Prompt Forge Studio: I Built a Prompt Engineering IDE + SDK

Source: Dev.to

Prompt Forge Studio: A Prompt Engineering IDE + PaaS Built on Gemini

If you’ve worked with LLMs, you already know the truth:

Most failures aren’t model problems.
They’re prompt problems.

Prompts are often vague, missing constraints, or poorly structured. In production, that means inconsistent outputs, wasted tokens, and higher latency. So I built Prompt Forge Studio.

Live Demo

Prompt Forge Studio | Master the Art of Prompt Engineering
Transform raw user intent into high‑quality, optimized AI prompts. The professional studio for prompt engineering.

Note: This is the result of my “vibe coding” and prompt‑engineering skills.

What It Is

Prompt Forge Studio is an Advanced Development Environment (ADE) for prompt engineering. Instead of sending raw text to Gemini, the system:

• Analyzes intent
• Injects structure and constraints
• Routes to the optimal model
• Caches deterministic outputs

It also logs execution telemetry. The goal is simple: treat prompts like infrastructure.

Key Features

• Cognitive Depth Control – Adjust how deeply the system expands and structures your prompt.
• AI Prompt Auditor – Critiques your prompt before execution to improve clarity and reduce token waste.
• Automatic Model Routing –
  • Long or reasoning‑heavy prompts → higher‑tier model
  • Simple prompts → fast, low‑latency model
• Exact‑Match Redis Caching – Repeated requests return in < 50 ms without calling the LLM.

V2: Prompt as a Service

The platform now exposes an API endpoint:

POST /api/v1/execute

Request body

{
  "version_id": "UUID",
  "variables": { /* ... */ }
}

The engine validates the request, checks the cache, routes intelligently, executes, and logs performance — all cleanly separated from the web app.

Built With

• Next.js 15
• React 19 + TypeScript
• Supabase (Postgres)
• Upstash Redis
• Google Gemini SDK
• Clerk Auth

We’re moving toward a world where prompts need versioning, performance optimization, and auditability. I’m building tooling for that future.

Would you use a Prompt PaaS in production? Feedback welcome.

What’s Next

• A dedicated CLI (almost done, just polishing stability).
• Beta testing – feel free to check it out when it’s released.