Back to Ruflo

Swarm SOTA Report — 2026-05-29

v3/docs/research/dream-cycle-2026-05-29-swarm.md

3.34.06.9 KB
Original Source

Swarm SOTA Report — 2026-05-29

TL;DR: Five 2026 arXiv papers redefine swarm consensus — decentralized Shapley credit, hierarchical PBFT at O(n) message complexity, and communication-policy decoupling each expose structural gaps in Ruflo's flat Raft consensus that will limit scaling past ~50 agents.


What's New in 2026

FindingSourceConfidence
SWARM+: hierarchical PBFT consensus scales to 1,000 agents with 97-98% scheduling latency improvement and ≤7.5% degradation under 50% agent failuresarXiv:2603.19431 (Mar 19, 2026)A
SwarmHarness: skill nodes self-organise into compute swarm without central authority via DHT + Shapley-value credit attributionarXiv (May 27, 2026)A
SLIM: decoupling communication from policy maintains performance under severe bandwidth constraints — direct drop-in for drone and LLM-agent swarmsarXiv (May 20, 2026)A
Auction-Consensus: RL-based bidding replaces hand-crafted functions in consensus task allocation, improving solution quality while preserving decentralised executionarXiv (May 20, 2026)A
CINOC: neural policies trained on 10-agent swarms transfer zero-shot to 500-agent populations via physics-informed operator learningarXiv (May 25, 2026)A
LangGraph v0.4 fastest on latency across 2,000-task benchmark; CrewAI 3× token overhead on single-tool flows vs LangGraphtensoria.fr benchmark (2026)B

Ruflo Current Capability

CapabilityStatusDetail
Consensus protocolRaft (leader-based)Good crash-fault tolerance; weak under 30%+ Byzantine
Max validated agents8 (CLAUDE.md default)No public stress-test beyond this
Credit attributionNoneAgents have no Shapley-style contribution weighting
Communication-policy couplingCoupledAll agents share same coordination channel; no bandwidth tiering
Scale-transfer primitivesNoneSwarm configs are static; no zero-shot scale extrapolation
Event-triggered messagingNoAll coordination is time- or call-driven
Swarm observability (ruview)Hooks + memory onlyNo structured audit trail or agent execution graph
Hybrid vector search (ruvector)NoHNSW only; no vector+keyword hybrid retrieval
Public recall benchmarkNoneNo published score for AgentDB/HNSW at 1M+ vectors

Competitor Comparison

FrameworkConsensusMax Agents (validated)Credit AttributionObservability
Ruflo (claude-flow v3.6)Raft8 (configured max)NoneHooks + memory (no trace graph)
LangGraph v0.4Checkpoint-based (no consensus)No stated limitNoneLangSmith native (full trace)
AutoGen 1.0 GAGroupChat selectorNo stated limitNoneAzure Monitor integration
CrewAI EnterpriseSequential/hierarchical fixed50+ (enterprise reported)NoneSOC2 observability hooks
OpenAI Agents SDKHandoff-based (no consensus)No stated limitNoneBuilt-in tracing, spans
SWARM+ (research)Hierarchical PBFT1,000 (benchmarked)Dynamic quorumNot productised

Benchmarks

SystemMetricValueGradeSource
SWARM+Scheduling latency improvement over baseline97-98%AarXiv:2603.19431
SWARM+Job completion under 50% agent failure≥92.5%AarXiv:2603.19431
SWARM+Agent scale1,000 agentsAarXiv:2603.19431
RLR consensusQuality of Delivery vs RaBFT (30% Byzantine)+126%BResearch Square rs-7522626
QdrantQuery latency at 100M vectors, 95% recallsub-100msBdasroot.net (Apr 2026)
MilvusThroughput at 1B vectors100K+ QPSBdasroot.net (Apr 2026)
WeaviateHybrid search recall at 1M vectors93.2%Bdasroot.net (Apr 2026)
LangGraphFastest latency in 2,000-task cross-framework test#1Btensoria.fr (2026)

SOTA Proof & Witness

This section is completed after file hash computation.

FieldValue
Session commitb8a49cfb82ffe3e1f376cc3be3776bdd4d8c9ab7
Report SHA-256712376e4ef16f9f00e24da2f5dabe4600825f7b103f98670298b50e77c09121e
Witness stamp8e3c2d8fb902014b739e54cf07b1b237d0168feb1a24c4e6e1d6ae47474f3588
Verifiersha256sum dream-gist-2026-05-29.md → concat session commit → sha256sum → must equal witness stamp

Scan: ruvector-integration

Source: dasroot.net (Apr 2026), digitalapplied.com (2026), callsphere.ai (2026)

Qdrant is 10-25% faster than Weaviate/Milvus on common workloads (Grade B). Weaviate hybrid search (vector + keyword) achieves 93.2% recall at 1M vectors (Grade B). Milvus handles 1B-scale collections at 100K+ QPS (Grade B). Ruflo's ruvector/AgentDB has no public benchmark, no hybrid retrieval, and no validated performance at 1M+ vectors. LanceDB costs ~$200-500/month at 10M vectors — 4-25× cheaper than alternatives.

Finding: ruvector benchmarking gap is the single highest-ROI documentation task: publishing HNSW recall@10 at 100K, 1M, 10M vectors would immediately differentiate vs all four competitors who publish vendor-only Grade B claims.


Scan: ruview-integration (Swarm Observability)

Source: codebase grep (no ruview module found), LangSmith docs, OpenAI Agents SDK tracing

Ruflo has no dedicated swarm execution trace module. post-edit and post-task hooks write to memory namespaces but produce no structured span-based trace graph. LangGraph ships LangSmith natively (full agent graph, token cost per node, replay). OpenAI Agents SDK emits OpenTelemetry-compatible spans out of the box. CrewAI Enterprise has SOC2 observability. Ruflo has none.

Finding (Grade C — single internal assessment): Absence of a structured trace graph is the primary barrier to enterprise adoption; every competitor has addressed this; Ruflo has not.


  1. ADR-132 (file today): Adopt hierarchical consensus topology for swarms >20 agents — replace flat Raft with SWARM+-style two-tier hierarchy (leader group + worker shards), reducing coordination overhead and enabling scale to 200+ agents without O(n²) message explosion. Implement in @claude-flow/cli hive-mind module.

  2. Benchmark ruvector now (implementation-level, no ADR): Add npx claude-flow performance benchmark --suite vector that runs recall@10 and p99 latency at 100K/1M/10M vectors using the existing HNSW index and publishes results to v3/docs/benchmarks/vector-search.md. This closes the Grade B gap vs all four external vector DBs.

  3. ruview stub (implementation-level, no ADR): Add @claude-flow/ruview plugin stub that wraps existing hook events into OpenTelemetry-compatible spans. Single-file implementation using @opentelemetry/sdk-trace-node as a dev dependency. Enables export to Jaeger/Zipkin with zero architectural change to existing hooks.