Please enable JavaScript.
Coggle requires JavaScript to display documents.
L5 - LangGraph Core Components - Coggle Diagram
L5 - LangGraph Core Components
What is LangGraph
Orchestration framework for building stateful & multistep LLM applications
Represents base logic into graphs of nodes (tasks) and edges (routing)
Features - Parallelism, Loops, Branching, Memory, Resumability
Workflows
Series of tasks to achieve goal, can be linear, branched, parallel, looped
Complex includes - retries, multiagent communication, tool augmented reasoning
Common Workflows
Prompt Chaining
Routing
Parallelization
Orchestrator Workers
Evaluator Optimizer
Graph
Nodes ( Single task/Python Function)
Edges (Sequential/Parallel/Conditional/Loop)
State
Shared memory which flows through workflow
holds datapoint in key-value pair
Shared across all nodes
Mutable
Evolves as move down the graph
Reducers
defines how updates from nodes are applied to shared states
Each key in the state can have its own reducer
Replace, Add, Merge
Execution Model
Define state (TypedDict Class)
Define Graph from StateGraph
Add Nodes & Edges
Compile
Invocation
SuperStep - one round of execution
Execution stops when no active node,
no messages passed through edges