The Guard is an agent responsible for verifying actions from other Agents in the optimistic messaging model. This includes polling for invalid states and attestations as well as submitting fraud reports.
The Guard operates with four main components:
streamLogs is the data-getter for the Guard. It works by instantiating a gRPC connection to Scribe, and puts logs in a channel for the origin and destination contracts on each chain in the config. From here, it verifies the logs’ order since the order of logs are very important for merkle tree construction.
Additionally, if the Guard unexpectedly stops in the middle of streaming logs, it will use the current database state to reconstruct the tree up to where the last log was, then continue to use gRPC.
receiveLogs is the data-processor for the Guard. It works by taking the logs streamed from streamLogs and parsing the logs into either a Message on the Origin.sol contract, or a Attestation on the Destination.sol contract. It then stores the data into the Guard’s database and builds the tree accordingly.
loadOriginLatestStates polls Origin states and caches them in order to make the latest data available.
submitLatestSnapshot fetches the latest snapshot from Origin and submits it on Summit.
updateAgentStatuses polls the database for RelayableAgentStatus entries and calls updateAgentStatus() once a sufficient agent root is passed to the given remote chain.
The fraud reporting logic can be found in fraud.go, which consists mostly of handlers for various logs. The two major handlers are handleSnapshotAccepted and handleAttestationAccepted, both of which verify states corresponding to the incoming snapshot/attestation, initiate slashing if applicable, and submit state reports to eligible chains.
Navigate to sanguine/agents/agents/guard/main and run the following command to start the Guard:
$ go run main.go
Then the Guard command line will be exposed. The Guard requires a gRPC connection to a Scribe instance to stream logs. This can be done with either a remote Scribe or an embedded Scribe.
For more information on how to interact with Scribe see the Executor README.
Guard ├── main: API server ├── cmd: CLI commands ├── db: Database interface │ └── sql: Database writer, reader, and migrations ├── main: CLI entrypoint