ASU Academic Studio — Slide Redesign Tool

Demo-grade slide redesign pipeline. Upload any .pptx, parse + classify each slide, review/regenerate/accept-reject in the UI, export a fresh .pptx styled under one of four ASU program themes (Studio, Law, Hugh Downs Communications, Psychology).

Status: prototype. No pilot, no auth, no persistence. Local-only.

Layout

• backend/ — FastAPI service (parse → classify → render pipeline).
• frontend/ — Static React mockup of the studio UI (Babel-in-browser).

Backend quickstart

cd backend
python3 -m venv .venv
.venv/bin/pip install -r requirements.txt pytest httpx
.venv/bin/python -m pytest               # 7 tests
.venv/bin/uvicorn app.main:app --reload  # http://127.0.0.1:8000

Endpoints

Method	Path	Purpose
POST	/jobs	multipart .pptx upload → {job_id}
GET	/jobs/{id}	job + parsed slide list
POST	/jobs/{id}/slides/{i}/regenerate	bump variant seed
PATCH	/jobs/{id}/slides/{i}	{status: accepted|rejected|pending}
POST	/jobs/{id}/export	{program_id, accepted_only} → rendered .pptx
GET	/jobs/{id}/assets/{image_id}	extracted image

Pipeline

1. app/pipeline/parse.py — python-pptx walks each slide, extracts title / bullets / first picture / speaker notes. Charts and tables flagged unsupported (preserved, not rebuilt).
2. app/pipeline/classify.py — heuristic-only mapping into title | bullets | bullets_image | section | quote | unsupported.
3. app/pipeline/render.py — programmatically builds a fresh deck per the chosen ProgramTheme (top maroon bar, gold accent, themed eyebrow + body).

Demo cuts

No charts/SmartArt fidelity, no LLM rewriting, no async queue, no PDF, no font embedding, no real ASU brand masters (programmatic approximation only).

Frontend

Static frontend/index.html loads app-minimal.jsx via Babel-in-browser. Today it runs against a hardcoded SAMPLE_DECK. Wiring it to the live backend is the obvious next step.

License

Internal prototype. No license attached.