RAILWISE-CLI AI Multi-Agent System Overview

RAILWISE-CLI AI Multi-Agent System Overview

Table of Contents

• Overview
• System Architecture
• Core Agents
• Installation & Configuration
• Command Line Interface
• Workflow Orchestration
• Integration with External AI Platforms
• Best Practices
• Limitations & Notes
• FAQ
• Changelog
• Related Documents

---

Overview

RAILWISE-CLI is an AI-powered multi-agent command-line system for engineering surveying launched by Ningbo RailWise Engineering Technology Co., Ltd. (RailWise). Built on the TypeScript/Bun technology stack, it integrates multiple professional domain AI Agents, providing full-process automation capabilities from data acquisition, processing to report generation for engineering surveying and safety monitoring digitalization scenarios.

Applicable Scenarios

Scenario	Description
Rail Transit Monitoring	Automated monitoring of control protection zones during shield tunneling construction
Bridge Health Monitoring	Monitoring of viaduct pier displacement, settlement, and inclination
Excavation Engineering Monitoring	Monitoring of surrounding building and pipeline deformation for deep excavation
Tunnel Engineering Monitoring	Tunnel convergence and ground surface settlement monitoring
Data Processing	Batch cleaning, transformation, and analysis of measurement data
Report Generation	Automated compilation of monitoring daily, weekly, and monthly reports

Version Information

Item	Version
RAILWISE-CLI	1.2.0
Bun Runtime	1.1.x
TypeScript	5.5.x
AI Engine	RailWise Neural Engine v2

---

System Architecture

RAILWISE-CLI adopts a Multi-Agent Collaboration Architecture, where agents coordinate through a message bus, with an Orchestrator uniformly scheduling tasks.

┌─────────────────────────────────────────────────────────────┐
│                    RAILWISE-CLI Architecture                  │
├─────────────────────────────────────────────────────────────┤
│  ┌─────────────┐  ┌─────────────┐  ┌─────────────┐         │
│  │  Survey     │  │  Analysis   │  │  Report     │         │
│  │  Agent      │  │  Agent      │  │  Agent      │         │
│  └──────┬──────┘  └──────┬──────┘  └──────┬──────┘         │
│         │                │                │                  │
│         └────────────────┼────────────────┘                  │
│                          ▼                                  │
│              ┌─────────────────────┐                        │
│              │   Orchestrator        │                        │
│              │   Workflow Engine /   │                        │
│              │   Scheduler           │                        │
│              └─────────────────────┘                        │
│                          │                                  │
│         ┌────────────────┼────────────────┐                │
│         ▼                ▼                ▼                │
│  ┌─────────────┐  ┌─────────────┐  ┌─────────────┐       │
│  │  Data       │  │  Model      │  │  External   │       │
│  │  Storage    │  │  Inference  │  │  Interface  │       │
│  │  PostgreSQL │  │  ONNX/WASM  │  │  API/SDK    │       │
│  └─────────────┘  └─────────────┘  └─────────────┘       │
└─────────────────────────────────────────────────────────────┘

Technology Stack

Layer	Technology	Description
Runtime	Bun	High-performance JavaScript runtime
Language	TypeScript	Type-safe development experience
AI Inference	ONNX Runtime / WASM	Local model inference, no cloud dependency
Data Storage	PostgreSQL / SQLite	Structured data persistence
Message Communication	Built-in EventBus	Asynchronous inter-agent communication
Configuration Management	TOML / JSON	Flexible configuration file formats

---

Core Agents

1. Survey Agent

The Survey Agent is responsible for interacting with total stations, levels, and other measurement equipment to execute automated measurement tasks.

Core Capabilities:

• Total station auto-targeting and reading
• Real-time measurement data acquisition and validation
• Intelligent optimization of measurement plans
• Equipment status monitoring and anomaly alerting

Example Invocation:

# Launch single measurement task
railwise-cli survey execute   --project "Raocheng Highway Utility Tunnel Project Section 1"   --station "TS-01"   --points "P1,P2,P3,P4,P5"   --rounds 3   --output-format json

# Launch continuous monitoring mode
railwise-cli survey monitor   --project "Line 3 Control Protection Zone"   --interval 300   --duration 4h   --alert-threshold 2.0

2. Analysis Agent

The Analysis Agent is responsible for in-depth analysis of measurement data, identifying trends, anomalies, and potential risks.

Core Capabilities:

• Time-series trend analysis
• Deformation rate calculation and prediction
• Anomaly detection (based on statistics and machine learning)
• Multi-point correlation analysis
• Automatic alert level determination

Example Invocation:

# Execute data analysis
railwise-cli analyze trend   --project "Line 3 Control Protection Zone"   --start-date 2026-07-01   --end-date 2026-07-08   --metrics "displacement,velocity,acceleration"   --output report.html

# Anomaly detection
railwise-cli analyze anomaly   --dataset "survey_data_2026Q3"   --method isolation-forest   --sensitivity high

3. Report Agent

The Report Agent is responsible for automatically generating professional monitoring reports based on analysis results.

