feat(sdk): integration with proving service by noa-starkware · Pull Request #480 · starkware-libs/starknet-privacy

noa-starkware · 2026-02-12T17:12:55Z

This change is

m-kus · 2026-02-16T14:17:31Z

  constructor(
    private readonly params: {
-      account: Account; // the user account (for signing)
+      account: AccountSignerRaw; // the user account (for signing)


Can you use Account::sign_message instead? Using Account interface allows seamless wallet plug in here. Any additional interface wallets need to implement increase friction and maintenance burden

Do you think this could work? The doc of sign_message: "Signs a JSON object for off-chain usage with the private key and returns the signature. This adds a message prefix so it can't be interchanged with transactions"

m-kus · 2026-02-16T14:21:01Z

+
+  async prove(invocation: ProofInvocationWithPayload): Promise<Proof> {
+    const inv = invocation;
+    const transactionPayload = {


Just curious why do you do this low level rather than using starknet.js helpers?

Since this isnt a trivial call with an array of Calls, do you know of any helper I could use for this non conventional use case?

This would probably do the trick:

import { type RpcProvider, ETransactionType } from "starknet"; const channel = (this.provider as RpcProvider).channel; const transactionPayload = channel.buildTransaction( { type: ETransactionType.INVOKE, contractAddress: invocation.contractAddress, calldata: invocation.calldata, signature: invocation.signature ?? [], nonce: invocation.nonce, resourceBounds: invocation.resourceBounds, tip: invocation.tip, paymasterData: invocation.paymasterData ?? [], accountDeploymentData: invocation.accountDeploymentData ?? [], nonceDataAvailabilityMode: invocation.nonceDataAvailabilityMode ?? "L1", feeDataAvailabilityMode: invocation.feeDataAvailabilityMode ?? "L1", }, "transaction" );

m-kus · 2026-02-16T14:24:27Z

 * without actually generating zero-knowledge proofs.
 */
-export class CallMockProofProvider implements ProofProviderInterface {
+export class CallMockProofProvider extends AbstractProofProvider {


i'd suggest adding mocked tests to sdk as well to run in ci

Working on mock prover for ci, is that what you meant?

m-kus · 2026-02-16T14:25:14Z

   */
-  protected async getViewingKey(): Promise<ViewingKey> {
-    return await this.viewingKeyProvider.getViewingKey();
+  protected getViewingKey(): Promise<ViewingKey> | ViewingKey {


why making it sync?

m-kus · 2026-02-16T14:26:10Z

@@ -0,0 +1,55 @@
+/**
+ * Abstract base for proof providers with shared getDefaultDetails implementation.


Is this abstraction needed? Could we instead update ProofProviderInterface ?

Both the mock and the real proof providers use the same get_details impl

A simpler alternative would be a plain helper function (avoids extra abstraction layer):

// Replace abstract class with:
function getDefaultProofDetails(chainId: StarknetChainId): ProofInvocationFactoryDetails {
return { nonce: 0n, resourceBounds: { ... }, chainId, ... };
}

m-kus · 2026-02-16T14:28:10Z

 | `VITE_ADMIN_ADDRESS` | Admin/minter account address |
 | `VITE_ADMIN_KEY` | Admin account private key |
 | `VITE_ACCOUNTS` | JSON array of user accounts (see `.env.example`) |
+| `VITE_PROVING_SERVICE_URL` | Optional. Proving service URL (e.g. `http://136.115.124.93:3000`). If unset, the app uses the mock prover (`execute_view` only) |


I'd suggest adding support to the demo in a separate PR, this one should successfully run mocked tests in CI and e2e scenario against integration env

m-kus · 2026-02-16T14:29:46Z

+  type ProofInvocationFactoryDetails,
+  type ProofProviderInterface,
+} from "starknet-sdk";
+import { IndexerClient } from "../src/indexer-client.js";


You can use ContractDiscoveryProvider instead of indexer btw, to simplify your e2e flow (just test sdk<>proving service)

I thought we wanted to test everything together, didnt we?

m-kus · 2026-02-16T14:32:52Z

@@ -5,16 +5,12 @@
 * MockServerAction[] callbacks as the proof output.


Note the signature validation error in CallMockProofProvider (see devnet tests)

m-kus · 2026-02-17T12:53:26Z

+}
+
+/** Check result: proof non-empty, proof_facts and l2_to_l1_messages are arrays. */
+function isProveTransactionResult(value: unknown): value is ProveTransactionResult {


Do we need to validate proving service response?

