Working on A.I. Generated Night Sky Viewing Itinerary Generator project

[John Martin]

I’ve been working on a personal project that some of you might be interested in. And I am very interested in feedback from you about improvements or issues that you might encounter.

Background – Before doing an outreach event, I like to create an agenda or itinerary of objects to look for with the group. I used to do this “manually”, but thought that a tool might be useful. I need to learn some new things and this little project felt like a good fit.

This tool takes your provided start date, start time, end date, end time and location and will generate a list of the more popular night sky objects viewable during that time. There is an option to print the results once generated. The data is obtained via interface with the OpenAI ChatGPT engine. Since the information is A.I. generated dynamically with each request, results may differ each time that it is run. Also, the OpenAI project is still evolving, so some of the information may be inaccurate.

For each object, it will list:

• Object name
• Summary of the object
• Approximate location of the object in the night sky
• Best time to view the object.

Give it a try HERE

This is the first version (V 1.0).  I have some challenges with formatting the results that OpenAI returns. Sometimes it returns the results in a table format, which breaks the parsing that I’ve written.

Known issues

• Sometimes OpenAI is wrong. It is still learning.
• Sometimes the data is returned in a table format, which breaks the parsing and will look weird
• The output isn’t very mobile friendly yet – working on that
• The list is typically only 5-10 items and sometimes misses the more interesting objects

Note – this project is hosted on a development server, performance may be an issue while it is in development. I’ll eventually get it out on an industrial server if the project gets any demand. Also, to keep costs down, I am using ChatGPT 4.o-mini instead of 4.0 . This mini version doesn’t do as deep of a dive into calculating the results. I might upgrade to 4.0.

Cheers!

Attachments:

1. 

   Screenshot-2024-10-07-at-4.07.05 PM.png
   

[Josef Chlachula]

John, this is great project!

I tried to generate for the period from October 7, 2024, between 22:15 and 23:45. See attached PDF.

Here is what I found not accurate:

#1 Cannot not be visible: Mars is rising around 23:45
#6 Cannot not be visible: M8 is setting down around 22:15
#7 Cannot not be visible: M42 is rising around midnight
#9 Visible only short time, in different constellation: Ceres 8.97 mag is in constellation Sagittarius, not Virgo and sets at 22:44
#10 Cannot not be visible: Copeland’s Septet is in Leo, not Pegasus and those galaxies are challenging to observe

Attachments:

1. NightSkyItinerary-by-Josef-2024-1007.pdf

[John Martin]

Thank you, Josef! This is very helpful. I wonder if I can direct the API to tell OpenAI only to use certain sources that we trust. I will work on it. thanks again!

[Randy H]

Very interesting John! I wonder if the search could customize a minimum horizon like 5 -10 degrees to refine usable targets.

[John Martin]

Thank you! That is good feedback. I will see if I can incorporate it and maybe even put it in the form as a variable for people to enter.

[Author]

Posts