Prompt Forge

# Case Study: App 03 — Prompt Forge ## 1. Overview Prompt Forge is a structured content generation workbench. It allows marketing teams to craft tailored content variations across different personas (e.g. Friendly, Technical, CFO) and formats (LinkedIn, Email, Landing Page). ## 2. Learning & Technical Goals - Implement pre-generation safety filtering with Azure Content Safety API. - Support real-time Server-Sent Events (SSE) streaming with structured JSON parsing. - Eliminate server-side prompt logging vulnerabilities by shifting history storage to client `localStorage`. ## 3. Architecture & Tech Stack - **Frontend:** Next.js 16, Tailwind CSS, TypeScript. - **Backend:** FastAPI, Python 3.12, Azure OpenAI, Azure Content Safety SDK. ## 4. Key Security & Cost Control Decisions - **Content Safety:** Automated inspection for hate speech, self-harm, and violence before initiating LLM calls. - **No Pseudo-Identity:** Removed `X-Client-ID` headers and server log endpoints (`/log`), preventing unauthorized log access. - **Streaming Hardening:** SSE streams pass `Cache-Control: no-store` headers and do not log individual text deltas. ## 5. Outcome Prompt Forge delivers fast, streaming content variations with strict safety bounds and zero prompt exposure on the backend.