Integrate OpenAI's GPT to explain why option selected is correct or wrong.

[kendricksin]

Users want to know why their selection is incorrect or why another answer is better
It may not be easy to find clear answers addressing the question directly on search

Provide the question, answer, and wrong/correct answer as prompt into Gemini 1.5

import google.generativeai as genai

genai.configure(api_key="YOUR_API_KEY")
model = genai.GenerativeModel("gemini-1.5-flash")
response = model.generate_content("
  Explain why the correct answer and wrong answer to the following question
  Which GCP Service provides distributed SQL storage and database?
  WRONG: Cloud SQL
  CORRECT: Cloud Spanner
  ")
print(response.text)

Copy pasting the question and answer into Gemini
Gemini Flash 1.5 is free and it has at least a decent knowledge of GCP, if we can bring Gemini to users that would be great, perhaps in future we could use Google SSO and let users pay for more expensive accurate models.

[danieldanielecki]

Thank you for the idea, sounds fantastic!

Is it something you'd be interested to pick up and create a Pull Request (PR)?

What kind of improvement are you finding over what we have with Ollama3?
The PR for that change: #19

How to run it/generate explanations: https://github.com/Ditectrev/Practice-Exams-Platform?tab=readme-ov-file#explanations

[kendricksin]

I will give it ago!

Looking at how it was implemented, it requires locally run ollama, not many people have that infra setup.

The improvement I intend to make would be to simply allow users to copy their Gemini API Key and consume their own resources.

[danieldanielecki]

I will give it ago!

Looking at how it was implemented, it requires locally run ollama, not many people have that infra setup.

The improvement I intend to make would be to simply allow users to copy their Gemini API Key and consume their own resources.

Interesting.

Can't be the improvement done based on current solution to use that resource, or some kind of fallback?

[danieldanielecki]

@kendricksin how's the progress on that?:)

[kendricksin]

sorry i havent got around to start this, maybe next month - btw thank you for helping me pass GCP ACE with ease

[danieldanielecki]

sorry i havent got around to start this, maybe next month - btw thank you for helping me pass GCP ACE with ease

@kendricksin, how's the progress on that?:)

[danieldanielecki]

@kendricksin ??

[danieldanielecki]

The title has changed, based on Discord voting where Community said it'd prefer to use OpenAI's GPT.

[danieldanielecki]

User will need to provide its own key to have them generated for some cheap pricing model.

[danieldanielecki]

For the most expensive pricing tier it'll be running on our API key, and Ollama will still stay on some cheap (separated) pricing model to have the explanations based on LLM running locally.