@aztec/pxe 0.0.1-commit.d3ec352c → 0.0.1-commit.d58ff9d0

package/src/tagging/constants.ts

@@ -1,2 +1,30 @@
-// Half the size of the window we slide over the tagging indexes.
-export const WINDOW_HALF_SIZE = 10;
+import { MAX_PRIVATE_LOGS_PER_TX } from '@aztec/constants';
+
+// This window bounds how far a sender's next tag index for a directional app tagging secret can run ahead of that
+// secret's highest finalized index before `PXE::proveTx` throws.
+//
+// MAX_PRIVATE_LOGS_PER_TX is a floor for ordinary transaction patterns: the private kernel accumulates every
+// surviving private log into a single tx-wide array of exactly this size, so a fresh secret's very first
+// ordinary tx can consume that many indexes before anything is finalized.
+//
+// The +20 on top is headroom for multiple pending ordinary txs to the same counterparty stacking up
+// before the first one is observed as mined. Only a run of near-MAX_PRIVATE_LOGS_PER_TX txs back-to-back would
+// exhaust this; ordinary transfers use far fewer logs per tx, so the same margin covers many more of those. Keep the
+// margin additive rather than another multiple of the floor. A larger window also makes recipient-side sync more
+// expensive, since discovery has to probe ahead of the last finalized index for every active secret.
+//
+// TODO(F-783): MAX_PRIVATE_LOGS_PER_TX is not a hard ceiling in general — a tx that creates and nullifies (squashes)
+// many notes/logs to the same secret can drive that secret's raw tag index arbitrarily higher, since indexes are
+// reserved at log emission time, before squashing is decided, and the kernel's reset/squash loop is not bounded by
+// MAX_PRIVATE_LOGS_PER_TX. No fixed window value closes that gap.
+export const UNFINALIZED_TAGGING_INDEXES_WINDOW_LEN = MAX_PRIVATE_LOGS_PER_TX + 20;
+
+// The number of tags probed per constrained secret in the first round.
+//
+// The probe doubles each round (2, 4, 8, ..., capped at UNFINALIZED_TAGGING_INDEXES_WINDOW_LEN) while every probed
+// index is a hit, stopping at the first missing tag. Constrained delivery is gapless, so a single missing tag proves
+// the stream has ended: at steady state this turns a full WINDOW_LEN probe into two tags. A secret K logs behind
+// catches up in ~log2(K) round-trips while the probe is still doubling (2, 4, 8, 16, 32), but once it saturates the cap
+// and advances WINDOW_LEN tags per round, deeper catch-up is linear at ~K/WINDOW_LEN rounds. Either way it beats both
+// the full window every round and one round per log.
+export const INITIAL_CONSTRAINED_PROBE_LEN = 2;

package/src/tagging/get_all_logs_by_tags.ts

@@ -0,0 +1,120 @@
+import type { BlockNumber } from '@aztec/foundation/branded-types';
+import type { AztecAddress } from '@aztec/stdlib/aztec-address';
+import type { BlockHash } from '@aztec/stdlib/block';
+import { MAX_RPC_LEN } from '@aztec/stdlib/interfaces/api-limit';
+import type { AztecNode } from '@aztec/stdlib/interfaces/client';
+import {
+  type LogIncludeOptions,
+  type LogResult,
+  type SiloedTag,
+  type Tag,
+  queryAllPrivateLogsByTags,
+  queryAllPublicLogsByTags,
+} from '@aztec/stdlib/logs';
+
+/** Optional block-range, effects opt-in, and pagination cap shared by both wrappers. */
+export type GetAllLogsByTagsOptions = {
+  /** Lower block bound, inclusive. */
+  fromBlock?: BlockNumber;
+  /** Upper block bound, exclusive. */
+  toBlock?: BlockNumber;
+  /**
+   * When set, each log also carries `noteHashes` and all `nullifiers` of its tx. Defaults to off — sender
+   * sync only needs `txHash`. The recipient-sync/log-service paths flip this on to build `PendingTaggedLog`
+   * / `LogRetrievalResponse` from note data.
