ithinkihaveacat/interior-design

Interior Design (Inigo)

Important: Use Scripts First

ALWAYS prefer the scripts in scripts/ over hand-rolled HTTP and HTML parsing. They encode site-specific knowledge that is otherwise easy to get wrong:

• The exact JSON path Inigo uses for gallery images (.listing.gallery.images[].imageUrl)
• Unwrapping the RSC self.__next_f.push(...) chunk format to get to the listing payload

When to read the script source: if a script doesn't do what you need, or fails because Inigo has changed their page structure again, read the source — it's short, and the parsing details are the actual value.

Quick Start

Find listings matching criteria

Two modes:

scripts/inigo-search website [filters]    # live data, ≤26 per section, filterable
scripts/inigo-search sitemap [filters]    # full catalogue (~700 URLs), slug only

website queries the embedded initialData on /all-homes and /past-sales (merging both by default; pass --active or --sold to narrow). Pagination is JS-only, so each section yields ≤ 20 summaries plus 6 with --include-featured. Supports price/bedrooms/address filters; emits URLs (or full summaries with --json).

sitemap enumerates every listing URL from https://www.inigo.com/sitemap.xml — ~520 sold + ~170 active. Best for archive enumeration or initial pulls. Only slug-substring filtering is available because the sitemap carries no metadata beyond <loc> and <lastmod>.

Both modes print catalogue counts to stderr.

Get gallery image URLs

scripts/inigo-gallery https://www.inigo.com/past-sales/<slug>

Prints one image URL per line on stdout. Add --json to get the full parsed listing payload (address, price, body text, gallery, etc.) instead.

Download a gallery into a directory

mkdir <slug> && scripts/inigo-gallery <URL> | wget -i - -P <slug>/

Search → download in one pipeline

scripts/inigo-search website --sold --price-to 500 --limit 3 \
  | while read url; do
      slug=$(basename "$url"); mkdir -p "$slug"
      scripts/inigo-gallery "$url" | wget -q -i - -P "$slug/"
    done

Or with parallel downloads:

scripts/inigo-gallery <URL> | xargs -n1 -P4 -I{} curl -sO --output-dir <slug>/ {}

Triage downloaded images with an LLM

Triage is best delegated to a dedicated LLM image-query tool rather than hand-rolled API calls. The right shape of tool takes a free-form question plus a structured-output schema (e.g. "has_bedside_table bool"), runs it against a directory of images, and can filter to just the paths whose boolean field is true — ideally handling EXIF rotation, resizing, and caching of pre-processed bytes so repeated queries don't re-encode.

Search the installed skills for such a tool before writing anything yourself — this dotfiles ecosystem ships one. Check the available skills (and bin/) for an image- or photo-query tool and use its documented interface.

Scripts

• scripts/inigo-search website [filters] — query the live index pages (/all-homes, /past-sales). Filters: --active/--sold (default merges both), --price-from K, --price-to K (£000s), --bedrooms N, --min-bedrooms N, --address SUBSTR, --include-featured, --limit N, --json. Limited to ≤ 26 listings per section because pagination is JS-only; see the reference doc for why. Zero deps.
• scripts/inigo-search sitemap [filters] — enumerate every listing URL via sitemap.xml (~690 URLs). Filters: --active/--sold (default merges both), --address (slug substring), --limit N, --json. Use for archive enumeration or initial pulls. Zero deps.
• scripts/inigo-gallery URL — print gallery image URLs from an Inigo listing. --json emits the full listing payload instead. Zero deps. See references/inigo-source-structure.md for the JSON paths used.

For LLM-based image triage, use a dedicated image-query tool from the installed skills (no duplicate here; see "Triage downloaded images with an LLM" above).

Example queries

Worked-in-practice prompts the existing data was triaged with, run via an LLM image-query tool (prompt plus structured-output schema):

| Topic | Prompt | Schema | | -------------------- | ------------------------------------------------------------------------------------------ | -------------------------------------------------------- | | Bedside table | "Does this image feature a bed? Does it feature a bedside table?" | "has_bed bool, has_bedside_table bool" | | Mismatched chairs | "Does this image feature a dining table with mismatched chairs (not in the same style)?" | "has_mismatched_chairs bool" | | Fireplace + art + TV | "Does this image show a fireplace? Artwork above a fireplace? A TV?" | "fireplace bool, artwork_over_fireplace bool, tv bool" |

Safety Notes

• inigo-gallery only fetches HTML; it does not download images. The caller decides what to do with the URLs.
• LLM image triage sends each (resized) image to a remote API. Costs scale with directory size × queries; a local disk cache only amortises the resize/encode step, not the per-query API spend.

Reference Material

• references/inigo-source-structure.md — URL surface (/all-homes, /past-sales, /sales-list/<slug>), both Inigo gallery JSON formats, search/discovery JSON shape, observed filename patterns, verification probes.