Skip to main content

​
Roadmap

Aligns with The Graph’s 2026 Technical Roadmap (“Experimental JSON-RPC Data Service”, Q3 2026).

​
MVP ✅ Complete

The full working system is shipped. The goal was a minimal, stable payment loop — no dead weight. Payment infrastructure
  • RPCDataService.sol — register, startService, stopService, collect
  • paymentsDestination — decouple payment recipient from operator key
  • TAP receipt validation, RAV aggregation, on-chain collect()
  • Integration tests against real Horizon payment contracts (mock staking only)
  • EIP-712 cross-language compatibility tests (Solidity ↔ Rust)
Gateway & routing
  • QoS routing — latency EMA (35%) + availability (35%) + block freshness (30%)
  • Capability tiers — Standard / Archive / Debug; per-method tier detection
  • Archive tier routing — inspects block parameters (hex numbers, "earliest")
  • debug_* / trace_* routing — per-chain capability map
  • 10+ chains — Ethereum, Arbitrum, Optimism, Base, Polygon, BNB, Avalanche, zkSync Era, Linea, Scroll
  • CU-weighted pricing — 1–20 CU per method
  • Geographic routing — region-aware score bonus
  • Cross-chain unified /rpc endpoint — chain via X-Chain-Id header
  • JSON-RPC batch support
  • WebSocket subscriptions — eth_subscribe / eth_unsubscribe proxied bidirectionally
  • Per-IP rate limiting, Prometheus metrics
Verification
  • Response attestation — provider signs every response with operator key; gateway verifies before forwarding
  • Quorum consensus — deterministic methods sent to 3 providers; majority wins; disagreements logged
Discovery & operations
  • RPC network subgraph — indexes RPCDataService events for dynamic provider discovery
  • Indexer agent — TypeScript; automates provider lifecycle (register, startService, stopService)
  • Consumer SDK — TypeScript; receipt signing, provider discovery, QoS-weighted selection
  • Docker Compose full-stack deployment
  • CI (Rust + Solidity)

​
Not planned

These were considered and deliberately excluded.
FeatureWhy
slash() / fraud proofsRPC responses have no canonical on-chain truth to slash against
Block header trust oracleDependency of slashing; dropped with it
EIP-1186 MPT proof verificationSame
Permissionless chain registrationGovernance allowlist is sufficient
GRT issuance / rewards poolOut of scope for this data service

​
Potential future work

  • TEE-based response verification — cryptographic correctness via trusted execution; requires enclave hardware and a security audit
  • P2P SDK — gateway-optional model; consumer connects directly to provider
  • Permissionless chain registration — bond-based governance; deferred until the allowlist becomes a bottleneck