Slipknot V4.1 Lite - Lightweight Topological Arbitration MCP Coprocessor

Slipknot V4.1 Lite - Lightweight Topological Arbitration MCP Coprocessor Underlying Mathematical Consensus Protocol for Multi-Agent Clusters | 7-Day Rapid Deployment Edition

🎯 Project Positioning Slipknot is an AI infrastructure project that combines Topological Data Analysis (TDA) with Multi-Agent Systems. Its core positioning is the "Underlying Mathematical Consensus Protocol for Multi-Agent Clusters" / "Lightweight Topological Arbitration MCP Coprocessor".

Exclusive Differentiation Barriers: Topological Consensus Court: Uses persistent graph Wasserstein/Bottleneck distances as a mathematically neutral yardstick to resolve decision-making conflicts and deadlocks among multiple Agents. Reward Evolution Flywheel: Agents feed back business rewards, and Bayesian optimization automatically iterates TDA hyperparameters, making the data increasingly accurate with use. Prototype of Federated Topological Privacy Computing: Only lightweight topological skeletons are uploaded while raw data remains local, ensuring compliance. Industry-Adaptive Plugins: The same topological computation results are automatically translated into actionable business instructions for Agents in corresponding industries.

🚀 Quick Start

Option 1: Pure FastAPI Version (Recommended, Zero Dependencies) bash Install basic dependencies pip install -r requirements.txt

One-click start python dev_start.py

Access services Health Check: http://127.0.0.1:8000/health API Docs: http://127.0.0.1:8000/docs MCP Endpoint: http://127.0.0.1:8000/mcp Tools List: http://127.0.0.1:8000/mcp/tools

Option 2: FastMCP Standard Version (Standard MCP Protocol) bash Create virtual environment and install full dependencies python -m venv venv source venv/bin/activate pip install -r requirements.txt

Start FastMCP service python -m slipknot.gateway

Access services SSE Endpoint: http://127.0.0.1:8000/sse HTTP Endpoint: http://127.0.0.1:8000/mcp

Option 3: Docker Deployment bash chmod +x docker-run.sh ./docker-run.sh

🧪 Running Tests bash Core functionality tests (no need to start the service) python scripts/demo_test.py

Gateway API tests (requires service to be started first) python dev_start.py & python scripts/gateway_test.py

📦 Project Structure text slipknot-lite/ ├── slipknot/ │ ├── init.py # Version declaration │ ├── core/ # TDA core computation layer │ │ ├── accel.py # Numba JIT acceleration functions │ │ └── engine.py # Topological analysis engine │ ├── consensus/ # Topological Consensus Court │ │ ├── models.py # Pydantic data models │ │ └── court.py # Consensus arbitration engine │ ├── storage/ # Storage backend (Dual-mode) │ │ └── backend.py # Memory/Redis automatic fallback │ ├── task/ # Task pool │ │ └── pool.py # Thread pool task scheduling │ ├── flywheel/ # Evolution flywheel │ │ ├── filter.py # Reward filter │ │ └── optimizer.py # Bayesian optimizer │ ├── plugins/ # Industry plugins │ │ ├── energy.py # Energy storage scheduling │ │ ├── quant.py # Quantitative trading │ │ └── fraud.py # Risk control & fraud │ ├── gateway.py # FastMCP standard gateway │ └── gateway_simple.py # Pure FastAPI gateway ├── scripts/ │ ├── demo_test.py # Core functionality tests │ └── gateway_test.py # Gateway API tests ├── tmp/ # Test data directory ├── dev_start.py # Local one-click start script ├── Dockerfile # Production container image ├── docker-run.sh # Container one-click deployment ├── requirements.txt # Dependency list ├── .env.example # Environment variables template └── README.md # Project documentation

🔧 Core Features

TDA Topological Analysis Engine Adaptive Sampling: Default upper limit of 3000, lower limit of 80, automatically adapts to data scale. UMAP Dimensionality Reduction: Manifold learning preserves topological structures. Persistent Homology Computation: Supports H0/H1/H2 three-order hole detection. Singularity Identification: Automatically identifies anomalous data points. Numba JIT Acceleration: Just-in-time compilation for core computational loops.

Topological Consensus Court Dual-Layer Distance Determination: Bottleneck for rapid screening + Wasserstein for precise computation. Three-Tier Verdicts: CONSENSUS_FAST / CONSENSUS / TOPOLOGICAL_DIVERGENCE. Weaker Agent Identification: Automatically marks the Agent with lower confidence during divergence. Mathematically Neutral Arbitration: Pure topological distance, free from subjective bias.

