feat: implement committee aggregation #537

GrapeBaBa · 2026-01-29T03:58:18Z

No description provided.

Signed-off-by: grapebaba <grapebaba@grapebabadeMacBook-Pro.local>

Copilot

Pull request overview

Implements committee-style attestation aggregation by introducing an aggregated-attestation gossip message, adding subnet-capable gossip topic specs, and splitting signature aggregation into (1) aggregating fresh gossip signatures vs (2) selecting stored aggregated proofs for block production.

Changes:

Added SignedAggregatedAttestation type and new .aggregation gossip topic/message for disseminating aggregated attestation proofs.
Reworked attestation signature handling: aggregators build proofs from gossip signatures; proposers select proofs from stored payloads for inclusion in blocks.
Introduced attestation committee/subnet configuration and updated node/network plumbing (topic specs, subscriptions, timing intervals).

Reviewed changes

Copilot reviewed 21 out of 21 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
pkgs/types/src/lib.zig	Re-exports the new aggregated attestation type for use across packages.
pkgs/types/src/block_signatures_testing.zig	Updates tests to cover the new aggregation helpers (`aggregateGossipSignatures`, `selectAggregatedProofs`).
pkgs/types/src/block.zig	Adds new aggregation helpers and refactors aggregation logic into separate phases (gossip aggregation vs stored-proof selection).
pkgs/types/src/attestation.zig	Introduces `SignedAggregatedAttestation` with JSON helpers and deinit.
pkgs/state-transition/src/mock.zig	Updates mock chain generation to use the new gossip-signature aggregation API.
pkgs/params/src/types.zig	Adds `ATTESTATION_COMMITTEE_COUNT` to preset config.
pkgs/params/src/presets/mainnet.zig	Sets `ATTESTATION_COMMITTEE_COUNT` for mainnet preset.
pkgs/params/src/lib.zig	Exposes `ATTESTATION_COMMITTEE_COUNT` as an active preset constant.
pkgs/node/src/validator_client.zig	Adjusts validator interval switch to match new intervals-per-slot.
pkgs/node/src/node.zig	Adds aggregator mode behavior, aggregated-attestation publishing, topic-spec subscriptions, and attestation subnet publish logic.
pkgs/node/src/network.zig	Switches gossip publish API to `publishWithTopic`.
pkgs/node/src/lib.zig	Re-exports `buildGossipTopicSpecs`.
pkgs/node/src/forkchoice.zig	Adds aggregator/subnet tracking, selective signature storage for committee aggregation, and ordered payload storage.
pkgs/node/src/constants.zig	Changes `INTERVALS_PER_SLOT` to 5 to introduce a dedicated aggregation interval.
pkgs/node/src/chain.zig	Updates block production to select stored aggregated proofs; adds committee aggregation + aggregated-attestation gossip handling.
pkgs/network/src/mock.zig	Updates mock gossip interface to accept topic specs and topic-aware publish.
pkgs/network/src/lib.zig	Exposes `GossipTopicSpec` from the network package.
pkgs/network/src/interface.zig	Adds subnet-aware topic encoding/decoding, new `.aggregation` topic/message, and updates gossip interface signatures.
pkgs/network/src/ethlibp2p.zig	Adds `.aggregation` message decoding/logging and uses configured topic specs for network subscription/publish.
pkgs/cli/src/node.zig	Adds `is_aggregator` option, builds gossip topic specs, and plumbs into node/network init.
pkgs/cli/src/main.zig	Updates ethlibp2p demo/init wiring to supply gossip topic specs.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-01-29T04:08:23Z

pkgs/node/src/node.zig

+        const subnet_id =
+            @as(usize, @intCast(validator_id)) % ATTESTATION_COMMITTEE_COUNT;
+        const subnet_topic = networks.GossipTopicSpec{ .topic = .attestation, .subnet_id = subnet_id };
+        try self.network.publishWithTopic(subnet_topic, &gossip_msg);
+        try self.network.publishWithTopic(.{ .topic = .attestation }, &gossip_msg);


Publishing the same attestation to both the subnet topic and the global .attestation topic will cause aggregator nodes (which subscribe to both) to process duplicate messages. Since the subnet_id is not propagated to handlers, these duplicates are indistinguishable and can inflate work/metrics. Consider publishing to only one topic (e.g. only subnet when ATTESTATION_COMMITTEE_COUNT > 1, or only global until subnet routing is fully supported).

Suggested change

const subnet_id =

@as(usize, @intCast(validator_id)) % ATTESTATION_COMMITTEE_COUNT;

const subnet_topic = networks.GossipTopicSpec{ .topic = .attestation, .subnet_id = subnet_id };

try self.network.publishWithTopic(subnet_topic, &gossip_msg);

try self.network.publishWithTopic(.{ .topic = .attestation }, &gossip_msg);

if (ATTESTATION_COMMITTEE_COUNT == 1) {

// With a single committee, there is no meaningful subnetting; use the global topic only.

try self.network.publishWithTopic(.{ .topic = .attestation }, &gossip_msg);

} else {

// With multiple committees, publish only to the subnet-specific topic to avoid duplicates.

const subnet_id =

@as(usize, @intCast(validator_id)) % ATTESTATION_COMMITTEE_COUNT;

const subnet_topic = networks.GossipTopicSpec{ .topic = .attestation, .subnet_id = subnet_id };

try self.network.publishWithTopic(subnet_topic, &gossip_msg);

