SolanaVault is a decentralized (DePIN) storage network for Solana blockchain data. A multi-stage compression pipeline shrinks blocks and account history 15-25x at sub-millisecond encode latency; NNG transport, Kademlia DHT discovery, and Byzantine consensus serve them back through the standard Solana JSON-RPC API — the read surface never changes. As throughput growth turns state and history into a shared infrastructure problem, SolanaVault pushes archival off the critical path: run a light client and pay per read, or operate a gateway / storage node on cheap commodity storage and earn for what you serve. The whole workspace is MIT-licensed — clone vault-core, vault-node, vault-rpc-decentralized, vault-cli and bring up your own network.
# Start the light client — drop-in replacement for mainnet-beta
$ ./target/release/vault-light-client start --balance 50000
[INFO] bootstrapping kademlia dht ... 217 peers
[INFO] gateway selected: vault://us-east-2.gw.cryptuon
[INFO] json-rpc listening on http://localhost:8899
# Point any Solana client at the local endpoint
$ curl -s http://localhost:8899 -d '{"jsonrpc":"2.0","id":1,"method":"getConfirmedBlock","params":[245000000]}'
{"jsonrpc":"2.0","id":1,"result":{ ... block 245000000 (146 bytes wire, decompressed in 28μs) ... }}
Throughput growth turns Solana state and history into a shared infrastructure problem. SolanaVault answers it with a DePIN storage network — every line of the stack is real, public, and benchmarked, from the same Rust workspace that powers your gateway.
Versioned strategies (v1, v2, v3, hybrid, production_v3) and an XGBoost-tuned variant adapt to block payloads in real time. Real Solana mainnet blocks and account history compress 15-25x at sub-millisecond encode latency — the same tonnage of state at a fraction of the storage bill.
The storage layer changes; the read surface does not. The light client exposes the same JSON-RPC as mainnet-beta — getConfirmedBlock, getTransaction, getSignaturesForAddress. Swap your endpoint, keep your client libraries, ship today.
A network of independent storage operators, not a data center. Binary messaging over NNG sockets gives microsecond inter-node latency; Kademlia DHT routes content addresses to the closest replica, with automatic failover and reputation-aware peer selection. No central authority owns the data.
Every block served carries a cryptographic proof verified by a Byzantine quorum. Liars get slashed and lose reputation. Honest storage nodes earn consensus rewards on every retrieval.
Light clients deposit a balance and pay per query. Gateway operators earn the bulk of that revenue; storage and consensus participants are rewarded by the staking + reward-distribution layer based on uptime, success rate, and verified storage proofs.
The memory subsystem combines RocksDB persistence with a tiered cache manager and a memory pool sized for production. Hot blocks stay in RAM; warm blocks live in compressed pages; cold blocks come from the network.
tracing + tracing-subscriber for structured logs, Prometheus metrics via metrics-exporter-prometheus, and a dashboard API that powers both the TUI and web consoles. Governor-based rate limiting at every endpoint.
SolanaVault decouples the work of serving block data from the work of storing it, so each role can be priced and scaled independently.
┌─────────────┐ ┌──────────────┐ ┌─────────────┐
│ Application │───▶│ Light Client │───▶│ Gateway │
│ (unchanged) │ │ (pay-per-use)│ │ (earns 85%) │
└─────────────┘ └──────────────┘ └─────────────┘
│
▼
┌─────────────────────────┐
│ SolanaVault P2P Network │
├──────────┬──────────────┤
│ NNG │ Kademlia DHT │
│ Transport│ Discovery │
├──────────┴──────────────┤
│ Byzantine Consensus │
│ + Reputation Scoring │
├─────────────────────────┤
│ Compression Pipeline │
│ v1 → v2 → v3 → hybrid │
│ + XGBoost adaptation │
├─────────────────────────┤
│ RocksDB Storage Engine │
│ + Tiered Cache Manager │
└─────────────────────────┘We respect the teams we compete with. Pick the tool that fits — and read the full breakdown before you commit.
Hosted Solana RPC and DAS API with strong DX, but priced per request and centralized on Helius infrastructure.
High-end Solana RPC with custom hardware and Geyser feeds. Excellent performance, still operator-centric and expensive at scale.
Multi-chain RPC provider with a polished console. Pricing is per-request and per-feature, and historical access is gated by plan.
Filecoin/IPFS-backed Solana historical archive. Strong for cold retrieval, weaker for live RPC and not optimized for sub-second queries.
It is genuinely peer-to-peer. Peer discovery uses Kademlia DHT, transport uses NNG sockets, and data integrity is enforced by Byzantine consensus. There is no central scheduler, no privileged registry, and no single party that can revoke access.
