Skip to content

tsingxuanhan/agent4science

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

164 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

NexusFlow — Agent4Science

Dynamic Cognitive Topology Engine for Ultra-Long-Horizon Multi-Agent Collaborative Reasoning

License: MIT Python 3.9+ Lines Phase

What is NexusFlow?

NexusFlow is a dynamic heterogeneous multi-agent framework that solves ultra-long-horizon complex tasks through cognitive division of labor and adaptive context management. Built on 7 iterative phases of development (27,000+ lines), it goes beyond traditional "decompose-and-execute" patterns by enabling genuine deep collaborative reasoning.

Key Innovation: Information asymmetry between agents is not a deficiency — it's a resource. By constraining what each agent can see, we force cognitive processes that no single agent (even with complete information) could achieve alone.

Core Architecture

┌─────────────────────────────────────────────────────────────────────────┐
│                    NexusFlow Dynamic Cognitive Topology                  │
├─────────────────────────────────────────────────────────────────────────┤
│                                                                          │
│  ┌──────────────────────────────────────────────────────────────────┐  │
│  │              PlannerAgent (Task Decomposition)                    │  │
│  │  Goal → TaskTree → Auto-detect CDoL nodes → Route              │  │
│  └──────────────────────┬───────────────────────────────────────────┘  │
│                         │                                               │
│  ┌──────────────────────┴───────────────────────────────────────────┐  │
│  │            DynamicTopologyRouter (Phase 6)                        │  │
│  │  ┌─────────────┐  ┌──────────────┐  ┌──────────────────────┐    │  │
│  │  │ Capability   │  │ Load-Aware   │  │ Experience Learning  │    │  │
│  │  │ Matching     │  │ Scheduling   │  │ & Pattern Cache      │    │  │
│  │  └─────────────┘  └──────────────┘  └──────────────────────┘    │  │
│  │                         │                                         │  │
│  │         ┌───────────────┼───────────────┐                        │  │
│  │         │ Sequential    │ CDoL Mode     │                        │  │
│  │         │ (star/chain)  │ (perspective  │                        │  │
│  │         │               │  decomposition│                        │  │
│  │         └───────────────┴───────────────┘                        │  │
│  └──────────────────────────────────────────────────────────────────┘  │
│                                                                          │
│  ┌──────────────────────────────────────────────────────────────────┐  │
│  │        Cognitive Division Engine (Phase 7 — Core Innovation)      │  │
│  │                                                                    │  │
│  │   PerspectiveDecomposer → CommunicationLayer → FusionJudge        │  │
│  │                                                                    │  │
│  │   Round 0: Independent reasoning (isolated context masks)         │  │
│  │   Round 1: Difference attribution (share conclusions, not data)   │  │
│  │   Round 2: Correction & convergence                               │  │
│  │   → False consensus detection (compare reasoning chains, not answers)│
│  └──────────────────────────────────────────────────────────────────┘  │
│                                                                          │
│  ┌──────────────────────────────────────────────────────────────────┐  │
│  │        Adaptive Context Manager (Phase 7)                         │  │
│  │                                                                    │  │
│  │   LocalContextWindow ← AdaptiveWindowController ← LazinessDetector│  │
│  │        │                                                           │  │
│  │   GlobalMemoryPool ← ForcedGlobalSync ← RetrievalHeadAgent        │  │
│  │                                                                    │  │
│  │   "Small window + NoPE" strategy: constrain local view,           │  │
│  │   periodically force global awareness injection                    │  │
│  └──────────────────────────────────────────────────────────────────┘  │
│                                                                          │
│  ┌──────────────────────────────────────────────────────────────────┐  │
│  │           Edge-Cloud Scheduler (Phase 6)                          │  │
│  │   Edge (Device) ←→ Fog (Gateway) ←→ Cloud (Server)              │  │
│  │   Privacy-first scheduling · 5 scheduling policies                │  │
│  └──────────────────────────────────────────────────────────────────┘  │
│                                                                          │
│  ┌──────────────────────────────────────────────────────────────────┐  │
│  │     Foundation: BaseAgent · A2A Protocol · Letta Memory System    │  │
│  │     8 specialized agents · 12 built-in tools · Guardrails        │  │
│  └──────────────────────────────────────────────────────────────────┘  │
│                                                                          │
└─────────────────────────────────────────────────────────────────────────┘

