fix: replace hardcoded model class with dynamic parameter

[FWangZil]

Relates to

issue #1439

Description

• What is the problem?
  In the generateText function, the model type was hard-coded, preventing the function from dynamically switching models based on external parameters.
• What does this PR do?
  1. Removes the hard-coded model class from generateText and generateMessageResponse.
  2. Use the parameter to specify the desired model type.

Risks

• Risk Level: Low
  • The change is confined to the internal logic of the generateText and generateMessageResponse function.

Background

• Previously, generateText and generateMessageResponse did not allow dynamic model switching based on incoming parameters, limiting flexibility and making it cumbersome to use different models in various scenarios.
• By removing the hard-coded approach, developers can easily specify or configure the desired model without modifying core code.

What kind of change is this?

• Bug fix (fixes a reported issue)
• New feature (adds functionality)
• Other (please describe)

Documentation changes needed?

• Yes
• No

Testing

Where should a reviewer start?

• Look at how generateText and generateMessageResponse now handles the incoming model parameter, and verify that any calling code is appropriately updated to use this new parameter.

Detailed testing steps

• Call generateText and generateMessageResponse with a specific model parameter, confirming that it correctly switches to the chosen model.

[monilpat]

LGTM - but please confirm that this works with a screenshot and if so remove hard coding of gpt-4o in generateObject generatedObjectedDEPRECATED etc thanks

[odilitime]

Just ran into something that could really use it

[FWangZil]

I found that I should deal with the judgment and choice of TiktokenModel of different AI providers more. Let me temporarily close this PR, improve it, and reopen it later.