No need IMO

m-kus · 2026-02-17T12:53:50Z

+  );
+}
+
+export type BlockId =


This already exists in starknet.js

m-kus · 2026-02-17T12:55:54Z

+// ============ Calldata ============
+
+/** Ensure calldata elements are 0x-prefixed hex (starknet signer/RPC expect this). */
+export function ensureHexCalldata(calldata: string[]): string[] {


It's used only once, perhaps worth just inlining?

m-kus · 2026-02-17T12:57:16Z

 export interface ProofProviderInterface {
-  /** Get the default factory details for creating proof invocations */
-  getDefaultDetails(): ProofInvocationFactoryDetails;
+  /** Get the default factory details for creating proof invocations (may be async e.g. to fetch nonce) */


I'd make it always async to avoid mixing async/sync ret types

Remove the comment, not relevant

Note that two mixed return types remain:

ViewingKeyProvider.getViewingKey(): returns Promise | ViewingKey

ProofProviderInterface.prove(): returns Proof | Promise

m-kus · 2026-02-17T12:58:10Z

+  l2_to_l1_messages: MessageToL1[];
+}
+
+export interface MessageToL1 {


should be in starknet.js

With the same name?

import type { RPC } from "starknet";
type MessageToL1 = RPC.RPCSPEC010.MSG_TO_L1;

Although it's quite ugly :) Can probably keep the custom type

m-kus · 2026-02-17T13:00:20Z

+/**
+ * Proving service URL. If a host is given (no scheme), defaults to http and port 3000.
+ */
+export function normalizeProvingServiceUrl(hostOrUrl: string): string {


Not sure if it's worth introducing a special helper for host + default port, just require users to provide a url in expected format

m-kus · 2026-02-17T13:07:49Z

+}
+
+/** Proof to bytes: u32[] packed big-endian, or base64 string decoded. */
+function proofToBytes(proof: number[] | string): Uint8Array {


Is proof returned as base64 string from the service? afaics it's serialized as list of integers https://github.com/starkware-libs/sequencer/blob/69b4c8d378580a1aec345f23dadd4766e036fe3b/crates/starknet_os_runner/src/server/http_server.rs#L35

m-kus · 2026-02-17T13:09:52Z

+}
+
+/** Map JSON-RPC error code to typed exception. */
+export function mapProvingServiceError(error: {


Separate classes for proving errors seems a bit of an overkill tbh, can just throw a detailed error message?

m-kus · 2026-02-17T13:10:37Z

+      senderAddress: poolAddressHex,
+      compiledCalldata,
+      version: (details.version ?? ETransactionVersion3.V3) as `${typeof ETransactionVersion3.V3}`,
+      nonce,


I'd suggest using existing methods from starknet.js for building/signing tx

Because its non conventional invoke I dont know if its possible

semgrep-code-starkware-libs · 2026-02-17T14:12:12Z

    "node_modules/starknet": {
      "version": "9.4.0",
-      "resolved": "git+ssh://git@github.com/m-kus/starknet.js.git#425082b072b705ac129d1898ea4fc9863fe367f9",
+      "resolved": "git+ssh://git@github.com/m-kus/starknet.js.git#19c1b5fcb227df60dc4d7e965cf1aea0699f543f",
+      "integrity": "sha512-26eJb/OY7nZA/bDGd8VTOwlp9NKBZsIjjYx1uqtr5ZRLiy2UsZQefWN5YfHiow1oyAzTO+dLtcfHUX410/TTSw==",
      "license": "MIT",
      "dependencies": {
        "@noble/curves": "~1.7.0",


Package "$KEY" is resolved from a non-standard source: "git+ssh://git@github.com/m-kus/starknet.js.git#19c1b5fcb227df60dc4d7e965cf1aea0699f543f". Expected packages to be resolved from the official npm registry (registry.npmjs.org). Unusual sources may indicate dependency confusion, supply chain attacks, or misconfigured registries.

🌟 Fixed in commit 3b8850e 🌟

ittaysw · 2026-02-22T06:34:04Z

+    // Mock proof provider returns proofFacts but proof.data may be undefined; use placeholder for sequencer
    const response = await this.setup.admin.executeFromOutside(outsideTransaction, {
-      proofFacts: callAndProof.proofFacts,
+      proofFacts: callAndProof.proof.proofFacts,


Why isn't the proof passed ?