7 Phases of Evolution

Phase Focus Key Modules
1 Generalization Infrastructure TaskTree decomposition, Tree-of-Thought, Self-reflection
2 Planning Engine Planner/Executor separation, Adaptive strategy
3 Tool Ecosystem CodeAct execution, 12 tools, MCP v2
4 Knowledge & Memory Letta 3-Layer Memory, Sleeptime consolidation, Multi-Hop RAG
5 AGI Core Autonomous processor, Meta-cognition, Cross-domain transfer, AgentOS
6 Dynamic Swarm Intelligence DynamicTopologyRouter, EdgeCloudScheduler, NexusFlow Dashboard
7 Deep Collaborative Reasoning CognitiveDivisionEngine, AdaptiveContextManager, False Consensus Detection

Phase 7: Core Innovations

1. Cognitive Division of Labor (CDoL) Engine

Traditional multi-agent systems share all information with all agents. CDoL does the opposite:

The Theoretical Flip:

  • Classical view: information asymmetry is a problem to be solved
  • CDoL view: information asymmetry is a cognitive resource to be exploited

How it works:

  1. PerspectiveDecomposer splits a task into isolated viewpoints (6 strategies: evidence-split, role-constraint, layer-separation, modality-split, time-slice, abstraction-level)
  2. Each agent receives a ContextMask — only sees a subset of the problem
  3. Round 0: Independent reasoning in isolation
  4. Round 1: Share intermediate conclusions (not raw data) via lossy communication channel
  5. Round 2: Attribution & correction
  6. FusionJudge detects 4 types of contradictions, including false consensus (same conclusion, contradictory reasoning paths)

Why it matters: The collaborative gain comes from the cognitive process forced by information constraints — the upper bound exceeds any single agent, even one given complete information.

2. Adaptive Context Manager

Inspired by Tsinghua & OpenBMB's hybrid attention research (2026), addressing the "large window laziness" problem:

  • LazinessDetector monitors 4 indicators: retrieval frequency, correction rate, confidence trend, information source diversity
  • AdaptiveWindowController dynamically adjusts context window size (512 → 32768)
  • GlobalMemoryPool serves as horizontal convergence layer across all agents
  • ForcedGlobalSync periodically injects global summaries (NoPE strategy: "small window + no positional encoding = forced global awareness")
  • RetrievalHeadAgent specialized retrieval-only agent for bridging information gaps

3. Seamless Architecture Integration

Both modules surgically embed into the existing 6-layer architecture through precisely defined injection points — no module stacking, clean backward compatibility.

Agent Roles

Agent Role Specialization
Miner Literature Explorer Deep research, paper discovery, knowledge mining
Assayer Knowledge Validator Cross-verification, fact-checking, quality audit
Caster Code Engineer Script generation, data processing, automation
Artisan Domain Expert Field-specific Q&A, concept explanation
Planner Strategy Architect Task decomposition, CDoL node detection
Executor Action Runner CodeAct-first execution, tool orchestration
Researcher Deep Analyst Multi-source investigation, evidence synthesis
Reviewer Quality Gate Output validation, consistency check

Module Overview

Module Lines Phase Description
BaseAgent ~2600 1-7 Core agent + context window hooks
VectorMemory ~1700 4 NGram+TF-IDF hybrid retrieval
A2A Protocol ~1030 1,7 Inter-agent + 6 CDOL message types
CognitiveDivisionEngine ~1500 7 Perspective decomposition + fusion
AdaptiveContextManager ~1520 7 Window control + laziness detection
DynamicTopologyRouter ~870 6 NetworkX-based dynamic routing
EdgeCloudScheduler ~535 6 Edge-Fog-Cloud scheduling
Autonomous ~780 5 6-stage goal processor
Meta-Cognition ~680 5 Self-assessment system
AgentOS ~640 5 FastAPI + stdio server
RecallMemory ~610 4 Episodic memory with temporal indexing
ArchivalMemory ~600 4 Compressed long-term storage
TaskTree ~620 1,7 Task decomposition + CDoL markers
Cross-Domain ~590 5 Analogical transfer (8 seeds)
MemoryManager ~510 4,7 3-layer + Global Pool
NexusFlow Demo ~590 6,7 Full demonstration
Dashboard ~370 6 Real-time monitoring UI

Total: 27,000+ lines across 34+ modules

Quick Start

Basic Usage

from agent4science import BaseAgent

agent = BaseAgent(
    name="research_assistant",
    model="pro",
    system_prompt="You are a scientific research assistant"
)

reply = agent.chat("Explain the mechanical properties of LC3 cement")

Cognitive Division Demo

from agent4science import CognitiveDivisionEngine
from demo.nexusflow_demo import demo_cognitive_division

# Run the full CDoL demonstration
# Simulates a low-carbon cement formulation decision
# with 3 agents in isolated perspectives
result = demo_cognitive_division()

