Connect Chat GPT to Yii2 and output the results through API in Yii2.

PHP, Web Programming — incorrectly specified categories?

Project translated automatically. Log in or register, to view the original

To the specified email, it is necessary to register an account in Chat GPT, save the access credentials appropriately in Redmine. Configure the operation of Chat GPT in such a way that when we call our API (implemented on Yii2), we can securely use the Chat GPT token and receive a response back in our API, which we will use on the front end. Also, for working with the API, we use Swagger, so it is very important that you have experience working with it from the backend side (without the ability to work with the API through Swagger, this work cannot be accepted).

Additionally, in our API, it is necessary to check whether a long string is not coming from the front end. We need to accept 1-2 words from the front end. (Most likely, during testing, we will be able to determine the appropriate string length and adjust it) to be agreed upon during the work process.

We need to use Chat GPT as a dictionary/wikipedia. On the front end, a person will click with the mouse on a specific word or phrase, we will send this to our API, then pass it to Chat GPT, and return the text to the front end.

The essence is to have a secure connection with Chat GPT, rather than connecting directly from the front end to it, as it is easy to seize the token from the front end.

If you have any clarifications - write, I would be very grateful, I will add them to the project. If you have similar experience in configuring chat for such a task - great.

It is very important: we work and best understand the process of time and materials evaluation, so to form an understanding of pricing, it is necessary to know the rate per hour and the estimated time for completing this task.

It will be necessary to interact with other team members. For task management, we use Redmine (understanding it is not mandatory, I will explain everything), for work correspondence Teams. Studios are requested not to disturb. I expect feedback more from middle or something in between middle and senior specialists for this project.

Update #1 from 9 December 2025

The Swagger is already installed and configured in the project, there is no need to install and configure from scratch. There is authorization, there are endpoints for user login, token refresh, and logout. It is also set up to work in 6 languages, 4 of which are already fully functional, while the others are waiting for translation by natives.

Update #2 from 9 December 2025

> On the front end, a person will click on a specific word or phrase with the mouse, we will send this to our API, then pass it to Chat GPT, and return the text to the front end.

Important! This project is only about the backend; this project does not involve any work on the frontend. The frontend is handled by a separate team member. I quoted part of the description to draw attention to it. The mention of the frontend was made solely for the purpose of providing a holistic view of the use of this feature.

Update #3 from 10 December 2025

Chat GPT, prompts: our service is a game for people studying Nutrition.

For example, there are "cards" with text:

Did you know that… magnesium helps enzymes in the liver neutralize toxins and reduces inflammation?
Antioxidants protect the liver from damage when processing toxins?
With an excess of toxins and stress, the liver itself becomes a source of free radicals?

In the text above, the highlighted words:

- magnesium
- enzymeAM
- toxins
- inflammation
- antioxidants
- stressE
- freeRADICALS

These are the words that may be unclear to players who have just started to familiarize themselves with Nutrition, and for which explanations of their meanings in the context of the science of Nutrition are needed, in clear, accessible language.

Above, I highlighted the endings of words in uppercase to show that they are not always in the required form and will not always be in the required form for queries, for example, in Google (perhaps? this will also need to be adjusted additionally in the prompt)

Example:

What is a free radical? - this is one
What is freeRADICALS - this is already different...

Update #4 from 10 December 2025

Caching responses for identical words, to reduce requests to GPT

Initially, there are 3 options envisioned:

Follow-up caching (I personally am not very in favor of this type, as we may run into Linux limitations)
In the database: create an md5 hash from the word, store it in a table with id, md5, and GPT explanation (do not use full-text search, but look for exact matches by md5, to avoid overloading the database with requests)
Attach INDEX and UNIQUE to the words and language tag and then use %LIKE%
At this stage, "semantic caching" is not being considered
Your option

In any case, I am ready to hear your thoughts on the best of the above options, or an alternative that would be better than the ones described above.

Update #5 from 11 December 2025

Data reception and transmission (Request, Responses) occurs in the form of JSON

Proposals 8 Withdrawn 1 Rejected 12 Discussions 11

Viktor Serobaba 4 December 2025

Ось чорновий варіант уже зробив. Залишається додати нюанси.
Помилки:

Запитання:

1. Написи для документації які

2. чи треба перевіряти символи (перевірку слів та довжину шкірного слова додано).

3. Є формат відповіді який треба обов'язково зробити (для відповіді та помилок)?

Якщо вам потрібний новий аккаунт на платформі чату жпт то треба емайл
Ну і для того щоб додати модуль потрібно буде доступ до проекту

Якщо є питання, пишіть.

Viktor Serobaba 4 December 2025

ще:
Я не знаю де саме він буде використовуватися, якщо можна посилання дайте сайту, де буде працювати.

Завтра після 12-го буду вдома і можу доробити все. зараз 1:00 вийшло.

Також у майбутньому для скорочення витрат та безпеки можна ще:
1. додати кешування відповідей у внутрішню базу, або вбудований файловий кеш
2. Додати якийсь елемент безпеки щоб трохи ускладнити просес запитів від сторонніх ботів.

Viktor Serobaba 4 December 2025

ще:
Я не знаю де саме він буде використовуватися, якщо можна посилання дайте сайту, де буде працювати.

Завтра після 12-го буду вдома і можу доробити все. зараз 1:00 вийшло.

Також у майбутньому для скорочення витрат та безпеки можна ще:
1. додати кешування відповідей у внутрішню базу, або вбудований файловий кеш
2. Додати якийсь елемент безпеки щоб трохи ускладнити просес запитів від сторонніх ботів.

Serhii Melnichuk

4 December 2025

> довжину шкірного слова додано
слово "шкірного" не зрозуміло

також на скрінах ви обрізали частину Swagger`у, не видно чи у вашому є авторизація, зазвичай є іконка замочку, бо в нас своя авторизація є