Core Capabilities:

• Automated generation of daily, weekly, and monthly monitoring reports
• Automatic chart drawing and layout
• Automatic citation of standard clauses
• Multi-format output (PDF/Word/HTML)
• Report quality self-check

Example Invocation:

# Generate daily report
railwise-cli report daily   --project "Raocheng Highway Utility Tunnel Project Section 1"   --date 2026-07-08   --template "gb50911-2013"   --output "Daily_Report_20260708.pdf"

# Generate weekly report
railwise-cli report weekly   --project "Line 3 Control Protection Zone"   --week 2026-W27   --include-charts   --output "Weekly_Report_2026W27.docx"

4. Coordinator Agent

The Coordinator Agent is responsible for task scheduling and resource allocation among multiple agents.

Core Capabilities:

• Workflow definition and execution
• Task priority management
• Agent load balancing
• Error recovery and retry
• Execution log recording

---

Installation & Configuration

Environment Requirements

Item	Minimum	Recommended
OS	macOS 12+ / Windows 10+ / Linux	macOS 14 / Windows 11 / Ubuntu 22.04
Bun	1.1.0	Latest stable
Memory	4 GB	8 GB
Disk	2 GB free	10 GB free
Network	Access to RailWise API	Stable internet connection

Installation Steps

# Install globally with npm
npm install -g @railwise/cli

# Or install with Bun
bun install -g @railwise/cli

# Verify installation
railwise-cli --version
# Output: RAILWISE-CLI v1.2.0

# Initialize configuration
railwise-cli init

Configuration File

Configuration file located at ~/.railwise/config.toml:

[core]
workspace = "/Users/wangjiawei/RailWise/projects"
log_level = "info"
timezone = "Asia/Shanghai"

[api]
base_url = "https://api.railwise.cn/v2"
api_key = "rw_live_xxxxxxxxxxxxxxxx"
timeout = 30

[ai]
model_path = "~/.railwise/models"
enable_local_inference = true
max_concurrent_agents = 4

[survey]
default_instrument = "total_station"
measurement_rounds = 3
precision_threshold = 0.5

[report]
default_template = "gb50911-2013"
output_format = "pdf"
company_name = "Ningbo RailWise Engineering Technology Co., Ltd."

Note: First run of railwise-cli init will guide you through configuration file generation. API Key can be obtained from the RailWise Developer Platform.

---

Command Line Interface

Global Options

railwise-cli [global options] <command> [subcommand] [options]

Global Options:
  -v, --version          Show version number
  -h, --help             Show help information
  -c, --config <path>    Specify configuration file path
  -o, --output <path>    Specify output path
  -f, --format <format>  Output format: json, yaml, table, html
  --verbose              Show detailed logs
  --dry-run              Simulate run, no actual operations

Command List

Command	Description	Subcommands
survey	Survey operations	execute, monitor, calibrate, status
analyze	Data analysis	trend, anomaly, correlation, forecast
report	Report generation	daily, weekly, monthly, custom
agent	Agent management	list, status, start, stop, logs
workflow	Workflow management	run, list, create, validate
config	Configuration management	get, set, reset, validate
auth	Authentication management	login, logout, status, refresh

Usage Examples

# View all agent statuses
railwise-cli agent list

# Launch specific workflow
railwise-cli workflow run   --name "daily_monitoring"   --project "Line 3 Control Protection Zone"   --schedule "0 */4 * * *"

# Validate configuration file
railwise-cli config validate

---

Workflow Orchestration

RAILWISE-CLI supports defining complex workflows through YAML files to achieve multi-agent collaboration.

Workflow Definition File

workflow_daily_monitoring.yaml

name: daily_monitoring
version: "1.0"
description: Daily monitoring workflow — execute every 4 hours

schedule:
  cron: "0 */4 * * *"
  timezone: "Asia/Shanghai"

project:
  id: "proj_rw_2026_001"
  name: "Raocheng Highway Utility Tunnel Project Section 1"

steps:
  - id: step_1
    name: "Data Acquisition"
    agent: survey
    action: execute
    params:
      station: "TS-01"
      points: ["P1", "P2", "P3", "P4", "P5"]
      rounds: 3
    timeout: 1800
    retry:
      max_attempts: 3
      backoff: exponential

  - id: step_2
    name: "Data Analysis"
    agent: analysis
    action: trend
    depends_on: [step_1]
    params:
      metrics: ["displacement", "velocity"]
      alert_threshold: 2.0
    timeout: 300

  - id: step_3
    name: "Report Generation"
    agent: report
    action: daily
    depends_on: [step_2]
    params:
      template: "gb50911-2013"
      include_charts: true
    timeout: 120
    condition:
      if: "step_2.alert_level != 'normal'"
      then: "continue"
      else: "continue"

  - id: step_4
    name: "Alert Notification"
    agent: notification
    action: send
    depends_on: [step_2]
    condition:
      if: "step_2.alert_level in ['yellow', 'orange', 'red']"
    params:
      channels: ["email", "sms", "app"]
      template: "alert_{{alert_level}}"

Execute Workflow

# Manual workflow execution
railwise-cli workflow run --file workflow_daily_monitoring.yaml

# Register scheduled workflow
railwise-cli workflow register --file workflow_daily_monitoring.yaml

# View workflow execution history
railwise-cli workflow logs --name daily_monitoring --limit 10

---

Integration with External AI Platforms

RAILWISE-CLI supports integration with mainstream AI development platforms through MCP (Model Context Protocol).

