Proposal - AI API Evaluation Framework -MVP

[Aman071106]

PR Description

This PR addresses Issue #618 by introducing a Functional Prototype (MVP) for the AI API Evaluation Framework.

It establishes the foundational UI and logic to allow users to benchmark AI models (e.g., GPT-4 vs Gemini). The implementation includes a dedicated Evaluations page, a comprehensive creation wizard, and a results visualization engine.

Architecture Choice: Pure Dart Implementation

We intentionally chose a Pure Dart implementation to maintain the project's single-binary portability and architecture consistency.
Reason:

• No External Dependencies: Users do not need to install Python or manage virtual environments. The app remains "Download and Run".
• Network Native: Dart is optimized for asynchronous I/O, making it perfect for handling concurrent API evaluation of multiple models.
• Consistency: The entire codebase remains in a single language, simplifying maintenance and contribution.

Data Flow

1. Input: User configures task (Type, Scoring, Batch) via the EvaluationsPage Dialog.
2. State Management: EvaluationsNotifier (Riverpod) manages the queue of evaluation tasks.
3. Execution (Mock/Future Real):
   • Current: Simulates network delays and random scores.
   • Future: Will utilize Dart's http package to call the respective AI APIs (OpenAI, Gemini, etc.) directly and compute metrics (e.g., Levenshtein distance, Exact Match) in-memory.
4. Visualization: The UI listens to state changes and renders the results implementation.

Changes

• lib/models/evaluation_model.dart: Added data models for EvaluationModel and EvaluationResult.
• lib/providers/evaluation_providers.dart: Implemented StateNotifier for managing the evaluation lifecycle.
• lib/screens/evaluations/evaluations_page.dart:
  • Implemented the List View for evaluations.
  • Added the Create Dialog with "Batch Size", "Task Type", and "Scoring" configurations.
  • Integrated a Dual-View Results Card (Table + Charts).
• lib/screens/dashboard.dart: Added "Evaluations" to the navigation rail.

Related Issues

• AI API Eval Framework #618

Screenshots

Evaluations Tab	Create Dialog

Results Table	Results Charts

Checklist

• I have gone through the contributing guide
• I have updated my branch and synced it with the project main branch before making this PR
• I am using the latest Flutter stable branch
• I have verified the fix manually (UI flows and Mock execution)

Added/updated tests?

• No, and this is why: This is a UI/UX Prototype. Unit tests will be added when the real http execution logic is implemented in the next phase.

OS on which you have developed and tested the feature?

• Windows