Web Navigation

• Key points are in the visual diagram above.

Real web tasks aren't single-page interactions. Research, shopping comparison, and compliance checks require navigating across sites, handling redirects, and maintaining context across pages.

A web navigation agent is a research assistant who clicks through Google results, opens promising links, skims pages, and bookmarks the best sources.

Scenario

Agent task: 'Find the cheapest direct flight NYC→London next Friday.' Navigates to 3 comparison sites, handles cookie banners, extracts prices, compares, returns best option with booking link.

Solution

In Browser & Computer Use Agents, apply Web Navigation to this scenario: Agent task: 'Find the cheapest direct flight NYC→London next Friday. Identify the inputs, run the technique, validate the output, and note one thing you would monitor in production.

Do this before moving to the next module — reading alone is not enough.

Open the Code Walkthrough below and run it locally. Change one parameter related to Web Navigation (e.g. model, temperature, top_k, or tool name), observe the difference in output, and write 2–3 sentences explaining what changed.

Highlighted lines show where Web Navigation happens in the code.

1# Web Navigation — minimal example2from openai import OpenAI3
4client = OpenAI()  # create API client5
6# Ask the model to explain this topic7response = client.chat.completions.create(  # core API call for Web Navigation8    model="gpt-4o-mini",9    messages=[10        {"role": "system", "content": "You explain web navigation clearly."},11        {"role": "user", "content": f"What is web navigation?"},12    ],13    temperature=0,14)15print(response.choices[0].message.content)  # show output for debugging

• Treating Web Navigation as a black box without evaluation
• Ignoring cost and latency in production
• Skipping error handling for web navigation

Quick recap — the most important points from this module.