Supported AI Platforms

Platform	Integration Method	Status
Cursor	MCP Server	✅ Supported
Claude Desktop	MCP Server	✅ Supported
Windsurf	MCP Server	✅ Supported
VS Code + Copilot	Extension plugin	🔄 In development
JetBrains	Plugin	📋 Planned

MCP Server Configuration

~/.cursor/mcp.json

{
  "mcpServers": {
    "railwise": {
      "command": "railwise-cli",
      "args": ["mcp", "serve"],
      "env": {
        "RAILWISE_API_KEY": "rw_live_xxxxxxxxxxxxxxxx"
      }
    }
  }
}

After configuration, RAILWISE-CLI functionality can be invoked through natural language in Cursor:

User: "Execute measurement task for Line 3 control protection zone"
AI: Call railwise-cli survey execute --project "Line 3 Control Protection Zone" ...

See MCP Server Integration Guide for details.

---

Best Practices

1. Configuration Management

• Use environment variables for sensitive information (API Key, passwords)
• Include project configurations and workflow definitions in version control
• Use independent configuration files for different environments (dev/test/prod)

# Environment variable examples
export RAILWISE_API_KEY="rw_live_xxxxxxxxxxxxxxxx"
export RAILWISE_WORKSPACE="/data/railwise/prod"

2. Workflow Design

• Break complex tasks into multiple small steps for easier debugging and retry
• Set reasonable timeout and retry strategies for each step
• Use depends_on to clarify dependencies between steps
• Add conditional judgments to key steps to avoid unnecessary execution

3. Performance Optimization

• Enable local model inference (enable_local_inference = true) to reduce network latency
• Reasonably set max_concurrent_agents to avoid resource contention
• Use --batch-size parameter for chunked processing with large data volumes

4. Logging and Monitoring

# View real-time logs
railwise-cli agent logs --follow

# Export execution report
railwise-cli workflow report --name daily_monitoring --format html

---

Limitations & Notes

Limitation Item	Description
Concurrent agent count	Default maximum 4, adjustable via configuration
Single measurement point count	Recommended not to exceed 500 points
Report generation size	Single report not to exceed 100 pages
Local model support	Only ONNX format models supported
Network dependency	Some features require connection to RailWise API
Operating system	ARM32 architecture not currently supported

Note: Before using in production environments, please verify workflow correctness in test environments.

---

FAQ

Q1: Command not found after installation?

Ensure Bun/npm global installation path is added to system PATH:

# Check installation path
which railwise-cli
# or
where railwise-cli

# If not found, add path
export PATH="$PATH:$(bun pm bin -g)"

Q2: How to handle API Key expiration?

# Re-login
railwise-cli auth login

# Or manually refresh Token
railwise-cli auth refresh

Q3: How to troubleshoot workflow execution failure?

# View detailed logs
railwise-cli workflow logs --name <workflow_name> --verbose

# Validate workflow definition
railwise-cli workflow validate --file <workflow_file>

# Single-step debugging
railwise-cli workflow run --file <workflow_file> --step <step_id> --dry-run

Q4: How to customize report templates?

# Export default template
railwise-cli report template --export --name gb50911-2013 --output ./templates/

# Use custom template after modification
railwise-cli report daily --template ./templates/my_custom_template.html

Q5: Which total station brands are supported?

Currently supported total station brands for automated interfaces:

• Leica (徕卡)
• Trimble (天宝)
• Topcon (拓普康)
• Sokkia (索佳)
• South (南方测绘)

Other brands please contact technical support for adaptation solutions.

---

Changelog

Version	Date	Changes
1.2.0	2026-07-08	Added MCP Server support; optimized workflow engine; added anomaly detection algorithm
1.1.0	2026-05-20	Added report agent; supported custom templates; optimized measurement precision
1.0.0	2026-03-15	Initial release; supported survey, analysis, and report agents

---

Related Documents

• WorkWise AI Workbench Integration Guide
• RAILWISE-TSM Total Station Automated AI Tools
• RAILWISE-OS Monitoring Business AI Assistant
• MCP Server Integration Guide
• API REST Reference Documentation
• SDK TypeScript Guide

---

Technical Support: For questions, please contact the RailWise Technical Support Team

• Email: support@railwise.cn
• Phone: 0574-XXXXXXXX
• Documentation Feedback: GitHub Issues