+   */
+  includeEffects?: boolean;
+  /**
+   * Maximum number of logs returned per tag per round-trip. Capped at `MAX_LOGS_PER_TAG` on the node
+   * (rejected if higher). Defaults to `MAX_LOGS_PER_TAG`. Mainly useful for tests that need to force
+   * pagination at a small page size.
+   */
+  limitPerTag?: number;
+};
+
+/**
+ * Splits tags into chunks of MAX_RPC_LEN, paginates each chunk via the stdlib per-tag cursor helper,
+ * then stitches the results back into a single array preserving the original tag order.
+ */
+async function getAllPagesInBatches<T extends Tag | SiloedTag, Opts extends LogIncludeOptions>(
+  tags: T[],
+  fetchAllPagesForBatch: (batch: T[]) => Promise<LogResult<Opts>[][]>,
+): Promise<LogResult<Opts>[][]> {
+  if (tags.length === 0) {
+    return [];
+  }
+
+  if (tags.length <= MAX_RPC_LEN) {
+    return fetchAllPagesForBatch(tags);
+  }
+
+  const batches: T[][] = [];
+  for (let i = 0; i < tags.length; i += MAX_RPC_LEN) {
+    batches.push(tags.slice(i, i + MAX_RPC_LEN));
+  }
+  const batchResults = await Promise.all(batches.map(fetchAllPagesForBatch));
+  return batchResults.flat();
+}
+
+/**
+ * Fetches all private logs for the given tags, automatically paginating per-tag via `afterLog` cursors.
+ *
+ * @param aztecNode - The Aztec node to query.
+ * @param tags - The siloed tags to search for.
+ * @param anchorBlockHash - Reference block for the Aztec node query, throws if block is not found there (typically
+ *   because of reorgs).
+ * @param options - Optional `fromBlock`/`toBlock` range and `includeEffects` opt-in.
+ * @returns An array of log arrays, one per tag, containing all logs across all pages.
+ */
+export function getAllPrivateLogsByTags<Opts extends GetAllLogsByTagsOptions = GetAllLogsByTagsOptions>(
+  aztecNode: AztecNode,
+  tags: SiloedTag[],
+  anchorBlockHash: BlockHash,
+  options: Opts = {} as Opts,
+): Promise<LogResult<Opts>[][]> {
+  return getAllPagesInBatches<SiloedTag, Opts>(
+    tags,
+    batch =>
+      queryAllPrivateLogsByTags(aztecNode, {
+        tags: batch,
+        referenceBlock: anchorBlockHash,
+        fromBlock: options.fromBlock,
+        toBlock: options.toBlock,
+        includeEffects: options.includeEffects ?? false,
+        limitPerTag: options.limitPerTag,
+      }) as Promise<LogResult<Opts>[][]>,
+  );
+}
+
+/**
+ * Fetches all public logs for the given tags from a contract, automatically paginating per-tag via `afterLog` cursors.
+ *
+ * @param aztecNode - The Aztec node to query.
+ * @param contractAddress - The contract address to search logs for.
+ * @param tags - The tags to search for.
+ * @param anchorBlockHash - Reference block for the Aztec node query, throws if block is not found there (typically
+ *   because of reorgs).
+ * @param options - Optional `fromBlock`/`toBlock` range and `includeEffects` opt-in.
+ * @returns An array of log arrays, one per tag, containing all logs across all pages.
