Racing Competition
Every query is dispatched to 3–5 nodes simultaneously. First correct response wins 70% of the payment, verifiers split the rest. Slow nodes earn nothing.
Phase 1 complete · Devnet next
Solana indexing that pays on outcome, not on uptime.
StreamSync is a decentralized indexing protocol where 3–5 independent operators race to answer every query in under 10 milliseconds. If they miss the SLA, you don't pay. If they return the wrong result, they get slashed.
< 10 ms
enforced query SLA
3–5×
racing redundancy per query
193+
passing core-library tests
What you get
Features that are protocol, not promises
Sub-10ms SLA, enforced in code
If no node returns a verified answer within 10ms, the customer is not billed. The contract pays on outcome, not effort.
Distributed DuckDB sharding
A hash ring partitions Solana state across operator clusters. DuckDB merges sub-query results locally so analytical SQL stays fast even at multi-billion-row scale.
Four node specializations
ZK reconstruction, cache optimizer, speed runner, and archive. Operators pick the role that fits their hardware and get matched to queries that play to it.
Settle in STRM, SOL, USDC, or SPL
Payments aggregate into 5-minute batches and settle on Solana. Revenue splits 50% to nodes, 20% to treasury, 20% to data providers, 10% to governance — fixed in the protocol, not a vendor decision.
Slashing for incorrect answers
Verification nodes flag mismatched responses. Bad results are slashed from the staked 10,000 STRM, so accuracy isn’t a promise — it’s collateral.
No vendor lock-in by design
Multiple independent operators run from day one. If one disappears, every other operator is already serving traffic. There is no migration step.
Architecture
A pipeline tuned for tail latency
01 · gateway
Race dispatch
Selects the top-N operators by region, reputation, and specialization, then fires the same query to all of them.
02 · NNG mesh
Sub-millisecond messaging
Operators talk to each other over NNG, not HTTP, keeping the cross-shard cost near the kernel floor.
03 · DuckDB
Per-shard analytical SQL
Each operator runs DuckDB over its assigned partition. Sub-query results merge locally before the response goes back.
04 · settlement
5-minute batches on Solana
The settlement engine sums earnings per operator and writes one batched payout. Gas costs become a rounding error.
Honest comparisons
Where StreamSync fits in your stack
We benchmark against the providers people actually use. Pick what fits — and switch when the math changes.
StreamSync vs Helius
→Strong webhooks and developer UX, but a single vendor. StreamSync gives you the same query surface with multi-operator redundancy.
StreamSync vs Triton One
→Geyser-grade validator streams. StreamSync sits a layer up and gives you SQL with an enforceable latency SLA.
StreamSync vs QuickNode
→Multi-chain hosting with friendly tooling. StreamSync is Solana-native and only bills when the SLA is met.
StreamSync vs Shyft
→gRPC and webhooks on a managed stack. StreamSync mirrors the surface area but routes traffic through a competing operator marketplace.
FAQ
Questions teams ask before they switch
+ Is StreamSync a hosted RPC provider?
No. It is a protocol. You point your client at the StreamSync gateway, and queries are routed to whichever independent operators are racing fastest. There is no single company you have to trust.
+ What kinds of queries does it answer?
Account lookups, program-derived data, decoded SPL/Metaplex state, transaction traces, and analytical SQL across historical indices. The query layer is DuckDB, so anything expressible in SQL is fair game.
+ How is sub-10ms latency possible across a decentralized network?
Three things: NNG for inter-node messaging, query racing across nearest operators, and aggressive predictive caching by the cache-optimizer node class. The 10ms SLA is measured at the customer-facing gateway in the same region.
+ How do I pay?
Through the payment gateway, which accepts STRM, SOL, USDC, or any whitelisted SPL token. Customers prepay a balance; the protocol debits per successful query and settles to operators in 5-minute batches on Solana.
+ Can I run a node?
Yes. Stake 10,000 STRM, run streamsync init, pick a role (speed runner, cache, ZK, or archive), and join the network. Cooldown to unstake is 7 days. Operators are scored on latency, accuracy, and uptime, and rewards scale with reputation.
+ How does it handle compressed NFTs and ZK accounts?
A dedicated zk-reconstruction node class rebuilds compressed account state from on-chain proofs. These nodes earn a premium because the work is compute-heavy and customers value not having to run it themselves.
+ What happens if an operator goes offline mid-query?
The query was raced to 3–5 nodes. Whoever else finishes first still wins. The offline operator earns nothing and accumulates uptime debt against its reputation score, which lowers future query selection probability.
+ Is it production-ready today?
Phase 1 development is complete: core libraries (193+ tests), the STRM token program, payment gateway, settlement engine, racing competition, node specializations, gossip protocol, and cluster management. Devnet is next, mainnet follows.
Ready to stop paying for failed RPCs?
Drop in the StreamSync gateway, set a prepay balance in USDC, and let three operators race for every request. Settle the bill in five minutes.