Here is the enhanced, professional-grade prompt designed to get the best possible architectural and UI/UX output from an LLM.

Enhanced Prompt: Llama 3 Dataset Inspector & Workbench

🎭 Role

Act as a Senior Full-Stack Software Engineer and UI/UX Product Designer with deep expertise in LLM fine-tuning workflows, Data Engineering, and React/Next.js development. You specialize in building developer tooling that balances high-performance data handling with intuitive, clean interfaces.

🌐 Context

We are building [APPLICATION_NAME], a specialized dashboard for Data Scientists and ML Engineers to audit, clean, and refine JSONL datasets intended for training Meta’s Llama 3 models. The goal is to move beyond text editors and provide a robust environment to visualize data structure, estimate token usage, and perform quality control before pushing data into the training pipeline.

🛠️ Task Instruction

Design the technical architecture and UI specifications for the dataset viewer. Your response must include:

1. System Architecture: Define the technology stack (e.g., Next.js, TanStack Table for data grid, Tailwind CSS, Tiktoken for tokenization) that allows for handling large JSONL files (100MB+) without browser crashes.
2. UI Component Design:
   • Dual-View Engine: A toggle mechanism to switch between the raw JSON structure and a "Chat Bubble" view (styled like a standard LLM conversation).
   • Token Metrics: A strategy for calculating Llama 3-specific tokens (Cl100k_base encoding) in real-time.
   • AI Quality Overlay: A design for a "Quality Score" badge (0.0–1.0) with an optional "reasoning" tooltip.
   • Filtering & Search: A robust search bar to filter by semantic keywords or specific JSON keys.
3. Data Export Module: Define the logic for client-side serialization to export selected subsets into CSV or Parquet formats.
4. UX Guidelines: Maintain a "Developer-First" aesthetic—minimalist, high-contrast, and keyboard-shortcut friendly.

⚖️ Constraints & Tone

• Tone: Professional, technical, and pragmatic.
• Clarity: Avoid fluff. Use industry-standard terminology.
• Formatting: Use tables for feature comparisons and code blocks for proposed data schemas or component structures.
• Exclusions: Do not provide generic boilerplate code; focus on high-level patterns, architectural decisions, and specific libraries that solve the problem of large-scale JSONL handling.

📝 Output Format

Please structure your response using the following headings:

• Architectural Overview

• UI/UX Component Specifications

• Implementation Strategy for Large Datasets

• Proposed Data Schema (JSONL)

• Recommended Tech Stack

Placeholders

• [APPLICATION_NAME]: Enter the name of the tool (e.g., "LlamaVault").
• [PRIMARY_TARGET_USER]: Define the primary persona (e.g., "ML Engineers").
• [DATA_LIMIT]: Define the expected dataset size (e.g., "Up to 500k rows").