+ */
+export function getAllPublicLogsByTagsFromContract<Opts extends GetAllLogsByTagsOptions = GetAllLogsByTagsOptions>(
+  aztecNode: AztecNode,
+  contractAddress: AztecAddress,
+  tags: Tag[],
+  anchorBlockHash: BlockHash,
+  options: Opts = {} as Opts,
+): Promise<LogResult<Opts>[][]> {
+  return getAllPagesInBatches<Tag, Opts>(
+    tags,
+    batch =>
+      queryAllPublicLogsByTags(aztecNode, {
+        contractAddress,
+        tags: batch,
+        referenceBlock: anchorBlockHash,
+        fromBlock: options.fromBlock,
+        toBlock: options.toBlock,
+        includeEffects: options.includeEffects ?? false,
+        limitPerTag: options.limitPerTag,
+      }) as Promise<LogResult<Opts>[][]>,
+  );
+}

package/src/tagging/index.ts

@@ -1,6 +1,20 @@
-export * from './tag.js';
-export * from './constants.js';
-export * from './siloed_tag.js';
-export * from './utils.js';
-export { DirectionalAppTaggingSecret } from '@aztec/stdlib/logs';
-export { type PreTag } from '@aztec/stdlib/logs';
+/**
+ * This directory contains the sender and recipient synchronization algorithms that use tagging.
+ *
+ * The objective of the sender sync algorithm is to determine which tags have already been used by a sender, thereby
+ * deciding which tag should be used next.
+ *
+ * The objective of the recipient sync algorithm is to fetch and sync the corresponding logs.
+ *
+ * @module tagging
+ */
+
+export { syncTaggedPrivateLogs } from './recipient_sync/sync_tagged_private_logs.js';
+export { syncSenderTaggingIndexes } from './sender_sync/sync_sender_tagging_indexes.js';
+export { persistSenderTaggingIndexRangesForTx } from './persist_sender_tagging_index_ranges.js';
+export { UNFINALIZED_TAGGING_INDEXES_WINDOW_LEN, INITIAL_CONSTRAINED_PROBE_LEN } from './constants.js';
+export { getAllPrivateLogsByTags, getAllPublicLogsByTagsFromContract } from './get_all_logs_by_tags.js';
+
+// Re-export tagging-related types from stdlib
+export { AppTaggingSecret, Tag, SiloedTag } from '@aztec/stdlib/logs';
+export { type PreTag, type TaggingIndexRange } from '@aztec/stdlib/logs';

package/src/tagging/persist_sender_tagging_index_ranges.ts

@@ -0,0 +1,57 @@
+import type { Logger } from '@aztec/foundation/log';
+import type { PrivateKernelTailCircuitPublicInputs } from '@aztec/stdlib/kernel';
+import type { TaggingIndexRange } from '@aztec/stdlib/logs';
+import type { TxHash } from '@aztec/stdlib/tx';
+
+import type { SenderTaggingStore } from '../storage/tagging_store/sender_tagging_store.js';
+import { reconcileTaggingIndexRangesAgainstSurvivingTags } from './reconcile_tagging_index_ranges.js';
+
+/**
+ * Persists the tagging index ranges that a tx used as a sender, after reconciling them against the kernel's surviving
+ * private logs.
+ *
+ * The tagging index cache reserves an index for every log emission attempted during private execution, but the kernel
+ * may then squash some of those logs (e.g. when a note is created and nullified within the same tx). This function
+ * shrinks the recorded ranges to match the actual on-chain footprint before writing them, so that what we persist is
+ * consistent with what the network will see.
+ *
+ * @remarks Storing the recorded ranges in the DB may not be seen as necessary because we sync from chain before sending
+ * new logs, but the sync can only see logs already included in blocks. If we sent another transaction before this one
+ * was included from the same PXE, and that transaction contained a log with a tag derived from the same secret, we
+ * would reuse the tag and the transactions would be linked. Persisting in the DB prevents that linkage.
+ *
+ * @param store - The sender tagging store.
+ * @param recordedRanges - The tagging index ranges as recorded during private execution (pre-squash).
+ * @param publicInputs - The final kernel public inputs, used to determine which private logs survived squashing.
+ * @param getTxHash - Lazy accessor for the tx hash. Called only when there is something to persist, since computing
+ * the tx hash is expensive.
+ * @param jobId - Job context for staged writes to the store. See `JobCoordinator` for more details.
+ * @param log - Logger.
