📌 Stash

FastAPI · AI Summarization · Azure

Save it for later — and actually get the gist.

Stash tackles the “save for later but never revisit” problem. Paste a YouTube URL; it extracts the transcript, summarizes it with an LLM, and keeps everything organized — wrapped in clean DI, auth, metrics, and production DevOps.

Your stash
try:

↑ A scripted illustration — saving a video extracts a transcript and generates an AI summary from sample data.

01 — How it works

URL in, summary out.

🔗

Save a URL

Drop in any YouTube link and Stash persists it to your library.

📝

Extract transcript

The backend fetches and normalizes the video's transcript.

🤖

Summarize

An LLM (Groq's Llama) condenses it into a quick, skimmable summary.

02 — Engineering

Clean architecture, production-ready.

🧱

SOLID + DI

A ServiceContainer wires services; Protocols invert dependencies for testability.

🔐

Auth

JWT creation and password hashing; per-user saved videos.

❤️

Observability

Health and Prometheus metrics endpoints.

🧪

Tested

Unit + integration tests with coverage gates.

🐳

Containerized

Dockerized image, CI/CD via GitHub Actions, secret management.

☁️

Azure

Deployed to the cloud with auto-deploys from main.

03 — Stack

Built with.

⚡ FastAPI (Python 3.12) 🦙 Groq · Llama 🗄 SQLite 🔐 JWT auth 🐳 Docker 🔁 GitHub Actions 📊 Prometheus ☁️ Azure