986 5 months ago

Qwen2.5-Coder-SysML is a specialized language model designed to generate, interpret, and reason about SysML v2 models. It bridges natural language requirements with formal SysML v2 textual syntax, enabling automatic or semi-automatic generation of systems

tools
ollama run Babitdor/Qwen2.5-Coder-SysMLv2

Applications

Claude Code
Claude Code ollama launch claude --model Babitdor/Qwen2.5-Coder-SysMLv2
Codex
Codex ollama launch codex --model Babitdor/Qwen2.5-Coder-SysMLv2
OpenCode
OpenCode ollama launch opencode --model Babitdor/Qwen2.5-Coder-SysMLv2
OpenClaw
OpenClaw ollama launch openclaw --model Babitdor/Qwen2.5-Coder-SysMLv2

Models

View all →

Readme

Qwen2.5-Coder-7B-SysMLv2

Qwen2.5-Coder-SysMLv2 is a domain-specialized model adapted from Qwen 2.5 Coder to generate, understand, and reason about SysML v2 textual modeling. This model focuses specifically on structure modeling, option-space exploration, and low-syntax-error generation for automated MBSE pipelines.


๐Ÿš€ Overview

Property Description
Base Model Qwen 2.5 Coder
Fine-tuned Domain SysML v2 structural modeling + reasoning
Languages English + SysML v2 textual syntax
Primary Use Multi-option SysML structure generation
Format Support SysML v2, JSON
Deployment Ready vLLM โ€ข Ollama โ€ข LangGraph โ€ข MCP

๐ŸŽฏ Purpose

Qwen2.5-Coder-SysMLv2 is optimized for scenarios where you want to turn natural language requirements into SysML v2 architecture options โ€” reliably โ€” with significantly reduced syntax issues compared to direct SysML generation.

This model is especially suited for agentic orchestration pipelines where the SysML is either:

  1. directly returned as SysML syntax or
  2. generated as JSON โ†’ converted into valid SysML v2

๐Ÿง  Capabilities

  • โœ… SysML v2 Structure Generation Generates SysML v2 package / part definitions from requirement text.

  • ๐Ÿ” Multiple Architecture Options Generates usually 3 alternative model structures per prompt (for architectural trade-off exploration).

  • ๐Ÿงฉ Pipeline Friendly Supports JSON โ†’ XMI โ†’ SysML v2 conversion workflows (lowest syntax error mode).

  • ๐Ÿ” Domain Reasoning Good at structural decomposition of warehouses, manufacturing, logistics, robotics, automation, sensor networks, cyber-physical systems.

  • ๐Ÿงฑ Toolchain Integration Ready

    • LangGraph โ†’ multi-step modeling agents
    • MCP โ†’ SysML diagram rendering pipeline
    • vLLM & Ollama โ†’ inference serving

๐Ÿงฉ Example Usage Prompt Pattern

prompt

You are a SysML v2 structural model architect.

Given the requirement text below:
- generate 3 candidate SysML v2 structural architecture options
- do NOT include behavior
- output only valid SysML v2 textual syntax

Requirement:
"The warehouse needs an automated guided vehicle fleet, inventory storage racks, and a central planning server that dispatches missions to AGVs."

Example Output

Candidate 1: Hierarchical Design with Centralized Control package WarehouseSystem {

    // ** Requirements Section **
    requirement def <'REQ-WH-001'> DispatchingAccuracyRequirement {
        doc /* The system must ensure that AGVs reach their designated targets within ยฑ5 meters of accuracy. */
    }

    // ** Physical Objects Section **
    part def WarehouseRack {
        attribute totalCapacity :> ISQ::count = 5000; // Slots for storing items
    }

    part def AutomatedGuidedVehicle {
        attribute batteryLifeHours : Real = 12.0; // Duration before charging required
        attribute navigationPrecisionMeters : Real = 0.05; // Maximum deviation from planned path
    }

