Switch AI integration from OpenAI to Gemini.

[yesshreyes]

No description provided.

[RajashekarRaju]

I have applied below changes :

Previously, GeminiApiService was an object, which will it hard to mock or replace for testing.
Now it’s provided as a DI singleton and injected into DetailRepository, improving testability.

---

Replaced the KtorClientProvider singleton with proper DI managed instances of HttpClient and Json. This ensures a consistent configuration and allows us to reuse or override these components in other environments.

---

DetailRepository now takes GeminiApiService as a constructor parameter. This decouples the repository from hardcoded services and aligns it with clean architecture best practices.

---

DI definitions have been cleanly separated into type-based modules:

• httpClientModule for Ktor HttpClient and JSON
• apiModule for API services
• repositoryModule for data repositories
• viewModelModule for ViewModels
  This makes the setup modular, easier to reason about, avoids clutter in a single DI file.

[RajashekarRaju]

I'd move data classes CodeExample, Section, Syntax into KotlinTopicDetails itself, this sort of granularity for separation of each data class into own file is is unnecessary here and adds extra overhead in navigating the codebase.

Since these classes are tightly coupled to KotlinTopicDetails and not reused elsewhere, keeping them grouped within the same file improves readability and makes the model easier to maintain as a single unit of domain logic.