🚀 Getting Started with AgentOS

"""
CSV Processor Agent
Generated by AgentOS Copilot Extension
Safety Level: Standard | Policy: data-processing-v1
"""

from agent_os import KernelSpace, Policy
from agent_os.tools import create_safe_toolkit
from agent_os.validators import DataValidator

# Initialize kernel with safety guarantees
kernel = KernelSpace(
    policy="data-processing",
    audit_log=True,
    rate_limit="100/minute"
)

# Create toolkit with restricted file access
toolkit = create_safe_toolkit(
    preset="file-processor",
    allowed_extensions=[".csv", ".json"],
    max_file_size="50MB",
    output_dir="/tmp/agent-output"
)

@kernel.agent(
    name="csv-processor",
    version="1.0.0",
    description="Processes CSV files and generates summary reports"
)
async def csv_processor(input_path: str, output_format: str = "json"):
    """
    Process CSV files with full safety guarantees.
    
    Args:
        input_path: Path to the input CSV file
        output_format: Output format (json, markdown, html)
    
    Returns:
        dict: Processing results with metrics and summary
    """
    
    # Validate input (policy-checked)
    validator = DataValidator(toolkit)
    if not await validator.validate_file(input_path):
        return {"status": "error", "message": "Invalid input file"}
    
    # Read CSV file (restricted to allowed paths)
    data = await toolkit.file.read_csv(input_path)
    
    # Extract metrics using LLM (rate-limited, content-filtered)
    analysis = await toolkit.llm.analyze(
        model="gpt-4",
        prompt=f"""Analyze this CSV data and extract key metrics:
        Columns: {data.columns.tolist()}
        Row count: {len(data)}
        Sample: {data.head(5).to_dict()}
        
        Provide: summary statistics, trends, anomalies""",
        max_tokens=1000
    )
    
    # Generate report
    report = {
        "status": "success",
        "input_file": input_path,
        "row_count": len(data),
        "column_count": len(data.columns),
        "columns": data.columns.tolist(),
        "analysis": analysis,
        "generated_at": toolkit.utils.timestamp()
    }
    
    # Write output (restricted to /tmp/agent-output)
    output_path = f"/tmp/agent-output/report.{output_format}"
    await toolkit.file.write(output_path, report, format=output_format)
    
    # Log completion for audit trail
    kernel.audit.log("processing_complete", {
        "input": input_path,
        "output": output_path,
        "metrics": report
    })
    
    return report


if __name__ == "__main__":
    import asyncio
    
    result = asyncio.run(
        kernel.execute(csv_processor, "data/sales.csv", "json")
    )
    print(f"Processing complete: {result['status']}")

policy:
  name: "CSV Processor Policy"
  version: "1.0"
  agent: "csv-processor"
  
  # File System Rules
  filesystem:
    read:
      - pattern: "data/**/*.csv"
        action: allow
      - pattern: "data/**/*.json"
        action: allow
      - pattern: "**/*"
        action: deny
        message: "File access restricted to data/ directory"
    
    write:
      - pattern: "/tmp/agent-output/**"
        action: allow
      - pattern: "**/*"
        action: deny
        message: "Write access restricted to /tmp/agent-output/"
  
  # API Rules
  api:
    llm:
      rate_limit: "100/minute"
      max_tokens: 2000
      allowed_models:
        - "gpt-4"
        - "gpt-3.5-turbo"
      content_filter: true
    
    external:
      - domain: "*.openai.com"
        action: allow
      - domain: "*"
        action: deny
        message: "External API calls restricted"
  
  # Data Rules
  data:
    pii_detection: true
    pii_action: redact
    max_file_size: "50MB"
    allowed_formats:
      - csv
      - json
  
  # Audit Configuration
  audit:
    enabled: true
    log_level: "info"
    retention_days: 90
    events:
      - file_read
      - file_write
      - llm_call
      - error

name: CSV Processor Agent

on:
  workflow_dispatch:
    inputs:
      input_file:
        description: 'Path to input CSV file'
        required: true
        default: 'data/sales.csv'
      output_format:
        description: 'Output format'
        required: false
        default: 'json'
        type: choice
        options:
          - json
          - markdown
          - html
  
  # Optional: Run on schedule
  schedule:
    - cron: '0 6 * * 1'  # Every Monday at 6 AM

  # Optional: Run on file changes
  push:
    paths:
      - 'data/**/*.csv'

jobs:
  process:
    runs-on: ubuntu-latest
    permissions:
      contents: read
      actions: write
    
    steps:
      - name: Checkout repository
        uses: actions/checkout@v4
      
      - name: Set up Python
        uses: actions/setup-python@v5
        with:
          python-version: '3.11'
          cache: 'pip'
      
      - name: Install AgentOS
        run: |
          pip install agent-os
          pip install -r agents/csv-processor/requirements.txt
      
      - name: Validate Agent Policy
        run: |
          agent-os validate agents/csv-processor/policy.yaml
      
      - name: Run CSV Processor Agent
        env:
          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
          AGENT_OS_AUDIT_LOG: true
        run: |
          python agents/csv-processor/agent.py \
            --input "${{ github.event.inputs.input_file || 'data/sales.csv' }}" \
            --format "${{ github.event.inputs.output_format || 'json' }}"
      
      - name: Upload Results
        uses: actions/upload-artifact@v4
        with:
          name: processing-results
          path: /tmp/agent-output/
          retention-days: 30
      
      - name: Upload Audit Log
        uses: actions/upload-artifact@v4
        with:
          name: audit-log
          path: /tmp/agent-output/.audit/
          retention-days: 90