AIML 901

Recitation 5

Important – No Core Content
This recitation contains no core content. Everything in this document is exploratory only and will not be directly tested on the final exam.
That said, the patterns you see here (webhooks, front-ends, logging, and verifiable workflows) are highly relevant for both:

  • your final exam, and
  • your individual project.

1. Goals for this recitation

By the end of this session, you should be able to:

  • Explain what a webhook is and how webhooks work in n8n.
  • Describe how to connect n8n to a front-end tool such as Lovable (or another app that can send HTTP requests).
  • Understand the architecture of an end-to-end interview prep product:
    • Front-end form (company, role, job description, resume)
    • n8n workflow (webhooks, AI agent, memory, logging in Sheets)
    • Optional analysis workflow at the end.
  • Work through final-exam-style problems that require you to:
    • Inspect JSON inputs/outputs
    • Reason about node behavior
    • Identify verifiable answers that depend on running the workflow.

Remember: today is about design patterns and mental models, not about memorizing syntax.

2. High-level overview: the interview prep product

Our running example is an interview coach bot:

  1. A user fills in a form with:
    • Interviewee name
    • Company name
    • Job title
    • Job description
    • (Optional) resume upload
  2. When they submit the form:
    • The front end sends the data to n8n via a webhook.
    • n8n:
      • Creates a session_id for this user and interview.
      • Extracts text from the resume if present (or notes “No resume included”).
      • Calls an AI Agent that plays the role of the interviewer.
      • Logs the generated question into a Google Sheet for future reference.
      • Returns the first interview question to the front end.
  3. The user answers each question:
    • Their response is sent back to n8n through a second webhook.
    • n8n logs each answer and generates the next question, forming a chat-like conversation.
  4. At the end, the user can click “Analyze conversation”:
    • n8n pulls the full conversation from the sheet.
    • Builds a combined transcript.
    • Sends that to an Analysis Agent, which returns feedback on their performance.

This pattern is a template for many products:

  • Intake → Orchestration → Logging → Feedback.

3. What is a webhook (in n8n)?

A webhook is a URL that another service can call to trigger your workflow.

In n8n, a Webhook node:

  • Has a method (GET, POST, etc.).
  • Has a path (e.g., "39706c09-...").

  • Exposes a derived URL like:

    ```text https:///webhook/