    // ** Architecture Section (Hierarchical) **
    analysis def WarehouseFleetArchitecture {
        subject system;

        part node warehouseNetwork {
            part server: CentralPlanningServer;
            part rack: WarehouseRack;
            part *vehicle: AutomatedGuidedVehicle {
                part batteryModule; // Aggregates the battery
            }
        }

        flow dispatchingFlow from server to vehicle.batteryModule {
            then flow navigationFlow from vehicle.batteryModule to rack {
                then flow materialTransferFlow from rack to vehicle.batteryModule;
            }
        }

        assert requirement <'REQ-WH-001'> of dispatchingFlow;
    }
}

Candidate 2: Distributed Design with Decentralized Control package WarehouseSystem {

    // ** Requirements Section **
    requirement def <'REQ-WH-001'> DispatchingAccuracyRequirement {
        doc /* The system must ensure that AGVs reach their designated targets within ยฑ5 meters of accuracy. */
    }

    // ** Physical Objects Section **
    part def WarehouseRack {
        attribute totalCapacity :> ISQ::count = 5000; // Slots for storing items
    }

    part def AutomatedGuidedVehicle {
        attribute batteryLifeHours : Real = 12.0; // Duration before charging required
        attribute navigationPrecisionMeters : Real = 0.05; // Maximum deviation from planned path
        part serverInterface; // Aggregates interface with Central Planning Server
    }

    // ** Architecture Section (Distributed) **
    analysis def WarehouseFleetArchitecture {
        subject system;

        part node warehouseNetwork {
            part rack: WarehouseRack;
            part *vehicle: AutomatedGuidedVehicle {
                part batteryModule; // Aggregates the battery
                part serverInterface; // Aggregates interface with Central Planning Server
            }
        }

        flow dispatchingFlow from serverInterface to vehicle.batteryModule {
            then flow navigationFlow from vehicle.batteryModule to rack {
                then flow materialTransferFlow from rack to vehicle.batteryModule;
            }
        }

        flow controlFlow from CentralPlanningServer.serverInterface to AutomatedGuidedVehicle.serverInterface;
        part CentralPlanningServer {
            :>> vehicle;
            part serverInterface; // Aggregates interface for communication
        }
    }
}

Candidate 3: Microservices Style Design with Autonomous AGVs

package WarehouseSystem {
    // ** Requirements Section **
    requirement def <'REQ-WH-001'> DispatchingAccuracyRequirement {
        doc /* The system must ensure that AGVs reach their designated targets within ยฑ5 meters of accuracy. */
    }

    // ** Physical Objects Section **
    part def AutonomousVehicleController {
        attribute controlLatencySeconds : Real = 0.1; // Low latency for real-time control
    }

    part def WarehouseRack {
        attribute totalCapacity :> ISQ::count = 5000; // Slots for storing items
    }

    part def AutomatedGuidedVehicle {
        attribute batteryLifeHours : Real = 12.0; // Duration before charging required
        attribute navigationPrecisionMeters : Real = 0.05; // Maximum deviation from planned path
        part autonomousController: AutonomousVehicleController;
        part batteryModule; // Aggregates the battery
    }

    // ** Architecture Section (Microservices Style) **
    analysis def WarehouseFleetArchitecture {
        subject system;

        part node warehouseNetwork {
            part rack: WarehouseRack;
            part *vehicle: AutomatedGuidedVehicle {
                part autonomousController: AutonomousVehicleController;
                part batteryModule; // Aggregates the battery
            }
        }

        flow dispatchingFlow from CentralPlanningServer.serverInterface to AutomatedGuidedVehicle.autonomousController {
            then flow navigationFlow from AutonomousVehicleController to rack {
                then flow materialTransferFlow from rack to vehicle.batteryModule;
            }
        }

        part CentralPlanningServer {
            part serverInterface; // Aggregates interface for communication
        }

        part ChargingStation {
            attribute chargingRateAmps : Real = 50.0; // Amps per vehicle
            part serverInterface; // Aggregates interface for vehicle registration
        }
    }
}


Notes

This model is intended to be chained inside agent graph / multi-step workflows โ€” NOT to replace formal SysML tools โ€” but to massively accelerate first-pass structure modeling, alternative exploration, and early model formation inside engineering loops.

This is a research-focused domain adaptation intended for actual industrial MBSE work.