LangGraph Sidekick

Overview

LangGraph Sidekick is the week-4 Gradio co-worker from 4_langgraph. A Worker node calls tools until it answers or asks a question; an Evaluator checks success criteria and either ends the turn or sends the worker back with feedback.

System design

• Worker / evaluator loop. Conditional edges keep refining until criteria are met or user input is needed.
• Tool node in the graph. Playwright browse, search, Wikipedia, Python REPL, sandbox files, and Pushover sit behind a LangGraph ToolNode.
• Checkpointed thread. MemorySaver holds conversation state per Sidekick instance id; Gradio supplies the request and success criteria.

Product path

What a person can do: submit a request and criteria, reset the co-worker, read reply plus evaluator feedback. Diagram key: ✦ AI agent · purple = actor · white = service.

---
config:
  theme: default
  themeVariables:
    background: "#ffffff"
    fontSize: "14px"
---
flowchart TB
    person[Person]
    ui[Gradio Sidekick]
    worker["✦ Worker"]
    tools[ToolNode]
    evaluator["✦ Evaluator"]
    memory[MemorySaver]

    person -->|request and success criteria| ui
    person -->|reset co-worker| ui
    ui --> worker
    worker -->|tool calls| tools --> worker
    worker -->|text reply| evaluator
    evaluator -->|retry| worker
    evaluator -->|done or clarify| ui
    ui --> person
    worker -.-> memory
    evaluator -.-> memory

    classDef agent fill:#e8f0ee,stroke:#0f3d36,stroke-width:2px,color:#0d0d0d
    classDef service fill:#fafaf8,stroke:#c8c8c2,stroke-width:1px,color:#0d0d0d
    classDef actor fill:#f3eef8,stroke:#6b4f9a,stroke-width:2px,color:#0d0d0d
    class worker,evaluator agent
    class ui,tools,memory service
    class person actor

Layers live on Mental Model; role detail on Worker and Evaluator.