A blockchain-native compression engine with negotiated models for each serialization family.
ZChain is Zetako's proprietary lossless compression technology for blockchain data paths. This public repository documents the evolution of the product, benchmark methodology, blockchain-specific use cases, reproducible results, and the boundary between measured evidence and future integration work.
ZChain is not being developed as a generic “compress everything” library. A blockchain runtime already knows whether it is handling a block, receipt, RLP payload, signature, account key or consensus result. The product direction is to use that knowledge instead of spending CPU rediscovering it.
Goal: reduce bytes moved or stored by blockchain infrastructure while keeping reconstruction exact and processing cost low enough for the path being optimized.
The public history is intentionally visible. Each generation exists because the previous one exposed a concrete limitation.
| Public generation | Engineering name | What changed | Evidence / delta |
|---|---|---|---|
| ZChain 3 — Foundation | v3 | Original context-adaptive lossless codec | Public compatibility baseline |
| ZChain 4 — Fast Compatible | v4 | C hot-path optimization without changing tested v3 output | 74.27 → 113.36 MiB/s encode on the 64-file public corpus; same compressed byte count |
| ZChain 5 — Speed | Speed_First | New speed-oriented stream, lighter model, RAW bypass | 126.53 MiB/s encode / 74.67 MiB/s decode on the same corpus |
| ZChain 6 — Blockchain | ZCB2 / blockchain profiles | First blockchain-specific routing and Ethereum hex specialization | Established that crypto-heavy fields should not always be treated like ordinary structured bytes |
| ZChain 7 — Schema | ZCB3 | Runtime-assisted structure / opaque / reference model | Architecture validated in schema-assisted research microbenchmarks |
| ZChain 8 — Blockchain Engine | current engine | Negotiated models, stable public C ABI, ZCB1 frames, installable SDK, multi-chain registry and direct serializer adapters | Native Ethereum RLP schema model now SUPPORTED |
v3 / v4
structured bytes
↓
context model + arithmetic coding
Speed
↓
same concept, optimized for throughput
Blockchain
↓
payload-aware profiles
Schema
↓
structure | opaque crypto | known references
Blockchain Engine
↓
negotiated model selected by the blockchain runtime
The product therefore evolved from a standalone codec into a blockchain compression engine with explicit integration contracts.
ETHEREUM_SCHEMA is now SUPPORTED through a direct Alloy/Reth-compatible adapter. The adapter works from native Alloy transaction objects and calls zchain_bc_stream_* directly — no .ops replay file and no RLP parser inside the timed codec path.
Reference corpus:
- 24 real Ethereum blocks;
- 5,246 native Alloy transactions;
- legacy, EIP-1559, EIP-4844 blob and EIP-7702 transaction mix;
- 2,959,206 native RLP bytes.
| Codec | Output bytes | Savings | Encode | Decode |
|---|---|---|---|---|
| ZChain 4 — Fast Compatible | 1,102,352 | 62.75% | 79.50 MiB/s | 46.37 MiB/s |
| ZChain 5 — Speed | 1,102,219 | 62.75% | 91.24 MiB/s | 55.14 MiB/s |
ZChain 8 — ethereum-rlp compatible |
1,102,352 | 62.75% | 74.23 MiB/s | 46.12 MiB/s |
ZChain 8 — ETHEREUM_SCHEMA |
1,070,680 | 63.82% | 108.24 MiB/s | 75.51 MiB/s |
Against ZChain 5 Speed on the same RLP bytes, ZChain 8 Schema is 2.86% smaller, 18.6% faster to encode, and 36.9% faster to decode in codec-only mode.
The end-to-end adapter measurement is reported separately: 58.28 MiB/s serializer→frame encode and 74.02 MiB/s decode. This keeps codec throughput distinct from Alloy serialization and span-marking cost.
Open the direct Alloy/Reth supported benchmark →
The existing Apple M4 native C reference run remains the current public summary for Ethereum JSON, Solana RPC, CometBFT and Agave ledger-source payloads.
| Corpus | Payload | Raw bytes | ZChain bytes | Savings | Encode | Decode |
|---|---|---|---|---|---|---|
| Ethereum Reth JSON | ethereum-hex |
2,908,507 | 512,808 | 82.37% | 131.11 MiB/s | 114.69 MiB/s |
| Ethereum Reth JSON | ethereum-block |
2,908,507 | 559,234 | 80.77% | 91.83 MiB/s | 57.70 MiB/s |
| Solana mainnet RPC JSON | solana-rpc |
26,106,696 | 4,493,986 | 82.79% | 110.40 MiB/s | 67.93 MiB/s |
| CometBFT CosmosHub RPC JSON | cometbft |
968,332 | 295,048 | 69.53% | 67.29 MiB/s | 43.19 MiB/s |
| Agave ledger source | solana-shred |
879,258 | 222,165 | 74.73% | 81.81 MiB/s | 54.38 MiB/s |
Open the full multi-chain M4 benchmark report →
The negotiated ETHEREUM_HEX profile remains the supported model for Reth Ethereum JSON payloads.