Reward Self-Evolution Flywheel Reward Cleansing: Extreme value truncation + 3σ anomaly detection. Bayesian Optimization: Gaussian process surrogate model + gp_minimize. Industry Parameter Isolation: Independent hyperparameter optimization for each industry. Automatic Triggering: Triggers optimization for every 50 valid Rewards.

Industry-Adaptive Plugins Industry Plugin Core Capabilities Energy Storage Scheduling energy Load cycle identification, charge/discharge strategies, fault warning

Quantitative Trading quant Market cycle detection, position management, hedging strategies

Risk Control & Fraud fraud Syndicate identification, transaction loop closure, risk grading

🔌 API Endpoints

Pure FastAPI Version (gateway_simple.py)

Health Check http GET /health

Get Tools List http GET /mcp/tools Authorization: Bearer enterprise-agent-key-2026

Call MCP Tool http POST /mcp Authorization: Bearer enterprise-agent-key-2026 X-Agent-Role: energy_storage Content-Type: application/json

{ "name": "submit_tda", "arguments": { "csv_path": "tmp/grid_load.csv", "industry": "energy" } }

Available Tools submit_tda - Submit topological data analysis task Parameters: csv_path (file path), industry (industry type) Returns: data_id + task status get_insight - Get topological analysis results and industry instructions Parameters: data_id, agent_role (Agent role) Returns: Business insights + operational instructions + raw topological data arbitrate - Multi-Agent topological consensus arbitration Parameters: state_a, state_b (topological states of two Agents) Returns: Arbitration verdict + distance metrics + weaker Agent identification send_reward - Feed back business Reward to drive self-evolution Parameters: agent_id, industry, reward, params Returns: Cleansed Reward value

⚙️ Configuration

Environment Variables env Service Configuration PORT=8000 MAX_WORKERS=4

Storage Configuration USE_REDIS=0 REDIS_URL=redis://localhost:6379/0

Security Configuration AGENT_TOKEN=enterprise-agent-key-2026

TDA Default Parameters N_NEIGHBORS=15 MIN_DIST=0.1 HOLE_THRESHOLD=0.1 TOP_K=5

Default Token enterprise-agent-key-2026

🎯 Use Cases

Smart Manufacturing Production line multi-Agent scheduling conflict arbitration Equipment anomaly topological pattern recognition Quality fluctuation cycle detection

New Energy Energy storage cluster load scheduling Power grid peak-valley topological analysis Photovoltaic output prediction optimization

Quantitative Finance Multi-strategy Agent consensus arbitration Market cycle topological identification Risk topological early warning

Financial Risk Control Anti-fraud multi-Agent cross-validation Syndicate transaction topological detection Anomalous behavior pattern recognition

📈 Performance Features Numba JIT Acceleration: 5-10x speedup for core loops. Dual-Layer Distance Determination: Fast return in 80% of scenarios, 3-5x overall speedup. Adaptive Sampling: Automatic downsampling for large datasets to ensure response time. Dual-Mode Storage: Zero-dependency memory mode, scalable Redis mode. Thread Pool Concurrency: Supports parallel processing of multiple tasks.

🔒 Security Features File Sandbox: Only allows reading files within whitelisted directories. Token Authentication: All API calls require a Bearer Token. Audit Logs: All arbitration operations are written to audit logs. Parameter Validation: Strict type validation via Pydantic.

🚀 Version Evolution Version Positioning Core Features V1.0 Standalone Tool Basic TDA analysis + CSV input

V2.0 Enterprise Middleware FastAPI gateway + caching + authentication

V3.0 Platformization SDK + plugin system + visualization

V4.0 Consensus Protocol A2A communication + Topological Court + Evolution Flywheel

V4.1 Lite Rapid Deployment Lightweight architecture + zero dependencies + 7-day delivery

📝 Development Roadmap [ ] V4.2: Federated Topological Privacy Computing [ ] V4.5: Distributed Swarm Cluster [ ] V5.0: Cross-Industry Plugin Marketplace [ ] V5.5: Visual Topological Graph [ ] V6.0: ASI-Level Topological Awareness Kernel

🤝 Tech Stack Core Computation: NumPy, Numba, UMAP, Ripser, GUDHI Web Framework: FastAPI, Uvicorn Data Models: Pydantic Optimization Algorithms: Scikit-Optimize, POT MCP Protocol: FastMCP (Optional) Deployment: Docker

📞 Contact Project Repository: [GitHub] Technical Documentation: [Wiki] Issue Tracking: [Issues]

Slipknot V4.1 Lite - Empowering Multi-Agent Clusters with Mathematical-Grade Consensus