Skip to content
SineraDocs

SDK reference

Main classes, functions, data types and hashing helpers of the ATP TypeScript SDK candidate.

This page summarises the public surface of the ATP TypeScript SDK 0.1.0 candidate, targeting ATP 2.4.0. The package exports exact TypeScript signatures. This page names the main entries and what they do. For installation see the SDK overview.

Implementation candidate

Client construction

createAtpClient({ publicClient, deployment }) returns a frozen client bound to a cloned copy of your deployment. Construction and module import make no RPC request, signing, approval or submission. Standalone functions with the same names accept publicClient and deployment explicitly.

MethodPurpose
read.verify(options?)Verify the selected runtime and config bindings. Returns the config and an observation
read.atom(locator, options?)Read Core state and the receipt projection with provenance
read.view(module, method, args, options?)A verified, typed view getter
read.decisionData(atomId, package, route?)Case-bound payload, digest and decision preflight. Omit the route for an award
read.bindExecutor(address)Verify clone, factory and session bindings. Returns a clone-bound context
prepare.call(module, method, args, account)Encode an allowlisted public action
prepare.creation(input, payer, session?)Validate the creation context. Returns the call, a separate approval, typed data, fee and predicted Atom id
prepare.approval(token, spender, amount, owner)Prepare an explicit approval to the pinned vault
prepare.session(terms, root, now)Validate terms and prepare root session creation
prepare.receiptClaim(locator, payer)Prepare a post-final receipt claim
prepare.feeClaim(locator, kind, payer)Prepare payment of an unpaid fee obligation to its fixed recipient
prepare.metadata(locator, publisher, revision, input)Validate finalized root publication
simulate(call)Read-only canonical preflight with the actual caller
write(simulation, wallet)Submit after a fresh preflight and wallet checks
waitAndReadback(submitted, options?)Reconcile receipt, payload, canonical block, events and Core state

Data types

TypeMeaning
Deploymentbigint chain id and deployment block, bytes32 config id and module bindings. Core and registry are required
RuntimeBindingAddress and runtime hash of the complete deployed bytes
AtomLocatorChain id, exact Core address and bytes32 Atom id
ReadOptionsOptional finalized, safe or latest block tag, explicit block number, and maxAgeSeconds
ObservationChain id, block number and hash, timestamp, finality and observation time
CreationInputIntent, time policy, asset policy, resolution policy and the signatures or root proofs the route needs
PreparedCallModule, function, args, chain, config, target, account, calldata and zero native value
SimulationA prepared call with its decoded preflight result and observation
SubmittedSubmission status, transaction hash and the prepared call
BoundExecutorClone-bound deployment with executor, root, operator, token, session and active flag

Amounts, salts and wide integer fields use bigint. Use stringifyWire for JSON transport and structFromWire to rebuild validated structs.

read.atom returns state, outcome, config id, completion mode, receipt and observation. Its deadlines object reports unavailable, because Core has no full deadline getter. A provider failure throws. It never becomes "not found".

Public action allowlist

prepare.call encodes only public actions:

ModuleActions
AtomCorecreateAndEscrow, acceptOriginal, proposeRevision, acceptRevision, cancelPending, confirmCompletion, openDispute, executeTimeout
AtomCore session variantscreateAndEscrowWithSession, acceptOriginalWithSession, proposeRevisionWithSession, acceptRevisionWithSession, confirmCompletionWithSession
AtomCore dispute finalizersfinalizeDisputeByConsensus, finalizeDisputeByAward
DelegationManagercreateSession, revokeSession, revokeAll
SessionExecutorFactorycreateExecutor

The allowlist excludes governance and admin changes, direct module calls and vault shortcuts. Deadline, authorization and state checks stay with the contracts. Listing an action does not let an arbitrary caller act in a role it does not hold.

Hashing and typed data

HelperResult
hashStruct(name, model)Canonical struct hash with exact field order and widths
typedData(name, model, context)EIP-712 payload for the chain and contract you pass
intentDigest(intent, chainId, atomCore)Intent digest in the Core domain
atomId(chainId, core, rootBuyer, salt)Atom identity, independent of any global sequence
atomKey(chainId, core, atomId)Cross-deployment Atom key
configId(chainId, core, modules)Config commitment for the module set
assetId(chainId, registry, seller, objectRef)Seller-namespaced inventory id
hashTerms(terms)Hash of all delegation terms
sessionId(chainId, manager, terms, epoch, nonce)Predicted session id
executorSalt(root, operator, token, userSalt)Deterministic executor salt
predictExecutor(factory, implementation, salt)Predicted executor clone address
encodeAwardProof(pair) / decodeAwardProof(bytes)Exactly two distinct signer and signature entries

Design boundaries

LayerResponsibility
Hashing and signingOffline functions that authorize and submit nothing
Read and verificationNeeds no signer. A failed read is never reported as absence
PreparationTyped allowlist and explicit previews
Simulation and writeChecks the actual caller and context. A simulation cannot promise a future mined result
ReconciliationInclusion and protocol finality stay separate concepts

The Atom id binds chain, Core, root buyer and salt. Never use an application order id in its place. Intent typed data uses the Core as verifying contract. Root eligibility and dispute typed data use the arbitration case manager. A new Atom must also use a fee snapshot the deployment admits.

The contracts remain the authority. The SDK checks inputs early, but it does not replace on-chain rules. For the portal view of the same operations see Create an Atom.