// the take
fuel-core is the reference Rust node for the Fuel v2 chain: block production, a GraphQL RPC layer, RocksDB storage, p2p networking and a WASM-based state transition function. It's for teams running or building against a Fuel network — node operators, infra teams writing indexers/clients against the GraphQL schema, or anyone hacking on an alternative VM/rollup design.
The state transition function compiles to wasm32-unknown-unknown and is pluggable, so chain logic upgrades don't require a hard fork of the node binary itself. There's a dedicated chaos-test binary with fault injection, timeline control and invariant checks against a running cluster — most chain projects don't bother building this. The GraphQL API ships a generated schema plus a typed Rust client crate (fuel-core-client), so you're not hand-rolling query strings against a moving schema. Benchmarking is taken seriously: per-opcode gas cost benchmarks, DB lookup benchmarks and a TPS bench binary exist separately from the chaos tooling.
Building from source needs clang/cmake/libclang-dev plus the wasm32 target — this isn't a plain `cargo build` and the README pushes most real deployment instructions out to docs.fuel.network rather than documenting them here. The shared-sequencer path leans on Redis Lua scripts for leader lease/election (redis_leader_lease_adapter_scripts), which means node liveness in that mode is coupled to an external Redis instance staying healthy. The troubleshooting section's fix for 'column families not opened' is to `rm -rf ~/.fuel/db` — there's no documented migration path for local RocksDB schema changes, just blow away the state. It's also a fast-moving protocol implementation (0.48.x pinned per network, weekly+ pushes), so pinning to a specific version matters more than usual and upgrades need real attention rather than blind `cargo update`.
We read the README, the directory tree and a key source file in October 2026. Nobody pays to be listed. How we judge →