Create Vegan Food Itinerary

Purpose

Build a day-by-day vegan-food itinerary for a city covered by spinach.guide by synthesizing the site's Top Picks (Essentials), signature-dish rankings, dietary-core lists, and meal-occasion buckets (breakfast / lunch & brunch / dinner / sweet treats / etc.) into a structured plan for an arbitrary number of days, weighted against user-supplied constraints — preferences, allergies, cravings, price band, vegan-only vs. omni-friendly, and any anchor neighborhoods from an existing travel itinerary. Read-only; never books, reserves, or contacts a venue.

When to Use

• A user is traveling to one of the 158 ranked spinach.guide cities for N days and wants a meal-by-meal plan grounded in current rankings, not generic listicles.
• A user wants the plan filtered by hard constraints (gluten-free, nut allergy, soy-free) and soft preferences (Japanese cravings, $$ budget, walkable from Mission Bay, kid-friendly only).
• A user is staying in a specific neighborhood / has a fixed day-by-day location plan and wants meal stops bucketed closest to where they'll be each day.
• A user wants a "best of" weekend (2 days) vs. a deeper week-long (7 days) tour, and expects later days to dig past the obvious 15 essentials into hidden gems and dish-specific #1s.

Workflow

The site is a static Astro build behind Cloudflare with no anti-bot, no auth, no JS gating, and no JSON API — every page renders the full dataset server-side. The fastest path is direct HTTPS GETs against predictable URL slugs and parsing the markdown-rendered body. browse cloud fetch <url> returns the full HTML in a single hop (~80–120ms); browse open + browse get markdown body is a cheaper second option that yields pre-cleaned markdown. No browser session, no proxy, no --verified is required — confirmed against 5+ cities on iter-1. (The site does block Bytespider, CCBot, and meta-externalagent in robots.txt but allows everything else; default user-agents work.)

There is no /api/cities/... endpoint (returns 404). Everything is in the rendered HTML. Treat each page slug as the API.

URL atlas (the full vocabulary)

Not every city has every category. Smaller cities (e.g. Wellington, York) typically lack work-friendly, hidden-gems, or worth-the-occasion. Probe the sitemap or read the "On this page" nav of the city overview to confirm which sections exist before fetching. Hitting a missing category returns a 404 HTML page (custom Spinach 404, not an error envelope).

Steps

1. Resolve the city slug. Lowercase, hyphenated, no diacritics: san-francisco, new-york-city, ho-chi-minh-city, los-angeles. If the user gives a city name, slugify it; if uncertain, GET /sitemap-0.xml and grep for /cities/{candidate}/ — the sitemap is the source of truth.

2. Fetch the city overview (/cities/{slug}/). This single page is enough for a 1–2 day itinerary and gives you the full set of available category slugs for that city (read the "On this page" nav links to know what exists). Extract:

   • Intro paragraph (city's signature cuisine angle — e.g. "San Francisco leans Japanese ramen on signatures with 18 venues serving it")
   • Top 15 Essentials list (each card: rank, venue-slug, name, ★ rating, A–E grade, $ price band, distance-from-centre, cuisine description, one-line tip, dish rank badge if applicable e.g. #1 FOR RAMEN IN SAN FRANCISCO)
   • Signature dish sections (e.g. Best Ramen — top 4 venues per dish)
   • Fully Vegan / Omni Standouts (top 4 of each)
   • Each meal-occasion section's top 4

3. Fetch deeper category lists for longer itineraries. For 3+ days, also GET the explicit /breakfast-spots/, /lunch-brunch/, /dinner-destinations/, /sweet-treats/ pages — they return up to ~8 venues each with full at-a-glance metadata, today's hours, and the same one-line tips. For 5+ days, add /hidden-gems/ and any signature dish sub-rankings (/best/{dish}/{city}/) the user is craving.

4. Optional venue-detail enrichment. If the user asks for hours, address, lat/lon, or must-try dishes per stop, GET /venues/{venue-slug}/ for each shortlisted venue. The page exposes:

   • Spinach rating breakdown (food / service / atmosphere)
   • Vegan friendliness grade (A–E) — A = clearly labeled, knowledgeable staff, abundant options; E = one sad salad
   • Full weekly hours
   • Address + Google Maps directions URL with embedded destination={lat},{lon} — parse this to get coordinates for distance/cluster math
   • "Popular tips" (5 bullets of crowd wisdom)
   • "Must-try" dishes with popularity counts (e.g. breaded oyster mushrooms — popular (×15))
   • "Praised for" attribute tags (cozy atmosphere ×19, friendly service ×14)
   • Tags: Date Night, Dinner, Drinks, Small Groups, Special Occasion, Outdoor Seating, etc.
   • External Google Maps rating (sanity-check)