+ */
+export async function persistSenderTaggingIndexRangesForTx(
+  store: SenderTaggingStore,
+  recordedRanges: TaggingIndexRange[],
+  publicInputs: PrivateKernelTailCircuitPublicInputs,
+  getTxHash: () => Promise<TxHash>,
+  jobId: string,
+  log: Logger,
+): Promise<void> {
+  if (recordedRanges.length === 0) {
+    log.debug(`No tagging index ranges used in the tx`);
+    return;
+  }
+
+  const survivingTags = new Set(
+    publicInputs.getNonEmptyPrivateLogs().map(privateLog => privateLog.fields[0].toString()),
+  );
+  const reconciledRanges = await reconcileTaggingIndexRangesAgainstSurvivingTags(recordedRanges, survivingTags);
+
+  if (reconciledRanges.length === 0) {
+    log.debug(`All tagging index ranges used in the tx were squashed by the kernel`, { recordedRanges });
+    return;
+  }
+
+  const txHash = await getTxHash();
+  await store.storePendingIndexes(reconciledRanges, txHash, jobId);
+  log.debug(`Stored used tagging index ranges as sender for the tx`, { recordedRanges, reconciledRanges });
+}

package/src/tagging/recipient_sync/sync_tagged_private_logs.ts

@@ -0,0 +1,358 @@
+import { BlockNumber } from '@aztec/foundation/branded-types';
+import { isDefined } from '@aztec/foundation/types';
+import type { BlockHash } from '@aztec/stdlib/block';
+import type { AztecNode } from '@aztec/stdlib/interfaces/client';
+import type { AppTaggingSecret, LogResult } from '@aztec/stdlib/logs';
+import { AppTaggingSecretKind, SiloedTag } from '@aztec/stdlib/logs';
+import type { BlockHeader } from '@aztec/stdlib/tx';
+
+import type { RecipientTaggingStore } from '../../storage/tagging_store/recipient_tagging_store.js';
+import { INITIAL_CONSTRAINED_PROBE_LEN, UNFINALIZED_TAGGING_INDEXES_WINDOW_LEN } from '../constants.js';
+import { getAllPrivateLogsByTags } from '../get_all_logs_by_tags.js';
+import { findHighestIndexes } from './utils/find_highest_indexes.js';
+
+/**
+ * Fetches and syncs tagged private logs for multiple sender-recipient pairs, batching tag queries across all secrets
+ * into shared RPC calls. Handles both constrained and unconstrained secrets, which differ in their scan strategy
+ * but share the same batching infrastructure.
+ *
+ * # Unconstrained secrets
+ *
+ * For each unconstrained secret we sync logs that correspond to the tagging index range
+ * (highestAgedIndex, highestFinalizedIndex + WINDOW_LEN]
+ *
+ * highestAgedIndex is the highest index that was used in a tx that is included in a block at least
+ *                  `MAX_TX_LIFETIME` seconds ago.
+ * highestFinalizedIndex is the highest index that was used in a tx that is included in a finalized block.
+ *
+ * "(" denotes an open end of the range - the index is not included in the range.
+ * "]" denotes a closed end of the range - the index is included in the range.
+ *
+ * ## Explanation of highestAgedIndex
+ *
+ * highestAgedIndex is chosen such that for all tagging indexes `i <= highestAgedIndex` we know that no new logs can
+ * ever appear.
+ *
+ * This relies on the "maximum inclusion timestamp" rule enforced by the kernel and rollup circuits:
+ * - a transaction's maximum inclusion timestamp is at most `MAX_TX_LIFETIME` seconds after
+ *   the timestamp of its anchor block; and
+ * - a rollup only includes transactions whose inclusion timestamp is >= the L2 block's timestamp.
+ *
+ * Suppose some device used index `I` in a transaction anchored to block `B_N` at time `N`, and that block is now at
+ * least `MAX_TX_LIFETIME` seconds in the past. Then there is no possibility of any *other* device
+ * trying to use an index <= `I` while anchoring to a *newer* block than `B_N` because if we were anchoring to
+ * a newer block than `B_N` then we would already have seen the log with index `I` and hence the device would have
+ * chosen a larger index.
+ *    If that *other* device would anchor to a block older than `B_N` then that tx could never be included in a block
+ * because it would already have been expired.
