Thank you for sending your enquiry! One of our team members will contact you shortly.
Thank you for sending your booking! One of our team members will contact you shortly.
Duration 14 hours
Course Outline
Foundational Review of AutoGen
- Defining agents and group configurations
- Exploring function calling and role chaining mechanics
- Identifying limitations of built-in agents and the necessity for customization
Developing Custom Agents in Python
- Defining agent behavior through user_proxy and AssistantAgent subclasses
- Integrating role-specific logic and decision-making processes
- Building reusable agent modules and mixins for flexibility
Advanced Tool Integration & Routing
- Registering, binding, and invoking external tools
- Implementing conditional routing to direct inputs to specific tools
- Orchestrating multi-step toolchains and composite actions
Strategy for Planning & Context Management
- Designing task decomposers and intermediate planning structures
- Persisting context across chained agent interactions
- Implementing scoped memory to support long-running sessions
Robust Error Handling & Recovery
- Identifying and managing failed or incomplete interactions
- Utilizing agent-triggered retries and fallback logic
- Enhancing observability through logging, debugging, and response validation
Collaboration among Custom Multi-Agent Roles
- Coordinating specialist agents within dynamic groups
- Orchestrating reasoning loops and cooperative workflow structures
- Balancing role separation versus role blending in task assignment
Strategies for Real-World Deployment
- Optimizing performance and cost efficiency through token management and caching
- Integrating AutoGen workflows into web applications or data pipelines
- Ensuring security, observability, and seamless user feedback integration
Requirements
- Strong proficiency in Python programming
- Practical experience in developing LLM-based applications
- Understanding of function calling and multi-agent system architecture
Target Audience
- Senior developers
- Platform engineers
- AI architects
Testimonials (1)
I liked that he constantly provided examples but also offered time for individual work on what he presented.