AI Prompt Generator & Optimizer
Intelligent prompt engineering middleware transforming casual inputs into structurally optimized AI instructions.
// Prompt Generator: Dual-Mode Pipeline
export async function optimizePrompt(input: string, mode: 'fast' | 'advanced') {
const systemDirective = mode === 'advanced'
? 'Deconstruct input into: Persona, Context, Detailed Constraints, Output Format.'
: 'Synthesize input into a single concise, punchy high-yield prompt under 500 chars.';
const response = await openRouterClient.post('/chat/completions', {
model: 'deepseek/deepseek-r1-distill-llama-70b',
messages: [
{ role: 'system', content: systemDirective },
{ role: 'user', content: input }
]
});
return response.data.choices[0].message.content;
}Engineered with uncompromising integrity.
Prompt Generator acts as an intelligent abstraction layer between casual users and cutting-edge LLMs. Featuring dual generation modes (Fast Mode for concise prompts, Advanced Mode for multi-layered reasoning prompts with constraints), it offers real-time character telemetry, one-click clipboard copying, keyboard shortcuts, and dark glassmorphic styling.
System Specifications
- Live deployed on GitHub Pages with zero server hosting costs
- Dual-mode prompt generation tailored for standard and reasoning models
- Instant clipboard integration and Ctrl+Enter keyboard workflow
- Decoupled, modular React component architecture
Core Capabilities
Dual Optimization Modalities
Switch between Fast Mode (<500 chars for rapid Q&A) and Advanced Mode (500–1000 chars with role definitions, edge case constraints, and strict output schemas).
Real-Time Telemetry & Token Guard
Live character counting and length boundary validation to prevent runaway token expenditure.
DeepSeek R1 & LLM Integration
Direct OpenRouter API integration with robust try/catch exception handling and user-friendly toast notifications.
Automated GitHub Actions CI/CD
Continuous integration pipeline running Vite builds, environment secret injection, and automatic GitHub Pages deployment on push.