Google Sheets
Product and news records are edited in two dedicated spreadsheets using required headers and flexible column aliases.
Live platform
Project record 02
Updated 18 Jul 2026
Corporate website · Content automation & performance
AGVN Digital Platform
A corporate agricultural website that turns editable spreadsheet content into product pages, news articles and locally managed image assets—without requiring a traditional CMS.
01 · Overview
A business website that non-technical editors can keep current.
AGVN needed more than a set of static pages. Product and news information had to remain editable without turning every content change into a manual coding task.
The resulting platform combines a static corporate website with a lightweight publishing engine. Content is maintained in Google Sheets, validated and transformed by a Node.js generator, then committed back to GitHub as deployable HTML and locally managed images.
This architecture keeps the public website fast and inexpensive while giving the business a familiar operational interface for ongoing updates.
02 · Operating model
Separate editable sources from generated outputs.
The system has a clear ownership model: humans edit structured data and templates; automation manages the repetitive output.
Reusable HTML templates
Product and news templates define the design, SEO structure and shared layout of generated pages.
Node.js generator
The generator fetches rows, validates fields, maps content, resolves slugs and writes output into product and news directories.
Image normalization
Google Drive and remote image URLs can be downloaded, verified and stored under assets/generated-images.
GitHub Actions
An hourly schedule and manual trigger run the generator, then commit only generated pages and image assets when changes exist.
Static public website
The deployed result remains standard HTML, CSS, JavaScript and optimized assets—fast to serve and easy to inspect.
03 · Publishing architecture
Multiple fallbacks protect a simple editing experience.
The system assumes spreadsheet names and image inputs may change, so discovery, validation and retry logic sit between content editing and deployment.
Discover sheets
Prefer configured gid, match known tab names, inspect discovered tabs, then fall back to the first tab.
Fetch with fallback
Try OpenSheet and Google GViz CSV sources with timeout and retry handling.
Resilient accessValidate headers
Confirm required fields such as product name, slug and news title before generating output.
Safer generationNormalize images
Resolve Google Drive IDs, verify image responses, create deterministic filenames and update the manifest.
Stable assetsGenerate pages
Merge row data into reusable product and news templates and write static HTML outputs.
Consistent pagesCommit changes
Stage generated directories and assets; skip the commit entirely when no output has changed.
Clean history
Google Sheets→OpenSheet / GViz→Header validation→generate.js→HTML + images→GitHub Pages
04 · Key decisions
Automation succeeds when the source and output boundaries are explicit.
DR-01Accepted
Use Google Sheets instead of a traditional CMS
The business needed an editor-friendly workflow without introducing backend hosting, CMS maintenance or account management.
- Why
- Familiar editing, fast adoption and low operating overhead.
- Trade-off
- Headers and field discipline become part of the system contract.
DR-02Accepted
Generate static pages instead of rendering spreadsheet data in the browser
Product pages need stable URLs, crawlable content and predictable output even when the data source is temporarily unavailable.
- Why
- Stronger SEO, faster delivery and independence from runtime sheet availability.
- Trade-off
- Content changes wait for the next workflow run.
DR-03Expanded
Auto-detect spreadsheet tabs and validate headers
Hard-coded tab names made the sync fragile when sheets were renamed, reordered or expanded.
- Why
- Support
gid, preferred names, discovered tabs and required-field checks. - Trade-off
- The generator becomes more complex, but failures become more diagnosable.
DR-04Accepted
Download remote images into the repository
Editors may paste Google Drive links or other remote sources that are unsuitable for direct production use.
- Why
- Local assets create stable paths and remove dependency on fragile sharing URLs.
- Trade-off
- The repository stores generated binary assets and needs a manifest.
DR-05Accepted
Convert GitHub raw links to jsDelivr where appropriate
Direct raw asset delivery can create rate limits or inconsistent performance for a public website.
- Why
- CDN delivery improves caching and public asset reliability.
- Trade-off
- Cache invalidation and version references need attention.
DR-06Refined
Optimize mobile without removing core content
The first performance passes revealed heavy video, animation and external asset costs on mobile.
- Why
- Use lighter video behavior, non-blocking assets and desktop-only enhancement where appropriate.
- Lesson
- Performance fixes must preserve banners, product images, sheet sync and generated content.
05 · Performance and reliability
The fastest page is not useful if automation silently breaks it.
AGVN required two forms of reliability: the public website had to remain responsive, and the publishing workflow had to fail visibly instead of generating incorrect pages.
Non-blocking external CSS
Fonts and icon styles were shifted away from the critical rendering path.
Desktop-only enhancement
AOS behavior was limited and paired with a failsafe so content remains visible if scripts fail.
Lite video strategy
Heavy YouTube behavior was reduced on mobile while preserving the customer-proof section.
WebP, CDN and local generation
Asset mapping and generated image storage reduce broken or slow external dependencies.
Header validation before generation
The generator refuses invalid sources instead of overwriting working output with incomplete pages.
No-change commits are skipped
Git history remains cleaner and scheduled runs do not create meaningless revisions.
06 · Lessons learned
A lightweight CMS still needs contracts, validation and ownership.
Edit sources, not generated files
Templates and spreadsheet rows are durable inputs. Generated product and news pages can be overwritten by design.
Flexible input must end in strict validation
Supporting many tab names and aliases is useful only when required fields are verified before output.
Image handling belongs in the publishing pipeline
Expecting every editor to provide production-ready URLs creates recurring breakage and support work.
Automation should preserve the last good state
Empty or invalid sheet data should not erase working pages during a scheduled run.
Performance changes must respect business content
Removing heavy code is useful; accidentally hiding proof, product content or navigation is not.
Static architecture can support real operations
With a disciplined generator and workflow, a static site can behave like a maintainable publishing product.
07 · Next priorities
Strengthen observability and content governance.
The core publishing engine works. The next improvements should make issues easier to detect and content quality easier to control.
Add a generation summary report
Record selected tabs, row counts, generated pages, skipped records, downloaded images and warnings.
Formalize spreadsheet schema documentation
Publish required headers, accepted aliases and example rows for products and news.
Add generated-output integrity checks
Verify canonical URLs, image paths, required metadata and broken internal links before committing.
Improve asset lifecycle management
Detect orphaned generated images and apply consistent size or format rules.
Create content freshness indicators
Expose last successful sync and last content update where it helps editors and maintainers.
Project record status
Second encyclopedia record complete.
AGVN now has a durable public record covering its publishing architecture, automation decisions, performance work and operational lessons.