Bose

My e-commerce site is bose.com. We don't publish an llms.txt file yet — the emerging convention (llmstxt.org) that gives language models a concise, curated guide to a site.

In this repository:

1. Create /llms.txt (served at https://bose.com/llms.txt as text/plain or text/markdown) following the llms.txt format:
   - H1 with our brand name,
   - a one-paragraph blockquote summary of what we sell and who we serve,
   - sections linking to our most useful pages for an AI agent: bestsellers/category pages, shipping & returns policy, size guides, FAQ/support, and store locator if any.
2. Write the summary from this repository's real content (README, about page, homepage copy) — keep it factual, no marketing superlatives.
3. Keep it under ~200 lines, every link absolute.
4. Show me the full file content and where you wired it to be served.

Verify after deploy: curl https://bose.com/llms.txt

My e-commerce site is bose.com. A scan found no machine-readable product feed, which feed-based AI shopping agents (ChatGPT Shopping, Klarna, Google AI Mode) rely on. The site runs on salesforce, so prefer that platform's idiomatic way of serving these files/markup (theme templates, app settings, or metafields) over hand-rolled middleware where it exists.

In this repository:

1. Determine where product data lives (database models, CMS, platform API) and add a product feed endpoint — Google Merchant–compatible XML (RSS 2.0 with the g: namespace) at /feeds/products.xml, or JSON if that's more idiomatic here.
2. Include per item: id, title, description, link, image_link, price with currency, availability, brand, and gtin/mpn when we have them.
3. Paginate or stream if the catalog is large; cache for ~1 hour.
4. Link the feed from robots.txt or a <link rel="alternate"> in the layout, show me the diff, and print the first two feed items rendered from real data.