Lovable Prompt Creator - AI Prompt Generator

## APPLICATION OVERVIEW
The AI Reliability & Optimization Platform is a web application designed for developers to enhance the reliability and efficiency of their LLM (Large Language Model) applications. By providing tools for model evaluation, smart routing, observability, and output verification, this platform aims to ensure that LLM applications are correct, cost-efficient, observable, and self-improving.

## CORE FEATURES
1. **Model Evaluation (CI for LLMs)**: 
   - Utilize a golden dataset with versioning to evaluate model performance. Metrics include exact match, LLM-as-judge scoring, and semantic similarity. Outputs will provide pass/fail results and regression diff reports.

2. **Smart Routing Layer**: 
   - Implement a routing mechanism that selects models based on prompt complexity, historical success rates, along with latency tracking and cost per request. This will include fallback strategies for optimal performance.

3. **Observability + Failure Tracing**: 
   - Log all actions including prompts, intermediate steps, and tool calls. When failures occur, trace the breakdown, categorize the type of failure (hallucination, tool error, bad prompt), and provide insights for improvement.

4. **Auto-Doc Sync**: 
   - Detect changes in functions/APIs and highlight impacted documentation sections. Generate suggested updates to documentation without automatic merging, ensuring that documentation remains accurate.

5. **Output Verification Layer**: 
   - Run outputs through a secondary LLM critique and rule-based validators. The system will return a confidence score and flagged issues, emphasizing nuanced evaluations rather than a simple majority rule.

6. **User Interface**: 
   - A real UI that includes a prompt testing playground, evaluation dashboard, cost/latency charts, and a failure trace viewer to visualize performance metrics and improvements.

## DESIGN SPECIFICATIONS
- **Visual Style**: minimalist - Clean, simple design with plenty of white space, minimal color palette, and focus on typography
- **Color Mode**: Light theme with dark text on light backgrounds
- **Primary Color**: #1978E5 (accent for buttons, links, highlights)
- **Typography**: Use Inter from Google Fonts for headings, Inter for body text and UI elements
- **Border Radius**: 8px (moderately rounded) for buttons, cards, and inputs
- **Layout**: The main layout will consist of a top navigation bar, followed by a hero section, feature overview, and individual modules presented in card format, leading to a call-to-action section and a footer.

## TECHNICAL REQUIREMENTS
- **Framework**: React with TypeScript
- **Styling**: Tailwind CSS
- **UI Components**: shadcn/ui
- **State Management**: React Context API or Zustand (based on project complexity)

## IMPLEMENTATION STEPS
1. **Set Up Project**: Initialize a new React project with TypeScript and install Tailwind CSS and shadcn/ui.
2. **Create Basic Layout**: Set up the main layout structure including the navigation bar, hero section, and footer.
3. **Develop Core Features**:
   - Build the Model Evaluation module with input forms and output displays.
   - Implement the Smart Routing Layer with algorithm logic for model selection.
   - Create the Observability and Failure Tracing logging system.
   - Develop the Auto-Doc Sync feature for documentation updates.
   - Build the Output Verification Layer with validation components.
4. **Design User Interface**: Style components using Tailwind CSS and ensure responsiveness across devices.
5. **Integrate APIs**: Connect to GitHub webhooks and Slack for alerts and notifications.
6. **Testing**: Conduct thorough testing for all features, focusing on usability and performance.
7. **Deployment**: Prepare for deployment on a platform like Vercel or Netlify.

## USER EXPERIENCE
Users will interact with the application through a clean and intuitive interface, starting with the hero section that introduces the platform. They can navigate through core features seamlessly, using the evaluation dashboard to monitor their models' performance. The prompt testing playground allows users to easily test and optimize their LLM applications while the failure trace viewer provides detailed insights into errors, enabling developers to make informed decisions on improvements. The overall experience emphasizes simplicity, efficiency, and actionable insights.