Problem
What was needed?
Adding questions to Dilci meant manually splitting text and entering options, answers, and ordering one by one.
Case study
Question Pipeline Panel
A desktop content automation that converts raw exam text into structured questions with AI and publishes them to Dilci in one click.
Platforms
Desktop
Role
Design and development.
Tech stack
Electron · React · Gemini
Status
Launch-ready
Solution
How was it solved?
Pasted raw text is converted by Gemini into a strict JSON schema (question, options A-E, correct answer); a repair layer fixes broken or truncated model output, and a draft preview gates publishing. Approved questions are written to Firestore in a single atomic batch.
Outcomes
What came out of it?
Manual question entry reduced to minutes
A JSON-repair layer that auto-fixes malformed AI output
Word sets and Pro user management from the same panel
Build note
Build narrative
The panel runs a React interface inside an Electron shell with a local Express bridge. The content workflow is paste, convert with a chosen model, adjust in draft, publish in one click — turning the repetitive work of growing Dilci's question bank into a pipeline.