+ *
+ * Therefore, once we see that index `I` has been used in a block that is at least `MAX_TX_LIFETIME`
+ * seconds old, we can safely stop syncing logs for all indexes <= `I` and set highestAgedIndex = `I`.
+ *
+ * ## Explanation of the upper bound `highestFinalizedIndex + WINDOW_LEN`
+ *
+ * When a sender chooses a tagging index, they will select an index that is at most `WINDOW_LEN` greater than
+ * the highest finalized index. If that index was already used, they will throw an error. For this reason we
+ * don't have to look further than `highestFinalizedIndex + WINDOW_LEN`.
+ *
+ * # Constrained secrets
+ *
+ * Constrained delivery enforces a contiguous index sequence (each send proves the predecessor's nullifier,
+ * chaining back to the handshake), so there is only one sender and indexes are sequential. This means:
+ * - The lower bound is `highestFinalizedIndex + 1`. The `highestAgedIndex` mechanism is unnecessary because
+ *   the nullifier chain prevents out-of-order index usage: no device can fill in a lower index after a higher
+ *   one is already on-chain.
+ * - Because the sequence is gapless, the scan stops at the first missing index: that gap proves the sequence has
+ *   ended and no further logs exist. We exploit this with a doubling first-miss probe (see
+ *   `INITIAL_CONSTRAINED_PROBE_LEN` in ../constants.ts), so a
+ *   steady-state sync with no new logs costs two tags instead of the whole window. The probe length is in-memory
+ *   state scoped to one sync call, so every sync restarts at the initial probe.
+ * - The upper bound is the same as unconstrained: `highestFinalizedIndex + WINDOW_LEN`. Advancing the probe is
+ *   decoupled from persisting the finalized index, so unfinalized logs at the top of the run are still fetched
+ *   while only the finalized prefix is persisted.
+ *
+ * # Batching across secrets
+ *
+ * Instead of running one RPC call per secret, we merge tags from all pending secrets (both constrained and
+ * unconstrained) into a single flat array, make one batched `getAllPrivateLogsByTags` call (which internally
+ * chunks at MAX_RPC_LEN), then split results back per secret using tracked offsets. Only secrets whose window
+ * advanced are kept for the next iteration.
+ */
+export async function syncTaggedPrivateLogs(
+  secrets: AppTaggingSecret[],
+  aztecNode: AztecNode,
+  taggingStore: RecipientTaggingStore,
+  anchorBlockHeader: BlockHeader,
+  finalizedBlockNumber: BlockNumber,
+  jobId: string,
+): Promise<LogResult[]> {
+  if (secrets.length === 0) {
+    return [];
+  }
+
+  const anchorBlockNumber = anchorBlockHeader.getBlockNumber();
+  const anchorBlockHash = await anchorBlockHeader.hash();
+  const currentTimestamp = anchorBlockHeader.globalVariables.timestamp;
+
+  // Read stored indexes from the db and compute the initial [start, end) range for each secret
+  let pending = await getIndexRangesForSecrets(secrets, taggingStore, jobId);
+  const allLogs: LogResult[] = [];
+
+  while (pending.length > 0) {
+    // Compute tags for all pending secrets and fetch logs in batched RPC calls
+    const logsPerSecret = await fetchLogsForSecrets(pending, aztecNode, anchorBlockNumber, anchorBlockHash);
+
+    const nextRound = await Promise.all(
+      pending.map(async (pendingSecret, i) => {
+        const logsFoundWithSecret = logsPerSecret[i];
+        if (logsFoundWithSecret.length === 0) {
+          // No logs found, no need to update indexes or advance window.
+          return undefined;
+        }
+
+        allLogs.push(...logsFoundWithSecret.map(({ log }) => log));
+
+        // Persist new indexes. If the finalized index moved forward, the window advances
+        // and we need another round for this secret.
