System breakdown

Inside DeerFlow 2.0: ByteDance SuperAgent Architecture & LangGraph Execution Engine

A production teardown of DeerFlow 2.0—analyzing supervisor routing, sandboxed Docker execution, declarative SKILL.md parsing, deep research pipelines, and Postgres durable checkpointing.

22 min read Verified 2026-08-08 4 primary sources

Canonical System Breakdown

DeerFlow 2.0 SuperAgent Architecture & LangGraph Execution Engine

RELEASE v2.0.4

2 CHAPTERS · 8 NODES · 7 EDGES

Ingress Security

HMAC SHA256

Constant-time verify (<50ms)

Async Queueing

Sidekiq + Redis

Multi-queue priority isolation

Real-time Fanout

ActionCable Pub/Sub

Redis backplane state broadcast

AI Action Engine

ActionService LLM

Streaming copilot replies (SSE)

Core Stack:Rails 7.1•Sidekiq 7•Redis 7.2•PostgreSQL 16•Vue.js 3

6 Verified Architectural Claims

Press1-6to switch views

Archify Interactive Mapv2.16.0

Inside DeerFlow 2.0: ByteDance SuperAgent Architecture & LangGraph Execution Engine — Archify Interactive System Map

JSON IR Open Standalone

Hotkeys:R Route ProbeL Role LensP Play StoryS Visual StyleT ThemeF Stage

Source Pinned & Grounded

Audited Architecture

v2.0.4

Verified Date:2026-08-07

Active Claims:6 verified

a4f891b234

Memory Layouts (2)

DeerFlow Pregel SuperAgent State Envelope

The unified state channel dictionary passed across Pregel supervisor nodes, subagent graphs, and checkpointers.

messages: List[BaseMessage]

next_agent: str

scratchpad: Dict[str, Any]

+2 more

Docker Sandbox Isolation Command Packet

Structured execution command payload sent to Docker daemon socket for sandboxed execution.

container_id: str (hex)

command: str

timeout_seconds: int32

+1 more

Engineering Concepts (4)

Agent Governance & Human-in-the-Loop Agent Control Loop LangGraph Pregel State Machine Loops Human-in-the-Loop & State Checkpoints

Primary Sources (4)

Building Effective AI Agents

Anthropic

Trustworthy Agents in Practice

Anthropic

LangGraph Agentic StateGraph Execution Engine Repository

LangChain Inc

DeerFlow 2.0 SuperAgent Harness and Deep Research Runtime Framework

ByteDance AI Lab

Reference Guides (2)

Deep Research & Autonomous Agent Swarms Reference Architecture

Production architecture blueprint for multi-query research decomposition, iterative web crawling, evidence synthesis, agent swarm topologies, scratchpad state isolation, and Postgres checkpointing.

SuperAgent Harness & Sandboxed Execution Architecture Teardown

Production reference architecture teardown of sandboxed SuperAgent execution harnesses—analyzing Docker container isolation, declarative SKILL.md dynamic parsing, bash guardrails, and durable state persistence.

Telemetry Alerts (3)

deerflow_supervisor_step_latency_seconds

histogram

Alert: p95 > 2.5s for 5m

Fix: Scale out Pregel worker replicas or check LLM API provider rate limits.

deerflow_sandbox_container_oom_total

counter

Alert: > 5 OOM kills in 15m

Fix: Increase sandbox container memory cap in docker_harness config.

deerflow_pregel_checkpoint_write_latency

histogram

Alert: p95 > 150ms

Fix: Check Postgres pool connection exhaustion or disk IOPS bounds.