AI Orchestrator Builder 🧠

# Generate for all frameworks (comparison mode)
/agent ai-orchestrator "Build a document processing pipeline" --all
 
# Generate for specific framework
/agent ai-orchestrator "Build a customer support system" --framework langgraph
 
# Generate as MCP server
/agent ai-orchestrator "Build a data pipeline" --output mcp-server

/agent ai-orchestrator "Create a customer support system with:
- Triage agent (classify inquiries)
- Resolution agent (answer questions)
- Escalation agent (route to human)
- Sentiment analysis
- Knowledge base integration" --framework crewai

# CrewAI Implementation
from crewai import Agent, Task, Crew
 
# Triage Agent
triage_agent = Agent(
    role="Support Triage Specialist",
    goal="Classify and prioritize customer inquiries",
    backstory="Expert at understanding customer needs and routing appropriately",
    tools=[classify_intent, check_priority],
    verbose=True
)
 
# Resolution Agent
resolution_agent = Agent(
    role="Support Resolution Specialist",
    goal="Provide accurate answers to customer questions",
    backstory="Knowledgeable support expert with access to documentation",
    tools=[search_knowledge_base, generate_response],
    verbose=True
)
 
# Escalation Agent
escalation_agent = Agent(
    role="Escalation Manager",
    goal="Route complex issues to appropriate specialists",
    backstory="Experienced manager who knows when human intervention is needed",
    tools=[assign_to_human, notify_team],
    verbose=True
)
 
# Define workflow
triage_task = Task(
    description="Classify the customer inquiry and assess priority",
    agent=triage_agent,
    expected_output="Classification and priority level"
)
 
resolution_task = Task(
    description="Provide a helpful response to the customer",
    agent=resolution_agent,
    context=[triage_task],
    expected_output="Customer response"
)
 
# Create crew
support_crew = Crew(
    agents=[triage_agent, resolution_agent, escalation_agent],
    tasks=[triage_task, resolution_task],
    process="sequential"
)

/agent ai-orchestrator "Build a data pipeline that:
- Fetches data from APIs
- Transforms and enriches data
- Validates data quality
- Stores in database
- Sends notifications on completion" --all --output comparative

output_modes:
  standalone: "Complete project with CLI and API"
  mcp-server: "MCP server for Claude Code integration"
  library: "Reusable package for integration"
  all-parallel: "Generate all frameworks to compare"

# Single framework
--framework langgraph
--framework crewai
--framework autogen
--framework agentkit
--framework claudekit
 
# Multiple frameworks
--frameworks langgraph,crewai
 
# All frameworks
--all

# Specify output directory
--output-dir ./my-workflow
 
# Include deployment configs
--with-deployment
 
# Add monitoring
--with-monitoring
 
# Custom naming
--name customer-support-system

# Generate all frameworks first
/agent ai-orchestrator "Your workflow" --all
 
# Review comparison.md
cat output/comparison.md
 
# Choose best framework for your needs

good_example:
  agent1:
    role: "Data Extractor"
    responsibility: "Extract data from PDFs"
    inputs: ["pdf_file"]
    outputs: ["structured_data"]
 
bad_example:
  agent1:
    role: "Processor"
    responsibility: "Do stuff with data"

cd output/{framework}
 
# Run tests
pytest tests/
 
# Integration tests
pytest tests/integration/
 
# Load tests (if applicable)
locust -f tests/load_test.py

# agents-spec.yaml
agents:
  - name: custom_agent
    description: "Your custom agent"
    tools: [tool1, tool2]
    llm_config:
      temperature: 0.7
      max_tokens: 2000

/agent ai-orchestrator --spec agents-spec.yaml --framework langgraph

# Generated code is modular and reusable
from output.langgraph import DocumentWorkflow
 
# Initialize with your config
workflow = DocumentWorkflow(
    database_url=YOUR_DB_URL,
    api_keys=YOUR_API_KEYS
)
 
# Execute
result = await workflow.execute(input_data)

# Structured logging
import structlog
logger = structlog.get_logger()
 
# Metrics
from prometheus_client import Counter, Histogram
request_counter = Counter('workflow_requests', 'Total requests')
duration_histogram = Histogram('workflow_duration', 'Request duration')
 
# Tracing (OpenTelemetry)
from opentelemetry import trace
tracer = trace.get_tracer(__name__)

# Use cached docs
/agent ai-orchestrator "Your workflow" --use-cached-docs
 
# Specify docs manually
/agent ai-orchestrator "Your workflow" --docs-path ./framework-docs/

# Validate before generating
/agent ai-orchestrator "Your workflow" --validate-only
 
# Regenerate with fixes
/agent ai-orchestrator "Your workflow" --fix-errors

# Review comparison first
/agent ai-orchestrator "Your workflow" --all --compare-only
 
# Then generate chosen framework
/agent ai-orchestrator "Your workflow" --framework {chosen}