Great let’s set up a ten part test.

First test the registration route and register a mock user we must retain that mock email and password for next tests because next we will test login for login we must retain the token for remainder of tests.

Remainder of tests are hitting remaining routes using our mock user and testing our interviews are creating user profile which is used in our activity ideas generation, then we must verify the users activity priority is calculated. So the interviews ask questions, using those answers AI generates a list of 50 activites that user should do in a 90-day plan. We must be able to include the interview question and answers in the 50 activity AI prompt.

By the time we make our 90-day plan builder requests we must have the user’s interview responses and their activity priority sequence and be able to test as a logged in user.

Once we get this far we are ready to really work! I need a testing platform to test with a logged in user.

We want to seriously upgrade the POST /plan:generate.

Truthfully THIS is what our multi-part test is for is to iterate and improve the 90-day plan generation process.

Currently our AI generated 90-day plan process is too simple.

AI generated 90-day plans are fundamentally just a series of startTs EndTs and activity titles.

We know every 90-day plan will have 90 8 hour sleep blocks so we still need the other 90*16 hours covered. Some plans will have more consistency than others. An excercise oriented plan for example will have lots of consistency in working out with variety in the work-out. A more balanced plan may have a lot of variety. We may need to consider running a series of AI requests to construct the plan perhaps using an orchestrator and then even multiple activity requests. The plan has to be cohesive overall which is why we don’t just run 90 one day planner requests.

I want one AI request to handle the big picture goals that drive subsequent requests that handles the calendar data.

And finally we must test calendar generation with the generated 90-day plan and iterate, improve optimize until the AI response can correctly instantiate a calendar.

Then we will need to involve a human to improve the quality and iterate on the ai prompts.

First setup the testing environment and get us to the point we can post a logged in user’s prioritized activities and interview answers to plan:generate.

Then lets start the plan:generate controller refactoring upgrading.