room.ventures

A static website for hotel room reviews, built with Hugo.

Overview

room.ventures is a static site that showcases and reviews hotel rooms. Photos are dumped into a folder, processed by an AI agent that extracts EXIF data, analyzes room quality, and generates static Markdown content pages with ratings. The final site features an interactive dark-themed map of all reviewed locations.

Tech Stack

• Static Site Generator: Hugo
• Content Format: Markdown with TOML front matter
• EXIF Extraction: ExifTool
• Image Analysis: llama3.2-vision via Ollama
• Interactive Map: Leaflet.js with OpenStreetMap tiles
• Map Tiles (Detail Pages): Static OpenStreetMap tile images
• Image Storage: Git LFS
• Container: Docker (Hugo build + nginx)
• CI/CD: GitHub Actions (build to GHCR, deploy via SSH)

Ingesting New Photos

Drop geotagged hotel room photos (JPEG) into photos/inbox/ and run OpenCode with:

based on the defined workflow, process all images in photos/inbox

The AI agent will automatically:

1. Deduplicate -- compute SHA-256 hashes and skip any photo already in photos/processed_hashes.json
2. Extract EXIF -- pull date, GPS coordinates, and camera info via ExifTool
3. Group -- cluster photos by timestamp proximity into hotel room visits
4. Reverse-geocode -- resolve GPS coordinates to hotel name, city, and country (prompts you if ambiguous)
5. Analyze -- assess each photo with llama3.2-vision for style, condition, luxury level, and cleanliness
6. Generate rating -- produce a composite 1-5 star rating
7. Download map tile -- fetch a static OpenStreetMap tile for the location
8. Create review page -- generate a Hugo content page in content/reviews/ with full metadata
9. Update map data -- add the location to data/locations.json for the homepage map
10. Move photos -- copy processed images to static/images/reviews/<slug>/
11. Update manifest -- add new SHA-256 hashes to photos/processed_hashes.json

Requirements

Photos must be JPEG files with EXIF GPS coordinates embedded (most smartphone cameras do this by default). The agent needs:

• ExifTool (installed automatically via Homebrew if missing)
• Ollama with llama3.2-vision:11b (pulled automatically if missing)

After Processing

Commit and push the generated files:

git add content/ data/ static/ photos/processed_hashes.json
git commit -m "Add new hotel reviews"
git push

The GitHub Actions pipeline will automatically build the Docker image and deploy.

Getting Started

Prerequisites

• Hugo (extended edition recommended)
• ExifTool
• Ollama with llama3.2-vision:11b
• Git LFS
• Docker (for containerized builds)

Setup

git lfs install
git lfs track "*.jpg" "*.jpeg" "*.png" "*.webp"

Local Development

hugo server -D

Docker Build

docker build -t room-venture .
docker run -p 8080:80 room-venture

Production Build

hugo

The generated site will be in the public/ directory.

Project Structure

room.ventures/
├── photos/
│   ├── inbox/                 # Drop raw photos here for processing
│   └── processed_hashes.json  # SHA-256 manifest for duplicate detection
├── content/
│   └── reviews/               # Generated review pages (Markdown)
├── static/
│   └── images/
│       ├── reviews/           # Processed review photos
│       └── maps/              # Static OpenStreetMap tile images
├── layouts/
│   ├── _default/              # Base layout, homepage with map
│   └── reviews/               # Review list and detail templates
├── data/
│   └── locations.json         # Location data for the homepage map
├── .github/
│   └── workflows/             # Build and deploy pipelines
├── Dockerfile                 # Multi-stage build (Hugo + nginx)
├── hugo.toml                  # Hugo configuration
├── AGENTS.md                  # AI agent instructions
└── README.md