ittaysw · 2026-02-22T06:40:10Z

+      fee_data_availability_mode: inv.feeDataAvailabilityMode ?? "L1",
+    };
+
+    const result = await this.provingService.proveTransaction("latest", transactionPayload);


if we prove 'latest' only, it means now the wallet needs to wait 10 blocks before transmitting the tx. There should probably be some configuration that allows to specify 'latest' (in case we're proving a new state change) or 'latest-verifiable' or relative number from current block number (nice to have) or explicit block hash (for max reliability)

noa-starkware self-assigned this Feb 12, 2026

noa-starkware marked this pull request as draft February 12, 2026 17:13

m-kus force-pushed the m-kus/sn-integration-demo branch 3 times, most recently from c3ef2de to 794ccf7 Compare February 15, 2026 14:45

noa-starkware force-pushed the noa/sdk/proving-service branch from 54fd897 to 9c7226b Compare February 16, 2026 09:23

semgrep-code-starkware-libs Bot reviewed Feb 16, 2026

View reviewed changes

Comment thread demo/discovery-service.toml

noa-starkware force-pushed the noa/sdk/proving-service branch from 9c7226b to 7b2c490 Compare February 16, 2026 09:34

noa-starkware marked this pull request as ready for review February 16, 2026 09:47

noa-starkware force-pushed the noa/sdk/proving-service branch from 7b2c490 to 66f0ce8 Compare February 16, 2026 09:51

noa-starkware requested review from Yoni-Starkware, ittaysw and m-kus February 16, 2026 10:16

m-kus reviewed Feb 16, 2026

View reviewed changes

m-kus force-pushed the m-kus/sn-integration-demo branch 10 times, most recently from de39ecd to f805a3a Compare February 17, 2026 12:27

m-kus reviewed Feb 17, 2026

View reviewed changes

m-kus force-pushed the m-kus/sn-integration-demo branch 2 times, most recently from 629ec28 to 4a8e35a Compare February 17, 2026 13:56

noa-starkware force-pushed the noa/sdk/proving-service branch from 1abc72b to 8742b39 Compare February 17, 2026 14:08

semgrep-code-starkware-libs Bot reviewed Feb 17, 2026

View reviewed changes

m-kus force-pushed the noa/sdk/proving-service branch from 0bafa59 to 96f2d5b Compare February 18, 2026 21:37

m-kus force-pushed the graphite-base/480 branch from 20cf470 to d5d2ec1 Compare February 18, 2026 21:37

m-kus changed the base branch from graphite-base/480 to m-kus/sn-integration-demo February 18, 2026 21:37

This was referenced Feb 18, 2026

feat(demo): deploy preview #512

Merged

feat(ci): add proving service docker build workflow and deploy docs #528

Merged

m-kus changed the base branch from m-kus/sn-integration-demo to graphite-base/480 February 19, 2026 15:06

m-kus force-pushed the graphite-base/480 branch from d5d2ec1 to 277d444 Compare February 19, 2026 15:41

m-kus force-pushed the noa/sdk/proving-service branch from 96f2d5b to 1092be3 Compare February 19, 2026 15:41

m-kus changed the base branch from graphite-base/480 to m-kus/sn-integration-demo February 19, 2026 15:42

m-kus force-pushed the m-kus/sn-integration-demo branch from 277d444 to faad3a9 Compare February 19, 2026 16:29

m-kus force-pushed the noa/sdk/proving-service branch from 1092be3 to acf6e20 Compare February 19, 2026 16:29

m-kus mentioned this pull request Feb 21, 2026

feat(sdk): rename package to @starkware-libs/starknet-privacy-sdk #533

Merged

ittaysw reviewed Feb 22, 2026

View reviewed changes

m-kus force-pushed the m-kus/sn-integration-demo branch from faad3a9 to 8a0efd1 Compare February 22, 2026 17:26

m-kus force-pushed the noa/sdk/proving-service branch 2 times, most recently from b01d143 to 39bb920 Compare February 22, 2026 18:19

m-kus force-pushed the m-kus/sn-integration-demo branch from 8a0efd1 to 4d1b5d9 Compare February 22, 2026 18:19

feat(sdk): integration with proving service

bd07bf3

m-kus force-pushed the noa/sdk/proving-service branch from 39bb920 to bd07bf3 Compare February 22, 2026 18:25

m-kus force-pushed the m-kus/sn-integration-demo branch from 4d1b5d9 to 796b988 Compare February 22, 2026 18:25

noa-starkware marked this pull request as draft February 23, 2026 08:36

noa-starkware closed this Feb 23, 2026

m-kus mentioned this pull request Feb 24, 2026

fix(e2e): add declare-class script and fix outgoing state sync #553

Closed

		@@ -0,0 +1,55 @@
		/**
		* Abstract base for proof providers with shared getDefaultDetails implementation.

		@@ -5,16 +5,12 @@
		* MockServerAction[] callbacks as the proof output.

Conversation

noa-starkware commented Feb 12, 2026 • edited by RoeiE-starkware Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

noa-starkware Feb 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

m-kus Feb 18, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

noa-starkware commented Feb 12, 2026 •

edited by RoeiE-starkware

Loading

noa-starkware Feb 18, 2026 •

edited

Loading

m-kus Feb 18, 2026 •

edited

Loading

semgrep-code-starkware-libs Bot Feb 17, 2026 •

edited

Loading