5. Apply user constraints. Run the candidate venue pool through the constraint stack in this order (each is a filter or a re-rank, not a hard cut unless the user marks it strict):

   Constraint type	Field on the venue card	Filter rule
   Hard allergies (gluten, nut, soy)	"Must-try" dishes + cuisine description	Drop venues whose only highlighted dishes contradict the allergy (e.g. drop dedicated bakeries for gluten-free; drop tofu-heavy spots for soy allergy). When the venue page has no contradicting signal, keep but flag.
   Vegan-only strict	"Fully plant-based" badge or /fully-vegan/ membership	Drop venues not in fully-vegan list.
   Price band	$, $$, $$$, $$$$	Drop venues outside the band.
   Cuisine cravings	Cuisine string + dish rank badges	Re-rank: boost venues with a #1 FOR {craving} badge to top of their slot.
   Neighborhood / day-anchor	distance from centre (km) on cards; lat/lon on venue page	When the user supplies an anchor (e.g. "we're in Mission Bay on day 2"), prefer venues with distance bands compatible with the anchor or geocode each candidate via the venue-page Maps URL and pick by haversine ≤ 2km from anchor.
   Family / date-night / group	Venue-page tag chips (Date Night, Special Occasion, etc.)	Match the night's intent to the tag.
   A–E grade floor	VFI grade letter on card	Drop below user's floor (default keep A–C, drop D/E unless they have a #1 dish rank).

6. Slot venues into the itinerary skeleton. Per day, fill four slots in this priority order, drawing from the filtered pool without repeating venues. Stop earlier if user requested fewer meals/day.

   Day N:
     Breakfast        ← /breakfast-spots/ pool, weighted by today's hours
     Lunch            ← /lunch-brunch/ ∪ /essentials/ (lunch-open) ∪ signature-dish #1s
     Coffee / Sweet   ← /sweet-treats/ (only if user opts in)
     Dinner           ← /dinner-destinations/ ∪ /worth-the-occasion/ (for "special" night) ∪ /essentials/ (dinner-open)
   

   Diversification rule: don't put two venues with the same dish-rank category on the same day (e.g. two ramen spots), and try to vary cuisine across the trip. Days 1–2 should weight toward Top 15 Essentials; days 3+ rotate in hidden gems, dish-#1s, and worth-the-occasion picks. End on a Worth the Occasion dinner if the trip is 3+ days.

7. Emit the itinerary as structured JSON (see Expected Output) with rationale strings citing the source page slug for each pick, so the user/agent can verify the recommendation against spinach.guide.

Browser fallback

When direct fetch fails (e.g. Cloudflare temporary block, transient 5xx) — none observed in testing, but the path is:

sid=$(browse cloud sessions create --keep-alive | node -pe "let s='';process.stdin.on('data',c=>s+=c).on('end',()=>process.stdout.write(JSON.parse(s).id))")
browse open "https://spinach.guide/cities/{slug}/" --remote --session "$sid"
browse get markdown body --remote --session "$sid"   # cleaner than HTML for synthesis

No --proxies or --verified needed; the site does not interrogate sessions.