Adaptive Context Demo

from demo.nexusflow_demo import demo_adaptive_context

# Shows window size adaptation + laziness detection
# across 10 simulation steps
demo_adaptive_context()

Autonomous Mode

from agent4science import AutonomousProcessor

processor = AutonomousProcessor(agent)
result = processor.process(
    "Design an experiment to test SSC with different nano-SiO₂ dosages"
)
# Intent → Decompose → Plan → Execute → Verify → Report

Configuration

All credentials via environment variables — no hardcoded secrets:

export LLM_API_KEY="sk-your-key-here"
export LLM_ENDPOINT="http://127.0.0.1:8083/v1/chat/completions"

Phase 7 Configuration (config.example.py)

# Cognitive Division Engine
CDOL_ENABLED = True
CDOL_MAX_ROUNDS = 2              # Max communication rounds
CDOL_MIN_BRIDGEABILITY = 0.3     # Min bridgeability threshold
CDOL_FALSE_CONSENSUS_THRESHOLD = 0.7

# Adaptive Context Manager
CONTEXT_WINDOW_DEFAULT = 4096
CONTEXT_WINDOW_MIN = 512
CONTEXT_WINDOW_MAX = 32768
LAZINESS_CHECK_INTERVAL = 5
GLOBAL_SYNC_INTERVAL = 10

Theoretical Foundation

Why Cognitive Division Works

The theoretical core is drawn from cognitive science and recent transformer architecture research:

  1. Bounded rationality as feature: Herbert Simon's bounded rationality suggests that decision-making under constraints produces different (not worse) cognitive processes. CDoL exploits this by design.

  2. Hybrid attention insight: Tsinghua & OpenBMB (2026) showed that "large window laziness" — models with larger context windows attend less carefully to relevant information. Our AdaptiveContextManager applies the counter-strategy: constrain local windows, force global sync.

  3. False consensus detection: Traditional ensemble methods compare final answers. CDoL compares reasoning chains — detecting when agents reach the same conclusion through contradictory exclusion processes. This catches errors that voting/averaging would miss.

Project Structure

agent4science/
├── agents/                          # Role-based agents
│   ├── domains/                     # Knowledge domain modules
│   ├── planner.py                   # Task decomposition + CDoL detection
│   └── __init__.py                  # Module registry
├── tools/                           # 12 built-in tools
├── demo/
│   └── nexusflow_demo.py           # Full Phase 6+7 demonstration
├── docs/                            # Design documents
├── iteration/                       # Version history (v3.0 → v5.0)
│
├── # Phase 7 Core (NEW)
├── cognitive_division_engine.py     # CDoL: decomposition + communication + fusion
├── adaptive_context_manager.py      # Adaptive window + global memory + laziness
│
├── # Phase 6 Core
├── dynamic_router.py                # Dynamic topology routing
├── edge_cloud_scheduler.py          # Edge-Fog-Cloud scheduling
├── dashboard.py                     # Real-time monitoring
│
├── # Foundation (Phase 1-5)
├── base_agent.py                    # Core agent class
├── autonomous.py                    # Autonomous goal processor
├── meta_cognition.py                # Self-assessment
├── cross_domain.py                  # Cross-domain transfer
├── continuous_learning.py           # Dual-path learning
├── agentos.py                       # AgentOS server
├── task_tree.py                     # Task decomposition
├── memory_manager.py                # Memory orchestration (4-layer)
├── a2a_protocol.py                  # Agent-to-Agent protocol
├── a2a_gateway.py                   # A2A network hub
├── ...                              # 20+ more modules
│
├── config.example.py                # Configuration template
├── requirements.txt                 # Dependencies
└── README.md                        # This file

Design Principles

  1. Safety First — Guardrails, circuit breakers, and anti-pattern detection at every layer
  2. Cognitive Division > Simple Decomposition — Information asymmetry as resource, not defect
  3. Adaptive over Static — Context windows, routing topology, and scheduling all adapt dynamically
  4. Observable — OpenTelemetry tracing + NexusFlow Dashboard for real-time monitoring
  5. Backward Compatible — All Phase 7 modules are optional; missing dependencies degrade gracefully
  6. Privacy-First Scheduling — Edge-cloud scheduler respects data sensitivity levels

License

MIT License — see LICENSE for details.


NexusFlow — Where cognitive diversity meets dynamic topology

About

Universal AGI Agent Framework — 5-layer cognitive architecture with meta-cognition, 3-tier memory, and knowledge orchestration on consumer hardware

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors