LangGraph review workflow

This example keeps the state visible as the graph drafts an answer with a model call, routes it through a conditional edge, and sends it to review when the request mentions a payment.

Requirements

• Python 3.10+
• An OpenAI API key set as an environment variable (OPENAI_API_KEY)

Steps for running this example in the shell

1. Install dependencies:

python -m venv .venv

# macOS/Linux:
source .venv/bin/activate

# Windows Command Prompt:
.venv\Scripts\activate.bat

# Windows PowerShell:
.venv\Scripts\Activate.ps1

pip install -r requirements.txt

2. Export your API key as an environment variable:

export OPENAI_API_KEY="sk-..." # Windows cmd: set OPENAI_API_KEY="sk-..." # Windows PowerShell: $env:OPENAI_API_KEY="sk-..."

3. Run the script:

python review_workflow.py

What it demonstrates

• Explicit input, draft, and needs_review state fields
• A drafting node that calls the model and writes its reply into the state
• A conditional edge that routes to human review or to the end of the graph
• Intermediate state preserved in the graph output

Output

When you run the script, it will print the final state of two runs. The first request ends after the drafting node, and the second one mentions a payment, so it also passes through the review node and its draft is prefixed with Reviewed:.