+        return pendingSecret.kind === AppTaggingSecretKind.CONSTRAINED
+          ? await processConstrainedResults(
+              pendingSecret,
+              logsFoundWithSecret,
+              taggingStore,
+              currentTimestamp,
+              finalizedBlockNumber,
+              jobId,
+            )
+          : await processUnconstrainedResults(
+              pendingSecret,
+              logsFoundWithSecret,
+              taggingStore,
+              currentTimestamp,
+              finalizedBlockNumber,
+              jobId,
+            );
+      }),
+    );
+
+    pending = nextRound.filter(isDefined);
+  }
+
+  return allLogs;
+}
+
+/** Reads stored indexes for each secret and computes the initial index range to query. */
+function getIndexRangesForSecrets(
+  secrets: AppTaggingSecret[],
+  taggingStore: RecipientTaggingStore,
+  jobId: string,
+): Promise<PendingSecret[]> {
+  return Promise.all(
+    secrets.map(async (secret): Promise<PendingSecret> => {
+      const currentHighestFinalizedIndex = await taggingStore.getHighestFinalizedIndex(secret, jobId);
+      const boundEnd = (currentHighestFinalizedIndex ?? 0) + UNFINALIZED_TAGGING_INDEXES_WINDOW_LEN + 1;
+
+      if (secret.kind === AppTaggingSecretKind.CONSTRAINED) {
+        // Constrained streams are gapless and resume at the finalized index, so probe a small initial window and stop
+        // at the first missing tag instead of always fetching the full window. The probe grows each round up to the
+        // window cap.
+        const start = currentHighestFinalizedIndex === undefined ? 0 : currentHighestFinalizedIndex + 1;
+        return {
+          kind: AppTaggingSecretKind.CONSTRAINED,
+          secret,
+          start,
+          end: Math.min(boundEnd, start + INITIAL_CONSTRAINED_PROBE_LEN),
+          boundEnd,
+          probeLen: INITIAL_CONSTRAINED_PROBE_LEN,
+        };
+      }
+
+      // Unconstrained secrets can have gaps, so they scan the whole window starting past the highest aged index.
+      const highestAgedIndex = await taggingStore.getHighestAgedIndex(secret, jobId);
+      const start = highestAgedIndex === undefined ? 0 : highestAgedIndex + 1;
+      return { kind: AppTaggingSecretKind.UNCONSTRAINED, secret, start, end: boundEnd };
+    }),
+  );
+}
+
+/**
+ * Computes siloed tags for all pending secrets' index ranges, fetches logs in one batched RPC call,
+ * and returns the results grouped back per secret.
+ */
+async function fetchLogsForSecrets(
+  pending: PendingSecret[],
+  aztecNode: AztecNode,
+  anchorBlockNumber: BlockNumber,
+  anchorBlockHash: BlockHash,
+): Promise<LogWithIndex[][]> {
+  // Determine the index range for each secret
+  const indexesPerSecret = pending.map(({ start, end }) => Array.from({ length: end - start }, (_, i) => start + i));
+
+  // Compute siloed tags for all indexes
+  const tagsPerSecret = await Promise.all(
+    pending.map(({ secret }, i) =>
+      Promise.all(indexesPerSecret[i].map(index => SiloedTag.compute({ extendedSecret: secret, index }))),
+    ),
+  );
+
+  const allTags = tagsPerSecret.flat();
+
+  // getAllPrivateLogsByTags handles MAX_RPC_LEN chunking internally. Recipient sync builds `PendingTaggedLog` from
+  // each log's note hashes and first nullifier, so we opt into effects. The `toBlock` cap (anchor block + 1,
+  // exclusive) tells the node to skip any logs in blocks past the anchor — the same guard previously enforced
+  // by an in-memory filter on the response.
+  const allResults = await getAllPrivateLogsByTags(aztecNode, allTags, anchorBlockHash, {
+    includeEffects: true,
+    toBlock: BlockNumber(anchorBlockNumber + 1),
+  });
+
+  // Split flat results back per secret using the known lengths
+  const logsPerSecret: LogWithIndex[][] = [];
+  let offset = 0;
+  for (const indexes of indexesPerSecret) {
+    const logsForSecret: LogWithIndex[] = [];
+    for (let i = 0; i < indexes.length; i++) {
+      for (const log of allResults[offset + i]) {
+        logsForSecret.push({ log, taggingIndex: indexes[i] });
+      }
+    }
+    logsPerSecret.push(logsForSecret);
+    offset += indexes.length;
+  }
+
+  return logsPerSecret;
+}
+
+/**
+ * Processes fetched logs for a constrained secret: persists the finalized index and advances the probe. See the
+ * "# Constrained secrets" section above for why the scan stops at the first missing index. Only `highestFinalizedIndex`
+ * is tracked (no aged index).
