docker

Author: solusops

README.md

@@ -98,6 +98,26 @@ Electron also generates a new random local-access token on every launch. The ren
 
 No cloud inference API key is required. The first synthesis can be slower while a model loads or prepares voice conditioning; later runs can reuse the warm model and cached conditioning.
 
+### Docker
+
+Docker packages the browser dev setup (frontend + backend, no Electron) into one container for local use. It is not a hosted deployment -- nothing publishes this container to a live URL.
+
+**Build and run locally** with Docker Compose, from the repository root:
+
+```powershell
+docker compose up
+```
+
+This builds the image from the root `Dockerfile`, starts both the Vite frontend (`5173`) and the FastAPI backend (`8765`), and persists `~/.voice-lab` inside the container to the named `voice-lab-data` volume so downloaded models and voice profiles survive a restart. `docker-compose.yml` also loads `backend/.env` and sets `ALLOWED_ORIGINS` to the localhost dev origins by default -- edit that `environment:` block if you're reaching the container from another machine on your LAN.
+
+**Or pull the prebuilt image** published by the `Docker` GitHub Actions workflow on every push to this branch:
+
+```powershell
+docker run -p 5173:5173 -p 8765:8765 ghcr.io/solusops/studybuddy:latest
+```
+
+Either way, the container does not bundle model assets -- run `npm run setup:voice-lab` (see above) once, either on the host before mounting `~/.voice-lab` in, or inside a shell in the running container, before using providers whose assets aren't installed yet.
+
 ## Shipped model assets
 
 `npm run setup:voice-lab` is the supported TTS download command. It uses the repositories and exact revisions below rather than floating branches.