N2C (Node-to-Client)¶

::: {.module} Cardano.Node.Client.N2C.* :::

Node-to-Client protocol implementation over Unix sockets.

Connection¶

runNodeClient
    :: NetworkMagic -> FilePath
    -> LSQChannel -> LTxSChannel
    -> IO (Either SomeException ())

Opens a multiplexed connection with two mini-protocols:

MiniProtocol 6 — LocalTxSubmission
MiniProtocol 7 — LocalStateQuery

For processes that also need ChainSync on the same physical connection, use runNodeClientFull, which adds MiniProtocol 5 (ChainSync) and takes an EpochSlots and a ChainSync application:

runNodeClientFull
    :: NetworkMagic -> EpochSlots -> FilePath
    -> N2CChainSyncApplication
    -> LSQChannel -> LTxSChannel
    -> IO (Either SomeException ())

Both helpers negotiate NodeToClientV_23. The shared codec configuration (Cardano.Node.Client.N2C.Codecs) uses CardanoNodeToClientVersion19 with a configurable EpochSlots (default 42 for devnet). Run either helper in a background thread — it blocks until the connection closes.

Channels¶

Create channels before starting the connection:

lsqCh  <- newLSQChannel 16   -- 16-slot query queue
ltxsCh <- newLTxSChannel 16  -- 16-slot submit queue

LocalStateQuery¶

The LSQ client batches queries in a single acquired session. Use queryLSQ for a one-shot query, or withAcquiredLSQ / queryAcquiredLSQ to run several queries against one acquired tip:

queryLSQ :: LSQChannel -> Query Block result -> IO result

LocalTxSubmission¶

Use submitTxN2C for low-level submission (it takes a consensus GenTx Block), or prefer the high-level mkN2CSubmitter, which wraps a Conway ConwayTx automatically.

submitTxN2C
    :: LTxSChannel -> GenTx Block
    -> IO (Either (ApplyTxErr Block) ())

Reconnect and readiness¶

For long-running followers, the Reconnect supervisor retries ChainSync with full-jitter exponential backoff and gates the first attempt on the Probe LSQ tip check, so chain-sync is never attached against a node still loading its ChainDB. See Architecture and the utxo-indexer manual.