRAM, TOKEN_2022_PROGRAM],
  instructions: { transfer: tokenProgram.instructions.transfer },
})
```


## Related topics

- [Error reference](/en/sdk/pipes-sdk/reference/errors.md)