Linea prover configuration reference

Advanced options

Some configuration keys are operational or environment-specific and are included for completeness. Their presence does not indicate that changing them is recommended or supported for every deployment. Verify their applicability against the relevant release and deployment guidance.

Defaults

Defaults are values defined in the prover Go source (config_default.go). They are not recommended production settings and may be overridden by deployment configuration. Sample config-*.toml files are never used as the source of defaults.

This reference lists Linea prover TOML configuration keys, grouped by section. Descriptions come from Go doc comments on the config structs; defaults are shown in the Default column.

Generated from LFDT-Lineth/lineth-monorepo@555fa72.

Options may differ in other Linea prover releases.

Top-level

Config key	Description	Default	Type	Allowed / required
assets_dir	AssetsDir stores the root of the directory where the assets are stored (setup) or accessed (prover).	—	string
environment	Environment is informational only — it indicates which environment (e.g. devnet, sepolia, mainnet) the config file targets. It is not used for setup paths or any runtime logic; all environments share the same setup and are differentiated only by layer2 public inputs.	—	string
log_level	LogLevel sets the log level for the logger.	—	uint8	required; gte=0; lte=6
version	if we supply as is to coordinator in responses, coordinator should parse semver	—	string	required; semver

Controller

Config key	Description	Default	Type	Allowed / required
controller.defer_to_other_large_codes	List of exit codes for which the job will put back the job to be reexecuted in large mode.	[137]	[]int
controller.enable_aggregation	EnableAggregation controls whether aggregation jobs are picked up.	true	bool
controller.enable_data_availability	EnableDataAvailability controls whether data-availability jobs are picked up.	true	bool
controller.enable_execution	EnableExecution controls whether execution jobs are picked up.	true	bool
controller.enable_invalidity	EnableInvalidity controls whether invalidity jobs are picked up.	true	bool
controller.prometheus.enabled	Enabled toggles the Prometheus metrics endpoint on or off.	—	bool
controller.prometheus.port	Port for the metrics endpoint; defaults to 9090 (Prometheus library default, not set by the prover).	—	int
controller.prometheus.route	The default implementation default to /metrics. The route should be prefixed with a "/". If it is not, the underlying implementation will assign it.	—	string
controller.retry_delays	The delays at which we retry when we find no files in the queue. If this is set to [0, 1, 2, 3, 4, 5]. It will retry after 0 sec the first time it cannot find a file in the queue, 1 sec the second time and so on. Once it reaches the final value it keeps it as a final retry delay.	[0, 1, 2, 3, 5, 8, 13, 21, 44, 85]	[]int
controller.retry_locally_with_large_codes	List of exit codes for which the job will retry in large mode	[77, 333, 2]	[]int
controller.spot_instance_reclaim_time_seconds	The number of seconds infra (AWS) waits before reclaiming a spot instance	120	int
controller.termination_grace_period_seconds	The number of seconds the controller should wait before killing a worker after receiving a SIGTERM	2700	int
controller.worker_cmd_large_tmpl	WorkerCmdLarge is the template string for spawning a large-mode worker process.	prover prove --config \{\{.ConfFile\}\} --in \{\{.InFile\}\} --out \{\{.OutFile\}\} --large	string
controller.worker_cmd_tmpl	WorkerCmd is the template string for spawning a standard worker process.	prover prove --config \{\{.ConfFile\}\} --in \{\{.InFile\}\} --out \{\{.OutFile\}\}	string

Execution

Config key	Description	Default	Type	Allowed / required
execution.can_run_full_large	CanRunFullLarge indicates whether the prover is running on a large machine (and can run full large traces).	—	bool
execution.conflated_traces_dir	ConflatedTracesDir stores the directory where the conflation traces are stored.	/shared/traces/conflated	string	required
execution.ignore_compatibility_check	IgnoreCompatiblityCheck indicates whether to ignore constaints version checking between trace files and zkevm.bin constraint files. Specifically, this check ensures that the zkevm.bin file used within the prover was generated from the same commit of linea-constraints as the generated lt trace file. Set this to true to disable compatibility checks (default: false).	false	bool
execution.keep_traces_until_block	KeepTraceUntil is an optional parameter (default to 0) that indicates a maximum block height under which the prover will not delete the execution traces from EFS. The block height that is compared to this value is the "end" block of the request range.	—	int
execution.limitless_with_debug	LimitlessWithDebug is only looked at when the limitless prover is activated. When set to true, the limitless prover will only run in debug mode and not produce any proof. This is useful to investigate bugs in the limitless prover. The field is optional and defaults to false.	—	bool
execution.prover_mode	ProverMode stores the kind of prover to use.	—	string	required; dev|partial|full|proofless|bench|check-only|limitless
execution.requests_root_dir	RequestsRootDir is the root directory for incoming job requests and completed results.	/shared/prover-execution	string	required
execution.serialization	Serialization enables loading the inner circuit (wizard IOP) from a serialized file on disk via memory-mapping, instead of compiling it at proving time. The file is produced during setup by protocol/serde.	false	bool

Data availability

Config key	Description	Default	Type	Allowed / required
data_availability.dict_nb_bytes	DictNbBytes is the number of bytes in the prover dictionary.	65536	int	gt=0
data_availability.dict_paths	DictPaths points to dictionaries used for decompressing executions data stored in the DA layer.	—	[]string
data_availability.max_nb_batches	MaxNbBatches is the maximum number of execution batches allowed in a single DA proof.	100	int	required; gt=0
data_availability.max_uncompressed_nb_bytes	MaxUncompressedNbBytes is the maximum number of bytes in the uncompressed blob payload.	—	int	required; gt=0
data_availability.prover_mode	ProverMode stores the kind of prover to use.	—	string	required; dev|full
data_availability.requests_root_dir	RequestsRootDir is the root directory for incoming job requests and completed results.	/shared/prover-compression	string	required

