Skip to content
StreamSync

How it works

Signals of speed, settled on Solana.

StreamSync is a control loop for indexing: one query, several racing operators, a verified winner, and a batched on-chain payout. Here is the full pipeline, the node roles, and how to plug in.

How it works

A pipeline tuned for tail latency.

One query races several operators. The fastest correct answer wins; verifiers confirm it; Solana settles the batch.

Customer query + prepay Gateway race dispatch top-N by region/rep speed-runner · DuckDB NNG mesh cache-optimizer · DuckDB predictive cache zk-reconstruction compressed state archive · DuckDB historical shards Verify first correct wins 70% Settle · Solana 5-min batches

The pipeline, stage by stage

01 · gateway

Race dispatch

The gateway selects the top-N operators by region, reputation, and specialization, then fires the same query to all of them at once. Customers can tune the racing set and SLA per query.

02 · NNG mesh

Sub-millisecond messaging

Operators coordinate over NNG (nanomsg-next-gen), not HTTP, keeping cross-shard cost near the kernel floor. A gossip protocol keeps membership and reputation in sync.

03 · DuckDB shards

Per-shard analytical SQL

A hash ring partitions Solana state across operator clusters. Each operator runs embedded DuckDB over its partition and merges sub-query results locally before the response returns.

04 · verify

First correct wins

Verification nodes confirm consensus on the answer. The first correct response takes 70% of the fee; verifiers split the remaining 30%. Mismatched answers are slashed from stake.

05 · settle

5-minute batches on Solana

The settlement engine sums earnings per operator and writes one batched payout to Solana every five minutes, so gas becomes a rounding error rather than a per-query tax.

Four node specializations

Operators pick the role that fits their hardware and are matched to queries that play to it.

speed-runner

low latency

Optimised to win the race on hot queries.

cache-optimizer

high memory

Predictive caching for repeat and popular queries.

zk-reconstruction

high compute

Rebuilds compressed / ZK account state from proofs.

archive

high storage

Serves deep historical shards and backfills.

Quickstart in five steps

  1. 1

    Wire the gateway

    Point your Solana client at the StreamSync gateway and set a payment token (STRM, SOL, USDC, or any whitelisted SPL) with a prepay balance.

  2. 2

    Set a racing policy

    Choose how many operators race each query and your latency SLA (default: 3–5 operators, 10ms). The gateway handles operator selection.

  3. 3

    Send queries

    Issue account lookups, decoded state reads, transaction traces, or analytical DuckDB SQL. You are billed only when a verified answer arrives under the SLA.

  4. 4

    Or stake and run a node

    Stake 10,000 STRM, run streamsync init, pick a role (speed-runner, cache, zk, or archive), and streamsync start to join the gossip mesh.

  5. 5

    Earn on reputation

    Win raced queries to earn 70% of the fee; verify others’ answers for a share. Latency, accuracy, and uptime feed a reputation score that scales future rewards.