[Phase 1.6] Basic OpenAI SDK Integration (Days 9-11)

[mfittko]

⚡ Basic OpenAI SDK Integration (Days 9-11)

Component: Agent Runtime Foundation | Simplified Scope: Basic single-agent execution
Parent Issue: #2 | Timeline: Days 9-11 of Phase 1

🎯 Objective

Implement minimal OpenAI Agents SDK integration for single-agent execution - foundation for Phase 2 workflows.

📋 Tasks (Simplified for Phase 1)

6.1 Basic SDK Integration

• 6.1.1 Set up OpenAI Agents SDK dependency

  • Install and configure OpenAI Agents SDK
  • Set up API client with authentication
  • Basic connection testing and validation
  • Error handling for API failures

• 6.1.2 Create simple AgentRunner class

  • Single agent creation from configuration
  • Basic message sending and response handling
  • Simple tool execution integration
  • Basic error handling and logging

• 6.1.3 Implement basic tool integration

  • Register MCP tools with agent
  • Tool execution through SDK
  • Result formatting for CLI output
  • Error propagation to user

6.2 Configuration Integration

• 6.2.1 Agent creation from JSON config

  • Parse agent configuration from schema
  • Convert to SDK format
  • Model and parameter validation
  • Clear error messages for invalid configs

• 6.2.2 Single-agent workflow execution

  • Execute one agent with given input
  • Tool execution during agent run
  • Collect and format results
  • Handle execution errors gracefully

6.3 CLI Integration

• 6.3.1 Integrate with CLI run command

  • Load agent from configuration file
  • Execute single-agent workflow
  • Display results in CLI
  • Proper exit codes for success/failure

• 6.3.2 Basic testing and validation

  • Unit tests for agent creation
  • Integration test with real OpenAI API
  • Example single-agent configuration
  • Error scenario testing

✅ Definition of Done (Simplified)

• Single agent can be created from JSON configuration
• Agent executes with basic tool access
• CLI agents-cli run --config single-agent.json works
• Results displayed properly to user
• Error handling provides clear feedback
• Basic integration test passes with real API

🔗 Dependencies

• Requires: Security Framework ([Phase 1.5] Security Layer & API Key Handling (Days 8-11) #7) and CLI ([Phase 1.4] CLI Framework & Interface (Days 4-7) #6)
• Enables: Basic Phase 1 functionality

🚫 NOT in Phase 1 (Moved to Phase 2)

• ❌ Multi-agent workflows and handoffs
• ❌ Complex workflow orchestration engine
• ❌ Agent registry and lifecycle management
• ❌ Advanced tool integration patterns
• ❌ Workflow state tracking and persistence
• ❌ Result aggregation from multiple agents

📚 References

• OpenAI Agents SDK Documentation
• AGENTS.md Knowledge Base

Note: This provides foundation for Phase 2 advanced workflows while keeping Phase 1 achievable.