SVI: A New CLI Tool to Streamline Prompt Engineering for AI-Assisted Coding

I wanted a better way to code with ChatGPT, so I built SVI

Developer Aleksandr Razinkin created SVI (Structured Vibe Coding) to eliminate the repetitive task of manually building prompts for ChatGPT. The tool automatically assembles complete prompts from reusable Markdown files, instructions, and project configurations.

Why This Matters

Engineers using AI coding assistants often spend more time preparing context-rich prompts than actually solving problems. This friction undermines the core promise of AI: accelerating development. SVI treats prompt engineering as a structured, composable artifact rather than an ad-hoc ritual—turning what was a manual bottleneck into a repeatable workflow that can be shared across teams.

Key Insights

• Prompt preparation overhead: Developers manually gather context per new file, leading to forgotten dependencies and inconsistent instructions (Razinkin, 2026).
• Composable prompt modules: SVI uses .svi.md files that combine instructions, code dependencies, reusable modules, and project config into one final prompt.
• Dual-mode workflow: Supports both clipboard-based copy/paste for chat UIs (ChatGPT/Claude/Gemini) and automated API calls via LLM.js providers including OpenAI, Anthropic, Gemini, OpenRouter, and Ollama (SVI GitHub repo).

Practical Applications

• • Use case: A developer starting a new module runs svi run -c to get a complete prompt on their clipboard—no more copying files manually or forgetting import paths.
• • Use case: Teams can define project-wide .svi.md templates ensuring all generated code follows consistent architecture patterns without repeating boilerplate in each chat session.

References:

• https://github.com/avrmsoft/svi