Talk Python To Me #507: Agentic AI Workflows with LangGraph
152 snips
Jun 2, 2025 Sydney Runkle, an open-source developer at LangChain, dives into the world of agentic AI workflows using LangGraph. She explains how integrating agentic frameworks can supercharge Python applications with large language models (LLMs). The discussion highlights the balance of AI-driven workflows, transparency in software development, and the importance of context and memory in enhancing user interactions. Runkle also shares insights on developing intelligent agents and managing application interrupts, showcasing practical examples that bridge creativity with responsibility in AI.
AI Snips
Chapters
Books
Transcript
Episode notes
Enable Real-Time Tool Use
- Allow LLM agents to access real-time data and tools to overcome training data limitations.
- Enable your LLM to decide which tools to use for dynamic, relevant outcomes.
Live Python Coding by LLM
- Perplexity AI wrote Python code live to answer a historical question accurately.
- This shows agentic LLM workflows can combine code and web queries effectively.
Temperature Controls LLM Behavior
- Lower temperature values make LLM outputs more deterministic and reliable for production.
- Higher temperatures are best suited for creativity and exploratory tasks.
