Course Objective & Scope
In real-world agent applications, pure ReAct often loses track over long horizons, while rigid Plan-and-Execute struggles when individual tool calls encounter unexpected outputs. This integrated course teaches how to build hybrid control architectures that plan at a high level, execute steps via scoped ReAct loops and persist progress in an explicit state schema.
Participants code the same business problem first with ReAct, then with Plan-and-Execute, and finally unify both inside a stateful, checkpointed LangGraph workflow. The deliverable is a resumable hybrid agent running on an organisational scenario.
Who Should Attend?
- Software developers who want to master core agent control loops and state management in one consolidated programme
- AI engineers designing complex multi-step assistants
- Technical teams moving agent prototypes into predictable production architectures
- Prerequisite: Python and basic experience with LLM APIs
Key Takeaways
- Compare the practical strengths and failure modes of ReAct and Plan-and-Execute loops
- Combine a high-level planner with step-scoped ReAct executors in a hybrid graph
- Manage plan progress, gathered evidence and tool outputs in a typed state schema
- Control step transitions deterministically using conditional edges and reducers
- Pause and resume multi-step plans from any checkpoint
Curriculum
The ReAct Loop and Short-Horizon Tactical Execution
- The Thought–Action–Observation loop and native tool calling
- Iteration caps, loop detection and where ReAct breaks down on long tasks
Plan-and-Execute and Strategic Task Decomposition
- Generating verifiable step-by-step plans via structured output
- Step execution, intermediate result aggregation and dynamic replanning triggers
Explicit State Schemas and Transitions
- Separating chat history from structured task state (plan, completed steps, artefacts)
- Pydantic/TypedDict state models, reducers and conditional transitions
Hybrid Architecture: ReAct Inside a Planned State Graph
- Plan-and-Execute at the outer graph with scoped ReAct workers at step nodes
- Checkpoint persistence, time-travel debugging and resuming interrupted runs
End-to-End Hybrid Agent Workshop
- Coding an enterprise research-and-action scenario with the hybrid architecture
- Benchmarking token cost, step count and task completion rate
Other Courses in This Category
-
Technical course on delegating work across specialist agents using control handoffs versus the agent-as-tool pattern, covering triage routing, context filtering and conversation ownership.
- Live Course
- 5 modules
-
Course on designing reusable skill packages, progressive context disclosure, bundled scripts/references and dynamic capability loading so agents scale without overflowing their context window.
- Live Course
- 5 modules
-
Course on cross-framework and cross-vendor agent collaboration using the A2A (Agent2Agent) protocol: Agent Cards, task lifecycle, streaming updates, enterprise authentication and MCP–A2A complementarity.
- Live Course
- 5 modules
-
Training on pausing agents for human approval on high-risk tool calls and low-confidence decisions, editing tool arguments, routing asynchronous escalations via Slack/Teams/web UIs and maintaining audit trails.
- Live Course
- 5 modules
-
Hands-on training for developers on building ReAct agents from scratch: the Thought–Action–Observation loop, native tool calling, stop conditions, iteration limits and failure analysis.
- Live Course
- 5 modules
-
Technical course on agents that split multi-step tasks into planner, executor and replanner roles, covering ReWOO and LLMCompiler-style parallel task graphs, plan validation and cost optimisation.
- Live Course
- 5 modules
Get in touch