| Path | Raw bytes | Final bytes | Savings | Encode | Decode |
|---|---|---|---|---|---|
Previous ethereum-block path |
2,908,507 | 559,234 | 80.77% | 91.83 MiB/s | 57.70 MiB/s |
Negotiated ethereum-hex |
2,908,507 | 512,808 | 82.37% | 131.11 MiB/s | 114.69 MiB/s |
On this run, the negotiated specialization is 8.30% smaller, 1.43× faster to encode, and 1.99× faster to decode than the previous Ethereum path.
Every page answers the same questions: what do we compress, why does it matter, how is it measured, what is proven today, and what comes next?
| Ecosystem | Current public status | Page |
|---|---|---|
| Ethereum / Reth | ETHEREUM_HEX and direct Alloy/RLP ETHEREUM_SCHEMA SUPPORTED |
Ethereum |
| EVM L2 / EVM-compatible | Ethereum JSON-hex model reusable; chain-specific benches pending | EVM L2 |
| Solana | Real mainnet RPC corpus measured; RPC profile supported | Solana |
| Agave | Ledger-source profile measured; validator-side schema integration remains research | Agave |
| Cosmos / CometBFT | Real CosmosHub RPC corpus measured; CometBFT profile supported | Cosmos / CometBFT |
| Bitcoin | Native block / transaction baseline established; specialized model next | Bitcoin |
| Substrate / Cardano / Sui / Aptos / TRON / others | Adapter-ready or research models | Model matrix |
Depending on the chain and integration boundary, the recurring targets are:
- RPC responses — blocks, receipts, transactions, traces, account data;
- native serialization — RLP and other chain-specific byte formats where the runtime already knows field identity;
- storage and archive paths — blocks, receipts, snapshots, exports, indexer datasets;
- state sync / snapshot movement — large structured batches where byte reduction can offset codec CPU;
- service-to-service transfer — when an explicit negotiated format is appropriate;
- validator or execution-client sidecars — shadow measurement before changing a critical path.
Compression savings on a benchmark payload are not automatically equivalent to mainnet bandwidth or database savings. Production claims require measurement in the actual path.
ZChain uses explicit maturity labels:
- SUPPORTED — specialized model with real-corpus evidence and an integration contract;
- ADAPTER_READY — serialization/integration boundary is defined, but specialized real-corpus validation is still required;
- RESEARCH — active model work or architecture experiment.
Public benchmark pages follow the same rules:
- lossless byte-for-byte round trip;
- exact workload and serialization named;
- native codec timing separated from I/O, hashing and process startup;
- compression and decompression throughput reported separately;
- serializer→frame measurements reported separately from codec-only throughput;
- build profile and hardware stated;
- synthetic architecture tests labeled as synthetic;
- no chain-specific production claim without a real chain-specific corpus.
ZChain is compared against selected established codecs — currently gzip, Brotli and LZMA2 — on the same documented payloads and named presets.
The comparison focuses on the tradeoff that matters for blockchain infrastructure: compressed bytes versus encode/decode cost.
/
├── README.md product evolution and headline results
├── docs/
│ ├── ETHEREUM.md Ethereum / Reth
│ ├── EVM_L2.md Base / Arbitrum / Optimism / Polygon / BNB / Avalanche
│ ├── SOLANA.md Solana mainnet data
│ ├── AGAVE.md Agave validator integration
│ ├── COSMOS_COMETBFT.md Cosmos / CometBFT
│ ├── BITCOIN.md Bitcoin model
│ ├── MODEL_MATRIX.md serialization-family maturity
│ ├── METHODOLOGY.md benchmark rules
│ ├── COMPRESSION_COMPARISON.md ZChain vs other codecs
│ ├── ZCHAIN_8_ETHEREUM_SCHEMA_DIRECT_ALLOY_SUPPORTED_REPORT.md
│ └── evidence / legacy reports detailed engineering evidence
└── assets/ readable public charts
Detailed historical reports remain in the repository as evidence even when a shorter product-facing page links to them.
Published here: methodology, benchmark reports, payload descriptions, product evolution, compatibility behavior, integration architecture and claim boundaries.
Kept private: proprietary codec core source, protected production releases, internal test infrastructure, unpublished datasets and customer-specific integrations.
For technical evaluation, partnership or licensing discussions: contact@zetako.ai
© Zetako. Proprietary technology. Public documentation in this repository does not grant rights to reproduce, reverse engineer or redistribute the protected ZChain implementation.