Build Log 1: Getting the Add-on to Actually Run
First working conversion today: a Google Doc exported to PDF and saved to the same Drive folder, via a menu item inside the document. This post is the honest account of what it took to get there.
What we built
Three files: a manifest (appsscript.json), the conversion logic (Code.gs), and a small modal dialog (Dialog.html). The conversion uses Google's own Drive API export endpoint, which means it's Google doing the rendering — no third-party PDF library, no page caps imposed by us. That was a deliberate choice based on the competitor complaints in the market research: silent truncation at 80 pages was a top complaint. Using native export sidesteps that entirely.
What actually broke
OAuth scope mismatch. The manifest lets you declare exactly which permissions the add-on requests. We declared minimal scopes. Apps Script issues OAuth tokens matching those declared scopes exactly — not what the code actually calls, not what the consent screen showed. The first version called DriveApp.getFileById(), which needs full drive scope, but we'd declared only drive.file. Result: PERMISSION_DENIED at runtime, even though the user had clicked through a consent screen that looked correct. Fixed by switching all Drive operations to the REST API via UrlFetchApp and using drive.readonly for reading, drive.file for creating.
Wrong export URL. The obvious export URL (docs.google.com/document/d/{id}/export) returned 404 when called with an Apps Script OAuth token. The correct endpoint is the Drive API v3 export: googleapis.com/drive/v3/files/{id}/export. Same result, right URL.
Drive API not enabled. Apps Script projects get auto-assigned to a GCP project. To call the Drive REST API from UrlFetchApp, that API has to be explicitly enabled in the GCP project. The auto-assigned project wasn't ours to configure, so we created a new GCP project, enabled Drive API there, and linked it to the script. This also required setting up an OAuth consent screen and adding a test user before authorization would go through.
clasp login. The default clasp OAuth client is rate-limited and returns access_denied for new users. Standard issue — worked fine once we authorized through the browser on a second attempt with a different account.
Current state
Docs → PDF works. Saves the PDF to the same Drive folder as the source document. Overwrites an existing PDF with the same name rather than creating duplicates. Error messages in the dialog are plain English, not HTTP codes.
Not built yet: Sheets → PDF, Slides → PDF, PDF → Docs. Scope reduction before Marketplace submission (we're using drive.readonly which is broader than ideal — we'll narrow it before asking Google to review).
Honest note on the process
Most of the time was OAuth and GCP configuration, not code. The actual conversion logic is about 30 lines. The surrounding infrastructure — GCP project, Drive API enable, consent screen, test users, scope declarations — took longer to sort out than the product code. That's a real fixed cost for anyone building on Google's platform and worth knowing upfront.