Input Agents (Analysis)

Input agents analyze source infrastructure code (Chef, PowerShell, or legacy Ansible) to understand configuration intent and create detailed migration specifications.

Purpose

Input agents serve as the “understanding” layer of X2A Convertor:

  • Parse source code using language-specific tools
  • Extract configuration logic, dependencies, and patterns
  • Document migration requirements in structured format
  • Validate completeness and accuracy of analysis
flowchart LR
    Source[Source Code<br/>Chef/PowerShell/Ansible] --> Detect{Technology<br/>Detection}
    Detect --> Agent[Input Agent]
    Agent --> Parse[Parse & Analyze]
    Parse --> Spec[Migration Specification]

    style Source fill:#ffebee
    style Agent fill:#e8f5e9
    style Spec fill:#e3f2fd

Agent Architecture

Technology Router

Located in src/inputs/analyze.py, routes to appropriate agent:

def analyze_project(user_requirements: str, source_dir: str):
    technology = detect_technology(source_dir)

    if technology == "chef":
        return chef_agent.analyze(user_requirements, source_dir)
    elif technology == "puppet":
        return puppet_agent.analyze(user_requirements, source_dir)
    elif technology == "salt":
        return salt_agent.analyze(user_requirements, source_dir)

Detection based on:

  • Chef: metadata.rb, Berksfile, .rb recipes
  • PowerShell: .ps1, .psm1, .psd1 scripts
  • Ansible: tasks/, handlers/, defaults/ role structure
  • Puppet: metadata.json, .pp manifests (framework ready)
  • Salt: top.sls, .sls state files (framework ready)

Available Agents

Chef Agent

Analyzes Chef cookbooks, recipes, templates, and attributes. Resolves dependencies from Chef Supermarket and generates migration specifications for converting Chef resources to Ansible modules.

PowerShell Agent

Analyzes PowerShell scripts, modules, and DSC (Desired State Configuration) configurations. Generates migration specifications for converting Windows automation to Ansible.

Ansible Agent

Modernizes legacy Ansible roles to follow current best practices. Performs in-place modernization with FQCN migration, loop updates, and module defaults.

Common Workflow Pattern

All input agents follow a similar workflow pattern:

  1. Scan Files: Discover and classify source files
  2. Analyze Structure: Extract configuration logic using LLM-powered analysis
  3. Write Report: Generate comprehensive migration specification
  4. Validate Analysis: Cross-check specification against structured analysis
  5. Cleanup Specification: Refine and finalize the migration plan

Each agent implements this workflow using a LangGraph-based state machine for reliable, observable execution.

Table of contents

Back to top