+ */
+async function processConstrainedResults(
+  pending: PendingConstrained,
+  logsWithIndexes: LogWithIndex[],
+  taggingStore: RecipientTaggingStore,
+  currentTimestamp: bigint,
+  finalizedBlockNumber: BlockNumber,
+  jobId: string,
+): Promise<PendingSecret | undefined> {
+  // Find where the contiguous run of indexes ends; all logs in the batch fall within this prefix.
+  const indexesWithLogs = new Set(logsWithIndexes.map(l => l.taggingIndex));
+  let firstMissingIndex = pending.start;
+  while (firstMissingIndex < pending.end && indexesWithLogs.has(firstMissingIndex)) {
+    firstMissingIndex++;
+  }
+
+  // Persist the highest finalized index found. The nullifier for a constrained-delivery log is emitted in the same
+  // transaction, guaranteeing the log cannot disappear once included, so we persist even when a gap stops the scan.
+  // This lets the next sync round skip already-finalized indexes.
+  const { highestFinalizedIndex } = findHighestIndexes(logsWithIndexes, currentTimestamp, finalizedBlockNumber);
+  if (highestFinalizedIndex !== undefined) {
+    await taggingStore.updateHighestFinalizedIndex(pending.secret, highestFinalizedIndex, jobId);
+  }
+
+  // Advancing the probe is decoupled from persisting the finalized index: keep scanning as long as the probe was
+  // entirely hits (no gap) and there is room before the protocol bound, even if none of those hits is finalized yet.
+  // Gating this on finalization would drop logs in unfinalized blocks, which sit at the top of the contiguous run.
+  // The bound re-anchors to any finalized index found this round.
+  const probeFullyConsumed = firstMissingIndex >= pending.end;
+  const boundEnd =
+    highestFinalizedIndex !== undefined
+      ? Math.max(pending.boundEnd, highestFinalizedIndex + UNFINALIZED_TAGGING_INDEXES_WINDOW_LEN + 1)
+      : pending.boundEnd;
+
+  // Double the probe each round, capped at the window (see INITIAL_CONSTRAINED_PROBE_LEN in ../constants.ts).
+  // The queried range is already bounded by boundEnd (a probe can never reach
+  // more than WINDOW_LEN past the finalized frontier); this cap just keeps the stored length from growing unbounded
+  // once the probe saturates the window.
+  const nextProbeLen = Math.min(pending.probeLen * 2, UNFINALIZED_TAGGING_INDEXES_WINDOW_LEN);
+
+  if (probeFullyConsumed && pending.end < boundEnd) {
+    return {
+      kind: AppTaggingSecretKind.CONSTRAINED,
+      secret: pending.secret,
+      start: pending.end,
+      end: Math.min(boundEnd, pending.end + nextProbeLen),
+      boundEnd,
+      probeLen: nextProbeLen,
+    };
+  }
+
+  return undefined;
+}
+
+/**
+ * Processes a single unconstrained secret's fetched logs: updates stored indexes and returns a new PendingSecret
+ * if the window needs to advance, or undefined if this secret is done.
+ */
+async function processUnconstrainedResults(
+  pending: PendingUnconstrained,
+  logsWithIndexes: LogWithIndex[],
+  taggingStore: RecipientTaggingStore,
+  currentTimestamp: bigint,
+  finalizedBlockNumber: BlockNumber,
+  jobId: string,
+): Promise<PendingSecret | undefined> {
+  const { highestAgedIndex, highestFinalizedIndex } = findHighestIndexes(
+    logsWithIndexes,
+    currentTimestamp,
+    finalizedBlockNumber,
+  );
+
+  // Store updates in data provider and update local variables
+  if (highestAgedIndex !== undefined) {
+    await taggingStore.updateHighestAgedIndex(pending.secret, highestAgedIndex, jobId);
+  }
+
+  if (highestFinalizedIndex === undefined) {
+    // We have not found a new highest finalized index, so there is no need to move the window forward.
