dcisionai-mcp-platform

ameydhavle/dcisionai-mcp-platform

3.2

If you are the rightful owner of dcisionai-mcp-platform and would like to certify it and/or have it hosted online, please leave a comment on the right or send an email to henry@mcphub.com.

DcisionAI MCP Server Platform is a production-ready server platform designed for manufacturing optimization, utilizing a 6-agent swarm intelligence system and comprehensive workflow orchestration.

Tools
5
Resources
0
Prompts
0

DcisionAI Platform

Overview

DcisionAI is the Intelligent Enterprise Decision Layer - a revolutionary platform that bridges the gap between AI chatbots and spreadsheets, providing mathematically proven optimal decisions with full transparency and interactivity.

Live Platform: https://platform.dcisionai.com

Architecture

Current Architecture (AgentCore Gateway + Qwen 30B + MCP Server)

  • AgentCore Gateway: Amazon Bedrock AgentCore Gateway for advanced agent orchestration
  • Qwen 30B Model: Qwen 3B Coder 30B for superior mathematical model generation
  • Enhanced Lambda Functions: 2GB memory, optimized for Qwen 30B integration
  • Real-time Processing: Live optimization with progress tracking
  • MCP Protocol: Model Context Protocol for seamless agent communication
  • MCP Server Distribution: Production-ready Python package for global distribution

Key Features

  • ✅ Qwen 30B Integration: Advanced mathematical model generation with Qwen 3B Coder 30B
  • ✅ Real AWS Bedrock Integration: No mock responses, all real AI models
  • ✅ AgentCore Gateway: Next-generation agent platform with extended execution
  • ✅ 21 Industry Workflows: Predefined optimization templates across 7 industries
  • ✅ Production Ready: AWS-hosted with global CDN distribution
  • ✅ MCP Server Distribution: pip install dcisionai-mcp-server for global access

MCP Server Distribution

Quick Installation

pip install dcisionai-mcp-server

Start Server

dcisionai-mcp-server start --host 0.0.0.0 --port 8000

Available Tools

  • classify_intent - Intent classification for optimization requests
  • analyze_data - Data analysis and preprocessing
  • build_model - Mathematical model building with Qwen 30B
  • solve_optimization - Optimization solving and results
  • get_workflow_templates - Industry workflow templates
  • execute_workflow - End-to-end workflow execution

IDE Integration

Works with Cursor, Kiro, Claude Code, VS Code, and other MCP-compatible IDEs.

Documentation

  • Package Documentation:
  • API Reference:
  • Quick Start:

📁 Repository Structure

dcisionai-mcp-platform/
├── main.py                          # Main platform entry point
├── runtime_config.json              # Runtime configuration
├── requirements.txt                 # Main requirements
├── dcisionai-mcp-server/            # MCP Server Distribution Package
│   ├── setup.py                     # PyPI package configuration
│   ├── pyproject.toml               # Modern Python packaging
│   ├── requirements.txt             # Package dependencies
│   ├── README.md                    # Package documentation
│   ├── dcisionai_mcp_server/        # Main package
│   │   ├── __init__.py             # Package initialization
│   │   ├── server.py               # Main MCP server
│   │   ├── tools.py                # 6 core optimization tools
│   │   ├── config.py               # Configuration management
│   │   ├── workflows.py            # Workflow manager
│   │   └── cli.py                  # Command-line interface
│   └── tests/                       # Test suite
├── domains/                         # Core domains
│   ├── manufacturing/               # Manufacturing domain
│   │   ├── agents/                  # AgentCore agents
│   │   │   └── DcisionAI_Manufacturing_Agent_v4.py
│   │   ├── mcp_server/              # MCP server components
│   │   │   ├── manufacturing_*_swarm.py  # Swarm implementations
│   │   │   ├── swarm_inference_profile.py
│   │   │   ├── consensus_mechanism.py
│   │   │   └── inference_profile_enhanced_swarm.py
│   │   └── tools/                   # Core manufacturing tools
│   │       ├── intent/              # Intent classification
│   │       ├── data/                # Data analysis
│   │       ├── model/               # Model building
│   │       └── solver/              # Optimization solving
│   ├── finance/                     # Finance domain
│   └── pharma/                      # Pharma domain
├── platform_core/                   # Platform management
├── shared/                          # Shared framework
├── tests/                           # Test suite
│   ├── unit/                        # Unit tests
│   ├── integration/                 # Integration tests
│   └── workflow/                    # Workflow tests
├── docs/                            # Documentation
│   ├── Architecture.md              # System architecture
│   ├── CUSTOMER_QUICK_START_GUIDE.md # Quick start guide
│   ├── index.md                     # Documentation index
│   └── openapi_specification.yaml   # API specification
└── archive/                         # Archived content
    ├── cleanup_archive/             # Previous cleanup
    ├── old_tests/                   # Archived test files
    ├── old_deployment/              # Archived deployment files
    ├── old_infrastructure/          # Archived infrastructure
    ├── old_documentation/           # Archived documentation
    ├── old_playground/              # Archived playground
    └── old_customer_portal/         # Archived customer portal

