CHANGELOG.md

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[5.8.0]

Added

• ProofRequestOptions.orgId is now also sent on session initialization (POST /api/sdk/init/session/), so the backend can attribute the session — and its SESSION_INIT analytics record — to the partner organization at creation time instead of waiting for the first status update. Omitted entirely when unset.

[5.7.0]

Added

• New ProofRequestOptions.orgId option for partner organization attribution (e.g. the SheerID orgId). When set, the SDK includes it in the body of its own session status updates (SESSION_STARTED, SESSION_CANCELLED) and embeds it top-level in the request template so the verification client forwards it on every status update it posts. Round-trips through toJsonString/fromJsonString. Optional and omitted entirely when unset; it has no effect on the verification process.

[5.6.2]

Fix

• Inclusion of parameters from JSON.parse($.claimData.parameters).paramValues along with JSON.parse($.claimData.context).extractedParameters.

[5.6.1]

Added

• RequestSpec.templateParamsMode ('separate' | 'merge', defaults to 'separate') controls how generateSpecsFromRequestSpecTemplate expands a template with multiple templateParams values. 'separate' is the existing/default behavior — one independent spec (and expected hash) per value, matching N separate proofs. 'merge' is new: it folds every value into a single spec whose responseMatches/responseRedactions gain one entry per value, matching one proof that bundles many values into a single claim (e.g. a customInjection provider that proves 30+ playlist names in one claim instead of 30 separate ones). See the README's "Provider Authors: Validating Dynamically Injected Claims" section for examples.

[5.6.0]

• Enable deferred deep links for android by default.
• Add opt-in for deferred deep links for iOS. Defaults to app clip.

[5.5.0]

Added

• New ReclaimProofRequest.cancelSession() method to deliberately cancel an in-progress verification (e.g. the user navigates away before a proof is produced). It tears down all local session machinery — the status polling interval, the 10-minute interval-ending timeout, and any portal UI (modal / tab / embedded iframe) — and marks the session SESSION_CANCELLED on the backend. onSuccess / onError are intentionally not invoked; cancellation is a silent, deliberate teardown rather than a verification outcome. Safe to call multiple times; a no-op when no session is active.
• New SessionStatus.SESSION_CANCELLED enum value. Marking a session cancelled lets the portal treat it as a clean terminal state, so an abandoned session is not later reported as a connection failure to the (typically shared) cancel/error callback URL — preventing a stale session from contaminating the next one in multi-session flows.

[5.4.1]

Fixed

• startSession now validates submitted proofs against the version the backend actually used for the session (read from session.providerVersionString on the status response) instead of the version resolved at init. Previously, if a provider's AI patch (e.g. 1.0.0-ai.6) was created mid-session and the proof was generated against that patch, the default-callback validation path threw UnknownProofsNotValidatedError because it kept asking for hashes against the init-time base version. Callers that pass an explicit verificationConfig to startSession are unaffected.

[5.4.0]

Added

• New ReclaimProofRequest.initWithSignature(applicationId, providerId, sessionAuth, options) static method that initializes a proof request from a pre-computed signature, so appSecret never has to be shipped to the client.
• New generateInitSignature(appSecret, providerId, timestamp) helper, intended for server-side use, that produces the signature initWithSignature expects (over canonicalized {providerId, timestamp}).
• Export generateAttestationNonce so backends can compute the TEE attestation nonce after sessionId is known and return it via the getAttestationNonce callback on initWithSignature. When acceptTeeAttestation is enabled, initWithSignature requires this callback and throws otherwise.

[5.3.0]

Added

• Request TEE attestations by default when initializing proof requests.
• Include the SDK-controlled TEE attestation version in request templates and attestation nonce context so attestors can keep older SDK clients on compatible proof formats.
• Add example support for server-side TEE attestation verification and displaying returned TEE proof details.

Fixed

• Support v3 TEE digest binding verification while keeping v2 verification compatible.

[5.2.0]

Changed

• Breaking: verifyTeeAttestation now takes (proof, appSecret) instead of (proof). It performs full GCP attestation verification and returns { isVerified, error? } instead of a boolean. The application ID is derived from appSecret automatically.
• Breaking: verifyProof config option renamed from verifyTEE: boolean to teeAttestation: { appSecret }. The app secret is now required so the SDK can verify application binding and recompute the attestation nonce.
• Breaking: Renamed isTeeVerified to isTeeAttestationVerified on VerifyProofResult. Now returns true when TEE verification passes and undefined when not requested (previously false).

Added

• GCP TEE attestation verification: OIDC token signature validation, platform claims, nonce/digest binding, application and session binding.
• New runTeeVerification(proofs, config) export for batch TEE verification that throws TeeVerificationError on failure.
• TeeVerificationError exported for catching TEE-specific failures.

[5.1.0]

Changed

• Refactor VerifyProofResult type into a discriminated union (VerifyProofResultSuccess and VerifyProofResultFailure).

Added

• Include publicData in successful VerifyProofResult (type VerifyProofResultSuccess) responses.
• Add verificationConfig parameter to startSession method.

[5.0.1]

Added

• Update type of hash field to include oprf-raw as a hashing method in ResponseRedactionSpec interface.