+    return undefined;
+  }
+
+  if (highestAgedIndex !== undefined && highestAgedIndex > highestFinalizedIndex) {
+    // This is just a sanity check as this should never happen.
+    throw new Error(
+      `Highest aged index (${highestAgedIndex}) must not exceed highest finalized index (${highestFinalizedIndex})`,
+    );
+  }
+
+  await taggingStore.updateHighestFinalizedIndex(pending.secret, highestFinalizedIndex, jobId);
+
+  // For the next iteration we want to look only at indexes for which we have not yet fetched logs while
+  // ensuring that we do not look further than WINDOW_LEN ahead of the highest finalized index.
+  const end = highestFinalizedIndex + UNFINALIZED_TAGGING_INDEXES_WINDOW_LEN + 1;
+  return {
+    kind: AppTaggingSecretKind.UNCONSTRAINED,
+    secret: pending.secret,
+    start: pending.end,
+    end,
+  };
+}
+
+/** Working scan state common to both secret kinds: the half-open [start, end) tag-index range to query this round. */
+type PendingSecretBase = {
+  secret: AppTaggingSecret;
+  start: number;
+  end: number;
+};
+
+/** Constrained scan state, carrying the doubling-probe bookkeeping that only the gapless constrained scan uses. */
+type PendingConstrained = PendingSecretBase & {
+  kind: typeof AppTaggingSecretKind.CONSTRAINED;
+  // Exclusive upper bound on the indexes this secret may probe this sync (highest finalized index + WINDOW_LEN + 1, the
+  // protocol bound on how far ahead of finalized a log can sit). `end` grows toward it each round.
+  boundEnd: number;
+  // Intended probe length for the round that produced `end`. The queried span can be smaller when boundEnd caps it.
+  probeLen: number;
+};
+
+/** Unconstrained scan state: a plain window with no probe bookkeeping, since the scan always covers the full window. */
+type PendingUnconstrained = PendingSecretBase & {
+  kind: typeof AppTaggingSecretKind.UNCONSTRAINED;
+};
+
+type PendingSecret = PendingConstrained | PendingUnconstrained;
+
+type LogWithIndex = {
+  log: LogResult;
+  taggingIndex: number;
+};

package/src/tagging/recipient_sync/utils/find_highest_indexes.ts

@@ -0,0 +1,34 @@
+import { MAX_TX_LIFETIME } from '@aztec/constants';
+import type { LogResult } from '@aztec/stdlib/logs';
+
+/**
+ * Finds the highest aged and the highest finalized tagging indexes.
+ */
+export function findHighestIndexes(
+  privateLogsWithIndexes: Array<{ log: LogResult; taggingIndex: number }>,
+  currentTimestamp: bigint,
+  finalizedBlockNumber: number,
+): { highestAgedIndex: number | undefined; highestFinalizedIndex: number | undefined } {
+  let highestAgedIndex = undefined;
+  let highestFinalizedIndex = undefined;
+
+  for (const { log, taggingIndex } of privateLogsWithIndexes) {
+    const ageInSeconds = currentTimestamp - log.blockTimestamp;
+
+    if (
+      ageInSeconds >= BigInt(MAX_TX_LIFETIME) &&
+      (highestAgedIndex === undefined || taggingIndex > highestAgedIndex)
+    ) {
+      highestAgedIndex = taggingIndex;
+    }
+
+    if (
+      log.blockNumber <= finalizedBlockNumber &&
+      (highestFinalizedIndex === undefined || taggingIndex > highestFinalizedIndex)
+    ) {
+      highestFinalizedIndex = taggingIndex;
+    }
+  }
+
+  return { highestAgedIndex, highestFinalizedIndex };
+}