# GitHub Pages deployment

The report is ready for branch-based publishing from `docs/`. It has not been pushed or deployed by this work.

## Activation

1. Review the local commits and push them to `main` when ready. No automatic push was performed.
2. On GitHub, open **MEK-0/Turbofan-RUL-Prediction-LSTM → Settings → Pages**.
3. Under **Build and deployment**, choose **Deploy from a branch**.
4. Select branch **main**, folder **/docs**, then **Save**.
5. Wait for GitHub's Pages deployment to succeed; check the Pages/Actions status before describing the site as live.

Expected URL:

<https://mek-0.github.io/Turbofan-RUL-Prediction-LSTM/>

No custom domain, workflow, framework build, package manager or server-side runtime is required. `docs/.nojekyll` keeps publication a plain static-file deployment.

## Checks after deployment

- Open the expected project URL, including its repository subpath and trailing slash.
- Switch English → Deutsch → English; check headings, captions, plots and table labels.
- Follow contents anchors and evidence downloads; inspect the page on a narrow viewport.
- Confirm both historical NASA figures and all generated evidence plots load.
- Verify the complete English content remains readable with JavaScript disabled.

All local assets and report documents use relative URLs (`./assets/…`, `./evidence/…`); no asset assumes the domain root. External repository and NASA references necessarily use absolute HTTPS URLs. The local curl example points to a separately running API and is documentation only.

## Local preview

From the repository root:

```sh
python -m http.server 8001 --directory docs
```

Open <http://127.0.0.1:8001/>. To simulate the exact Pages prefix, serve a temporary parent directory containing a `Turbofan-RUL-Prediction-LSTM` symlink to this checkout's `docs` directory, then visit `/Turbofan-RUL-Prediction-LSTM/`. Static checks are available through `python scripts/check_report.py` and `node scripts/check_i18n.cjs`.

## Scope

Publishing does not start FastAPI or expose Python/model artifacts. There is no live prediction form, API request, analytics script or remote runtime dependency on the report page. Model inference is described in the README and requires separate hosting plus a compatible scaler.

English is rendered directly in HTML. German translations and alternative plot files are selected by `assets/lang-data.js` and `script.js`. Source titles and historical diagram labels remain in their original language.