Day 9: LangChain & LangGraph Deep Dive

🎯 Cieľ dňa

Ovládnuť LangGraph primitívy — Nodes, Edges, State, Conditional Edges. Pochopiť prečo LangGraph nahradil LangChain agents.

core practice

Step-by-step: nodes, edges, states, conditional transitions od základov.

Deep dive do LangChain s real-world use cases. Updated pre 2026.

Graph-based štruktúry, persistent state management, conditional workflows.

Advanced: planning tools, sub-agent delegation, persistent memory — architektúra za Claude Code.

Chains vs Agents — Chains = deterministické sekvencie. Agents = autonómni decision-makeri s tool selection.

LangGraph Primitívy — Nodes (funkcie), Edges (spojenia), State (dáta cez graf), Conditional Edges (branching)

State Management — Core diferenciátor LangGraphu — stav naprieč krokmi, loops, branches, human-in-the-loop

Deep Agents — Planning tool (todo list), filesystem backend, sub-agent spawning — produkčný pattern

Buildni research assistant s LangGraph.

Vytvor graph workflow: Query Analyzer → Web Searcher → Summarizer → Fact Checker
Použi conditional edges na routing podľa query typu
Pridaj human-in-the-loop checkpoint pred finálnou odpoveďou
Testuj s rôznymi typmi otázok (fact, opinion, comparison)
Vizualizuj graph flow