Onchain storage

ABX uses reader contracts for content too large to keep in a normal storage slot. A metadata field holds the reader address and a pointer. The reader returns the original bytes.

Chunk store

AbxChunkStore writes bytes with SSTORE2. The payload lives in contract bytecode, behind a small data contract with no mutable storage.

One contract has a practical bytecode limit, so larger payloads are split across pointers. The reader joins them in order.

interface IAbxOnChainReader {
    function read(address pointer) external view returns (bytes memory);
}

The pointer names an SSTORE2 manifest. That manifest names the content chunks, which the reader joins in order.

Compression

ABX supports two different cases:

• reader-gzip stores compressed bytes that an offchain resolver expands.
• Onchain program templates may include a browser decompressor in the assembled document.

A Solidity metadata renderer cannot gunzip arbitrary bytes. Use plain reader for content that must be returned directly from an onchain tokenURI.

Limits and cost

Writing bytecode costs gas once. Reading and assembling it costs RPC execution on every view call. Large documents can pass deployment limits and still exceed a provider's eth_call limit.

Use the CLI's dry run and gas report before choosing this route:

abx deploy ./work.png --onchain-image --dry-run

There is no universal RPC ceiling. A document that works through one endpoint may fail through another. Code projects provide piecewise reads for large generated documents.

Onchain storage is a good fit when permanence and chain-only recovery justify the cost. IPFS or Arweave is usually simpler for large media.

Read and verify

The SDK reads the pointer, calls the reader, joins chunks, and decompresses when the representation requires it. A resolver can then serve those bytes over HTTP without changing their custody.

A field lock freezes the reader address and pointer. Use immutable readers and data contracts if the project claims the bytes are fixed.

See Storage for upload backends and Metadata for field representations.