}

Copilot · 2026-01-29T04:08:23Z

pkgs/node/src/node.zig

+    try topics.append(.{ .topic = .block });
+    try topics.append(.{ .topic = .attestation });
+    try topics.append(.{ .topic = .aggregation });
+
+    if (is_aggregator) {
+        if (validator_ids) |ids| {
+            if (ATTESTATION_COMMITTEE_COUNT == 0) return topics.toOwnedSlice();
+            const LocalSubnetSet = std.StaticBitSet(ATTESTATION_COMMITTEE_COUNT);
+            var local_subnets = LocalSubnetSet.initEmpty();
+
+            for (ids) |validator_id| {
+                const subnet_id = validator_id % ATTESTATION_COMMITTEE_COUNT;
+                local_subnets.set(subnet_id);
+            }
+
+            var subnet_it = local_subnets.iterator(.{});
+            while (subnet_it.next()) |subnet_id| {
+                try topics.append(.{ .topic = .attestation, .subnet_id = subnet_id });
+            }


buildGossipTopicSpecs always includes the global .attestation topic and (for aggregators) also adds per-subnet .attestation topics. Because inbound messages are decoded to GossipTopic.attestation without exposing subnet_id to the application layer, subscribing to both leads to duplicate processing if peers publish on both. Either avoid mixing global+subnet subscriptions, or plumb subnet_id through to the gossip callback so duplicates can be filtered intentionally.

Copilot · 2026-01-29T04:08:24Z

pkgs/node/src/chain.zig

+        while (true) {
+            const candidate_attestations =
+                try buildAggregatedAttestationsFromAttestations(self.allocator, selected_attestations.items);
+            var candidate_block = types.BeamBlock{
+                .slot = opts.slot,
+                .proposer_index = opts.proposer_index,
+                .parent_root = parent_root,
+                .state_root = undefined,
+                .body = types.BeamBlockBody{
+                    .attestations = candidate_attestations,
+                },
+            };
+            defer candidate_block.deinit();
+
+            var temp_state_ptr_opt: ?*types.BeamState = try self.allocator.create(types.BeamState);
+            errdefer if (temp_state_ptr_opt) |ptr| self.allocator.destroy(ptr);
+
+            const temp_state_ptr = temp_state_ptr_opt.?;
+            try types.sszClone(self.allocator, types.BeamState, pre_state.*, temp_state_ptr);
+            temp_state_ptr_opt = null;
+            defer {
+                temp_state_ptr.deinit();
+                self.allocator.destroy(temp_state_ptr);
+            }
+            try stf.apply_raw_block(self.allocator, temp_state_ptr, &candidate_block, self.block_building_logger);
+
+            var added_any = false;
+            var no_payloads = false;
+            self.forkChoice.signatures_mutex.lock();
+            defer self.forkChoice.signatures_mutex.unlock();
+            if (self.forkChoice.aggregated_payloads.count() == 0) {


selectAttestationsForBlock repeatedly clones pre_state and runs stf.apply_raw_block inside a while (true) loop. This makes block production potentially O(N) full state transitions per slot, and it also holds signatures_mutex while doing allocations (sszRoot, append, put). Consider narrowing the lock to only the aggregated_payloads.get calls and/or reworking the selection to avoid re-running the full state transition on each iteration (e.g., incremental updates or a bounded fixed-point loop).

Copilot · 2026-01-29T04:08:24Z

pkgs/types/src/block.zig

+    fn ensureAttestationAppendCapacity(self: *Self) !void {
+        const max = params.VALIDATOR_REGISTRY_LIMIT;
+        if (self.attestations.len() >= max or self.attestation_signatures.len() >= max) {
+            return error.Overflow;
+        }

-        // Group attestations by data root using bitsets for validator tracking
-        const AttestationGroup = struct {
-            data: attestation.AttestationData,
-            data_root: Root,
-            validator_bits: std.DynamicBitSet,
-        };
+        try self.attestations.inner.ensureTotalCapacity(self.attestations.len() + 1);
+        try self.attestation_signatures.inner.ensureTotalCapacity(self.attestation_signatures.len() + 1);


params.VALIDATOR_REGISTRY_LIMIT is a u32, but self.attestations.len() / self.attestation_signatures.len() are usize. This comparison will not type-check in Zig; cast the limit to usize (or define const max: usize = @intCast(params.VALIDATOR_REGISTRY_LIMIT)) before comparing and reserving capacity.

grapebaba added 2 commits January 28, 2026 22:10

feat: implement commitee aggregation spec

864d6ac

Signed-off-by: grapebaba <grapebaba@grapebabadeMacBook-Pro.local>

refactor: use 5 interval consensus

c285f44

Signed-off-by: grapebaba <grapebaba@grapebabadeMacBook-Pro.local>

Copilot AI review requested due to automatic review settings January 29, 2026 03:58

Copilot started reviewing on behalf of GrapeBaBa January 29, 2026 03:58 View session

Copilot AI reviewed Jan 29, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: implement committee aggregation #537

feat: implement committee aggregation #537

Uh oh!

GrapeBaBa commented Jan 29, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Jan 29, 2026

Uh oh!

Copilot AI Jan 29, 2026

Uh oh!

Copilot AI Jan 29, 2026

Uh oh!

Copilot AI Jan 29, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

feat: implement committee aggregation #537

Are you sure you want to change the base?

feat: implement committee aggregation #537

Uh oh!

Conversation

GrapeBaBa commented Jan 29, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Jan 29, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 29, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 29, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 29, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants