Foghorn

Network-quality judge

A composite A–F grade per indexer, fusing Foghorn's own correctness probing with The Graph's QoS oracle, on-chain stake and REO data — plus actionable verdicts and a live needs-attention triage.

What Foghorn actually tests

Most of an indexer's grade comes from network telemetry that applies to everyone; only correctnessis Foghorn's own measurement, and it only covers indexers serving the deployments Foghorn probes.

Directly probed by Foghorn — correctness

Block-pinned GraphQL queries sent through the gateway, responses canonicalised (JCS) and SHA-256 hashed; an indexer that returns minority (divergent) data versus consensus is flagged. Catches confident, well-formed wrongdata that QoS can't see.

Read from the network — applies to all indexers

QoS oracle: success rate (errors/400s), latency, chainhead lag, query volume — measured from real query traffic, not Foghorn.
On-chain / network subgraph: self-stake, allocations (coverage), REO eligibility.
Derived by Foghorn: sybil-swarm clustering and leech detection from roster patterns.

So today the composite leans on QoS / stake / coverage for most indexers; correctness is the differentiator wherever Foghorn has probed. "NR" = inactive / unrated.

Indexers graded

—

Needs attention

—

serving bad / no data

Open verdicts

—

Divergences (24h)

—

Needs Attention

Loading…

Indexer Grades

Sub-scores (top→bottom): Co Correctness · Av Availability · Fr Freshness · Cv Coverage · Va Value. Bar colour: green ≥75 · amber ≥50 · red <50.

Loading…

Verdicts

Loading…

Sybil Swarms

Non-deterministic Subgraphs

These deployments diverge across indexers every probe round — their mappings are non-deterministic (the subgraph's issue, not the indexers'). Indexers are not penalised for serving them.

Recent Divergences

Probe	Deployment	Query	Block	Clusters	Diff ops	When