AI Feature

The boilerplate includes an AI feature module with a streaming OpenAI integration. Use it as a ready foundation for shipping your own AI workflow without designing the request, response, and browser streaming loop from scratch.

What the `tools` app contains

<project_slug>/<project_slug>/tools/
├── apis.py
├── gpt.py
├── static/js/tools.js
├── templates/tools/cover_letter.html
├── urls.py
└── views.py

The key files are:

views.py: serves the included cover letter page
apis.py: receives front-end requests and returns a streaming response
gpt.py: centralizes the OpenAI client call
static/js/tools.js: holds the front-end JavaScript for the default AI flow
templates/tools/cover_letter.html: the default UI and access control

Enable OpenAI locally

Set your API key in the .env file, then reload the environment:

export OPENAI_API_KEY=your_openai_api_key
source .env

For Heroku production:

heroku config:set OPENAI_API_KEY=your_api_key

How the default flow works

The default implementation is a cover letter generator:

the page lives at /tools/cover-letter/
the front end posts user input to /tools/api/generate/
the API view builds a prompt
the OpenAI client streams text back to the browser
the UI renders the response progressively

That streaming flow is the reusable part. Replace the prompt, model choice, UI, and post-processing with whatever fits your own product.

Access control in the default implementation

The default template gates the feature behind request.user.is_paid. That gives you a clear starting point for tying AI access to billing.

If you want to test the page locally without paid access, update the condition in templates/tools/cover_letter.html:

{% if request.user.is_authenticated and request.user.is_paid %}

You can loosen or replace that guard however you want once you start building your real product rules.

What to customize first

Most teams should change these parts first:

the page copy and UX in cover_letter.html
the request payload shape in tools.js
the prompt construction in apis.py
the model call in gpt.py
the route names in urls.py