Invalidity

Config key	Description	Default	Type	Allowed / required
invalidity.can_run_full_large	CanRunFullLarge indicates whether the prover can run with large trace limits.	—	bool
invalidity.limitless_with_debug	LimitlessWithDebug is only looked at when the limitless invalidity prover is activated. When set to true, the limitless invalidity prover will only run in debug mode and not produce any proof. This is useful to investigate bugs in the limitless invalidity prover. The field is optional and defaults to false.	—	bool
invalidity.max_rlp_byte_size	MaxRlpByteSize specifies the maximum size of the RLP-encoded data, in bytes (this is the payload size without signature)	—	int	gte=0
invalidity.prover_mode	ProverMode stores the kind of prover to use.	—	string	required; dev|partial|full|limitless
invalidity.requests_root_dir	RequestsRootDir is the root directory for incoming job requests and completed results.	/shared/prover-invalidity	string	required

Aggregation

Config key	Description	Default	Type	Allowed / required
aggregation.is_allowed_circuit_id	IsAllowedCircuitID is a bitmask encoding which circuit IDs are allowed. Bit i (LSb to MSb) indicates whether circuit ID i is allowed. This should be computed based on the environment (testnet vs mainnet). Use circuits.GlobalCircuitIDMapping to set the appropriate bits.	—	uint64	required
aggregation.num_proofs	NumProofs is the list of supported aggregation batch sizes (circuit variants), e.g. [10, 20, 50, 100].	—	[]int	required; gt=0
aggregation.prover_mode	ProverMode stores the kind of prover to use.	—	string	required; dev|full
aggregation.requests_root_dir	RequestsRootDir is the root directory for incoming job requests and completed results.	/shared/prover-aggregation	string	required
aggregation.verifier_id	VerifierID is assigned to the proof and used by L1 contracts to select the Solidity Plonk verifier contract.	—	int	gte=0

Public input interconnection

Config key	Description	Default	Type	Allowed / required
public_input_interconnection.execution_max_nb_msg	ExecutionMaxNbMsg is the max number of L2-to-L1 messages per execution proof.	—	int	gte=0
public_input_interconnection.l2_msg_max_nb_merkle	L2MsgMaxNbMerkle is the max number of L2-to-L1 Merkle trees per aggregation; auto-computed when zero.	—	int	gte=0
public_input_interconnection.l2_msg_merkle_depth	L2MsgMerkleDepth is the depth of each L2-to-L1 message Merkle tree.	—	int	gte=0
public_input_interconnection.max_nb_circuits	MaxNbCircuits is an optional combined cap across all proof types; zero disables the check.	—	int	gte=0
public_input_interconnection.max_nb_data_availability	MaxNbDataAvailability is the max number of DA proofs per aggregation.	—	int	gte=0
public_input_interconnection.max_nb_execution	MaxNbExecution is the max number of execution proofs per aggregation.	—	int	gte=0
public_input_interconnection.max_nb_invalidity	MaxNbInvalidity is the max number of invalidity proofs per aggregation.	—	int	gte=0

Debug

Config key	Description	Default	Type	Allowed / required
debug.performance_monitor.active	Active enables granular per-step or per-round CPU profiling of the prover.	false	bool
debug.performance_monitor.profile	Profile selects the profiling scope: "prover-steps", "prover-rounds", or "all".	prover-rounds	string	prover-steps|prover-rounds|all
debug.performance_monitor.profile_dir	ProfileDir is the directory where performance profile files are written.	/shared/prover-execution/profiling	string
debug.performance_monitor.sample_duration	SampleDuration is the sampling interval used when collecting performance profiles.	1s	time.Duration
debug.profiling	Profiling indicates whether we want to generate profiles using the [runtime/pprof] pkg. Profiles the entire execution proof request-response lifecycle	false	bool
debug.tracing	Tracing indicates whether we want to generate traces using the [runtime/trace] pkg. Traces can later be read using the `go tool trace` command.	false	bool

Layer2

Config key	Description	Default	Type	Allowed / required
layer2.base_fee	No validator tag: presence is enforced in newConfigFromFile via viper.IsSet. `required` would reject 0, which is valid on gasless chains where every block has baseFeePerGas=0.	—	uint
layer2.chain_id	ChainID stores the ID of the Linea L2 network to consider.	—	uint	required
layer2.coin_base	CoinBaseStr stores the coinbase address of Linea as a string.	—	string	required; eth_addr
layer2.message_service_contract	MsgSvcContractStr stores the unique ID of the Service Contract (SC), that is, it's address, as a string. The Service Contract (SC) is a smart contract that the L2 network uses to send messages (i.e., transactions) to the L1 (mainnet). Use this field when you need the ETH address as a string.	—	string	required; eth_addr

Traces limits

The traces_limits section configures per-arithmetization-module row budgets used during execution proving.

It is a list of module entries under traces_limits.modules, each with:

• module (string, required) — module name prefix; the empty string is the default fallback
• limit (int, required, power of 2) — limit used in normal mode
• limit_large (int, required, power of 2) — limit used when large mode is active
• is_not_scalable (bool, optional) — when true, limitless scaling must not multiply this entry

Lookup uses longest-prefix matching after sorting entries in reverse alphabetical order. Large mode (for example via --large) selects limit_large instead of limit. When a scaling factor is applied and is_not_scalable is false, the selected limit is multiplied.

Concrete module tables are environment-specific; see the prover sample config-*.toml files for examples. Those samples are not portable defaults and are not reproduced here.