🚀 Quick Start

1. Run Customer Showcase (Recommended)

python3 run_customer_showcase.py

This will run a comprehensive customer demonstration with:

  • Real AWS Bedrock inference profiles
  • 18-agent swarm architecture
  • 4 real-world manufacturing scenarios
  • Complete optimization workflow
  • Performance metrics and health monitoring
  • No mock responses

Quick Mode: python3 run_customer_showcase.py --quick (2 scenarios, ~6-10 minutes)

2. Run End-to-End Test

python3 test_customer_scenario_e2e.py

This will run a complete E2E test with:

  • Real AWS Bedrock inference profiles
  • 18-agent swarm architecture
  • ACME Manufacturing scenario
  • No mock responses

3. Run Platform Demo

python3 main.py

This will demonstrate the platform architecture and domain management.

4. Deploy to AgentCore

cd domains/manufacturing/agents
python3 DcisionAI_Manufacturing_Agent_v4.py

🧪 Testing

Customer Showcase (Recommended)

  • File: test_customer_showcase_e2e.py
  • Runner: run_customer_showcase.py
  • Purpose: Comprehensive customer demonstration with real-world scenarios
  • Duration: ~15-20 minutes (full), ~6-10 minutes (quick mode)
  • Coverage: 4 manufacturing scenarios, complete 18-agent swarm workflow
  • Features: Performance metrics, health monitoring, detailed reporting

Main E2E Test

  • File: test_customer_scenario_e2e.py
  • Purpose: Complete workflow testing with real AWS Bedrock
  • Duration: ~15 minutes (with timeout considerations)
  • Coverage: Intent → Data → Model → Solver pipeline

Test Suite

  • Location: tests/ directory
  • Types: Unit, Integration, Workflow tests
  • Framework: Organized test structure

🔧 Configuration

Runtime Configuration

  • File: runtime_config.json
  • Purpose: AgentCore deployment configuration
  • Settings: Runtime name, image URI, role ARN

Requirements

  • Main: requirements.txt
  • AgentCore: domains/manufacturing/mcp_server/requirements.txt
  • Dependencies: AWS Bedrock, optimization solvers, MCP framework

📊 Performance

Recent E2E Test Results

  • Intent Classification: Variable (5 agents, real AWS Bedrock confidence scores)
  • Data Analysis: Variable (3 agents, real swarm consensus scores)
  • Model Building: Variable (4 agents, with timeout handling)
  • Solver Optimization: Variable (6 agents, with timeout handling)

Note: All execution times and confidence scores are real results from AWS Bedrock and swarm consensus. No mock data is used.

Timeout Configuration

  • Model Building & Solver: 300 seconds (5 minutes)
  • Other Agents: 120 seconds (2 minutes)
  • Connection: 60 seconds
  • Retries: 3 attempts with 1-second delay

🏭 Manufacturing Capabilities

Supported Industries

  • Automotive Parts Manufacturing
  • Production Line Optimization
  • Capacity Planning
  • Quality Control
  • Supply Chain Optimization

Optimization Types

  • Linear Programming (OR-Tools GLOP)
  • Mixed Integer Programming (OR-Tools SCIP)
  • High-Performance LP (OR-Tools HiGHS)
  • Python-based Optimization (PuLP)
  • Convex Optimization (CVXPY)

🔗 Integration

AWS Services

  • Bedrock: Real inference profiles
  • AgentCore: Runtime deployment
  • Cross-Region: Multi-region execution

Protocols

  • MCP: Model Context Protocol
  • FastMCP: Production MCP framework
  • HTTP: RESTful API endpoints

📚 Documentation

  • Customer Showcase: CUSTOMER_SHOWCASE_GUIDE.md - Complete customer demonstration guide
  • Architecture: docs/Architecture.md - Technical architecture and design
  • Quick Start: docs/CUSTOMER_QUICK_START_GUIDE.md - Get up and running in 5 minutes
  • API Spec: docs/openapi_specification.yaml - Complete API reference
  • Index: docs/index.md - Documentation overview

🗂️ Archive

All non-essential files have been moved to the archive/ directory:

  • cleanup_archive/: Previous cleanup efforts
  • old_tests/: Archived test files
  • old_deployment/: Archived deployment scripts
  • old_infrastructure/: Archived infrastructure files
  • old_documentation/: Archived documentation
  • old_playground/: Archived playground files
  • old_customer_portal/: Archived customer portal

🚨 Known Issues

Timeout Handling

  • Model building and solver tasks may experience timeouts
  • Timeout configuration has been increased to 300 seconds
  • System continues with consensus even if some agents timeout

Performance Optimization

  • Complex mathematical tasks require longer processing time
  • Consider using different models for computationally intensive tasks
  • Monitor AWS Bedrock quota and throttling

📞 Support

For issues or questions:

  1. Check the documentation in docs/
  2. Review test results in archive/old_tests/
  3. Examine the cleanup analysis in REPO_CLEANUP_ANALYSIS.md

DcisionAI Team | Copyright (c) 2025 DcisionAI. All rights reserved.