No. SolanaVault exposes the standard Solana JSON-RPC interface. Replace https://api.mainnet-beta.solana.com with http://localhost:8899 (or your gateway URL) and existing clients — Anchor, web3.js, solana-py — work unchanged.
On production Solana blocks we observe 15-25:1 typical ratios, with our v3 pipeline regularly exceeding 20:1. Compression takes under a millisecond per block; decompression takes 13 to 85 microseconds. The README shipping ratios are reproducible via vault-cli compress-demo.
Pay-per-use. Light clients deposit a balance (vault-light-client start --balance ...) and pay each gateway per query they route. Gateways keep the majority of revenue collected; storage and consensus participants are rewarded for contributing bandwidth, storage capacity, uptime, and verified proofs.
Yes. The workspace is MIT licensed. cargo build --release produces vault-light-client, vault-node, vault-rpc-decentralized, and vault-cli. You can run a private SolanaVault network end-to-end with zero managed dependencies — there is no central scheduler, registry, or coordinator to depend on.
A full archive node stores 100% of raw history on fast disk and carries significant storage and IOPS cost — you pay for all the data even if you read a fraction of it. A SolanaVault light client pays only for the blocks it actually fetches and inherits 95%+ storage savings from the compression pipeline. You retain the full Solana JSON-RPC surface and gain microsecond-range decompression (13-85us per block) on cached responses. If you already run archive nodes at scale and are happy with the cost, the win is smaller; the gains are largest for teams that read history heavily but do not want to own an archive node.
Instead of every team paying to run — or renting from someone who runs — a full archive node, a decentralized physical infrastructure (DePIN) network of independent storage operators holds compressed Solana data cooperatively, earns for serving reads, and exposes it through the same RPC methods callers already use. Durability is paid for by an incentive layer, not a single data center.
Solana produces more state per unit time than any other public chain that matters. As throughput keeps climbing, the account and history data that throughput produces grows faster than any single operator can economically hold on fast disk. That is state bloat: the compounding tax on every indexer, explorer, and read pipeline. SolanaVault treats that state as shared infrastructure — compressed 15-25x and served from a P2P operator set — rather than a per-operator cost center.
Agentic payments, verifiable/on-chain AI, and high-frequency RWA settlement generate machine-speed read traffic against historical and account data — exactly the data that lives on expensive archive nodes today. Sub-millisecond retrieval over a decentralized operator set is the kind of cheap, fast, non-single-vendor read layer that agent workloads need. SolanaVault is the storage-and-retrieval layer beneath the RPC those workloads already speak, not a new agent framework.
SolanaVault is under active development. It is a storage-and-retrieval layer, not a new L1, VM, or consensus replacement. Durability/replication guarantees, the on-chain incentive layer, and data-availability proofs are still hardening toward production — see the ROADMAP in the repository for the milestones and the "cheapest path to production" plan. The compression ratios, sub-millisecond reads, and drop-in RPC pass-through are demonstrable today; the production-grade DePIN guarantees are the work in progress.
The Byzantine consensus layer cross-checks responses across multiple replicas. Mismatches reduce the offending node's reputation. After a threshold the node is evicted from the routing tables and its stake is slashed via the economics layer (staking + reward distribution).
Not today. SolanaVault ships as an open-source workspace: vault-light-client, vault-rpc-decentralized, vault-node, and vault-cli. Anyone can operate a gateway, anyone can run a light client, and anyone can stand up a private network. The repo is the product.
Every part of SolanaVault is documented in the open — pick a role, read the engineering deep-dives, or see the honest comparisons.
Indexers & analytics, dApp/wallet backends, gateway operators, and storage/consensus nodes — find your role in the network.
Compression pipeline, drop-in Solana RPC, NNG transport, Byzantine consensus, gateway economics, and RocksDB caching in detail.
The four roles, NNG + Kademlia DHT discovery, Byzantine consensus, the compression pipeline, and the RocksDB storage engine.
Head-to-head breakdowns vs Helius, Triton One, QuickNode, and Old Faithful — read before you commit.
Decentralization, code changes, compression ratios, operator pay, self-hosting, and data integrity.
The engineering philosophy and the team building the SolanaVault DePIN storage network.
A deep walkthrough of the multi-stage compression that ships in vault-core, including the v3 production pipeline and the XGBoost-tuned variant.
How the consensus and reputation system protects clients from malicious or buggy storage nodes.
Moving a real production indexer from a hosted RPC provider to SolanaVault — the timeline and the surprises.
What it takes to run a profitable gateway in 2026 — hardware, networking, reputation, and operator economics.
The argument that compression — not consensus, not transport — makes decentralized blockchain storage economically inevitable.
Stop paying to own — or renting from someone who owns — a full archive node. Clone the workspace, cargo build --release, and run vault-light-client against the live DePIN network. MIT, no managed dependencies.