Advanced Agentic Reasoning
Explore best practices in agentic reasoning, covering planning, reflection, and multi-step logic within AI agents. Build a deep research agent using open-source frameworks like LangGraph and use modern evaluation tools to measure performance.
What you’ll learn:
- Agentic reasoning best practices: planning, tool-calling, and self-reflection.
- Understand planning agents vs. reflection agents and when to use each.
- Build a deep research agent using LangGraph.
- Evaluate complex agentic workflows with LangSmith and academic metrics.
🛠️ Project suggestion:
Create a LangGraph deep-research agent that:
• Plans a sequence of steps to answer a complex query
• Reflects and refines its plan after each iteration
• Integrates external tools (search, code execution)
• Logs intermediate reasoning with LangSmith and analyze using RAGAS metrics
🔗 Recommended Resources LangChain/LangGraph: Build Reflection‑Enabled Agentic – a hands-on Medium guide that shows how to implement reflection loops in agentic workflows https://www.newsletter.swirlai.com/p/building-ai-agents-from-scratch-part-8ca?utm_source=chatgpt.com
Planning for Agents – LangChain blog explaining planning techniques and structured reasoning in LLM agents https://blog.langchain.com/planning-for-agents/
LLM Agents – Prompt Engineering Guide – comprehensive overview of agent architectures including planning, memory, and tool use https://www.promptingguide.ai/research/llm-agents?utm_source=chatgpt.com