Site-Specific Gotchas

• No JSON API: /api/cities/... and any obvious REST shape return a custom 404 page. The Astro build emits the full dataset into rendered HTML/markdown — that is the API. Do not waste turns probing for /api/*, /graphql, or Supabase endpoints (the embedded Supabase client is only used client-side for photo-failure telemetry; it does not serve venue data).
• The site is Cloudflare-fronted but does not challenge bots: bb fetch (no proxies, no --verified) returns 200s consistently. robots.txt allows User-agent: * except Bytespider, CCBot, meta-externalagent. Keep request rate ≤ 1–2 req/s out of courtesy; no rate-limit observed below that.
• Category slug coverage is non-uniform across cities: Larger cities (SF, NYC, London, Berlin, LA) have all 11 sub-category pages; smaller cities (Wellington, York, Zurich) typically expose only 6–8 (commonly miss work-friendly, hidden-gems, and sometimes worth-the-occasion). Always read the "On this page" nav on /cities/{slug}/ before fetching sub-paths, or grep the sitemap. Hitting a missing category returns a custom 404 HTML page (not a JSON error).
• Slug variants for multi-word cities: new-york-city (not new-york), los-angeles, san-francisco, ho-chi-minh-city, tel-aviv. When in doubt, fetch /sitemap-0.xml (one file, ~250KB) and substring-match.
• Distance is from the city centre, not from the user: Every card shows X km from centre — this is centroid-relative, not user-relative. For neighborhood-anchored itineraries, you must geocode the user's anchor and the venue's lat/lon (from the venue page's https://www.google.com/maps/dir/?api=1&destination={lat},{lon} URL) and compute haversine yourself.
• Vegan Friendliness Index (VFI) grade ≠ Spinach rating: A venue can be ★ 5.0 (outstanding food) and D graded (limited vegan options) — the 5-star rating reflects the food's actual quality, the A–E grade reflects vegan-friendliness (options count, protein sources, menu clarity, staff knowledge). For a vegan-only audience, treat the grade as the primary filter and the star rating as the tie-breaker.
• "Fully plant-based" badge is the strict-vegan signal: When a venue card shows this badge or appears in /cities/{slug}/fully-vegan/, the kitchen has no animal products. Omni-standout venues require asking and may carry cross-contamination risk; flag these when the user has a strict-vegan stance.
• Dish-rank badges (#1 FOR RAMEN IN SAN FRANCISCO) only appear on cards in cities where that dish has ≥ a handful of contenders. If a user craves a dish not surfaced in the city's signature sections, check /best/{dish}/{city}/ directly — the per-dish page exists even when it didn't make the city's signature shortlist.
• Today's hours on category-list pages reflect the page's CDN cache time, not real-time. They're refreshed daily ("last refreshed today" banner); for time-of-day-sensitive plans (Sunday brunch, late-night dinner), open the venue page for the full weekly schedule rather than trusting the list card.
• Venue slugs are stable but contain numeric suffixes (e.g. aiso-124450, the-butterfly-joint-2421). Don't try to construct them — always lift them from anchor hrefs on the city/category page.
• Photos lazy-fail through a failover chain: The page embeds data-fallbacks="..." per <img> and reports broken photos back to Supabase. This is a UI concern only — has no impact on data extraction.
• /best/{dish}/ (city-omitted) is a cross-city dish ranking, not a global venue list — useful only when the user wants to compare cities for a specific dish, not for itinerary building within one city.
• 159 cities ≠ 158 cities: The homepage advertises 158 cities; the actual sitemap count drifts daily as cities are added. The "All N cities" link in the homepage is the live count. Don't hardcode.

Expected Output

Emit one JSON object with the shape below. The rationale strings cite the spinach.guide page each pick was sourced from so the user/agent can audit.

{
  "city": "san-francisco",
  "city_display": "San Francisco",
  "days": 3,
  "user_constraints": {
    "strict_vegan": true,
    "allergies": ["gluten"],
    "cravings": ["ramen", "ice_cream"],
    "price_band_max": "$$$",
    "vfi_floor": "B",
    "day_anchors": [
      { "day": 1, "neighborhood": "Mission", "lat": 37.7599, "lon": -122.4148 },
      { "day": 2, "neighborhood": "Marina", "lat": 37.8030, "lon": -122.4378 },
      { "day": 3, "neighborhood": "Downtown", "lat": 37.7879, "lon": -122.4075 }
    ],
    "meals_per_day": ["breakfast", "lunch", "dinner"],
    "end_on_special_occasion": true
  },
  "itinerary": [
    {
      "day": 1,
      "anchor": "Mission",
      "meals": [
        {
          "slot": "breakfast",
          "venue": "judahlicious",
          "venue_url": "https://spinach.guide/venues/judahlicious-2080/",
          "name": "Judahlicious",
          "cuisine": "Vegan Cafe, Burritos & Bowls",
          "rating": 5.0,
          "vfi_grade": "A",
          "price": "$$",
          "fully_vegan": true,
          "distance_from_centre_km": null,
          "today_hours": "8 AM-3 PM",
          "must_try": ["Nekked Burrito"],
          "rationale": "From /cities/san-francisco/breakfast-spots/: fully-plant-based cafe, rating 5.0 A, matches strict_vegan + gluten-friendly burrito-bowl menu."
        },
        {
          "slot": "lunch",
          "venue": "menya-kanemaru-golden-ramen",
          "venue_url": "https://spinach.guide/venues/menya-kanemaru-golden-ramen-2363/",
          "name": "Menya Kanemaru Golden Ramen",
          "cuisine": "Japanese Ramen, Vegan Curry",
          "rating": 5.0,
          "vfi_grade": "A",
          "price": "$$",
          "dish_rank": { "dish": "ramen", "rank": 3, "scope": "san-francisco" },
          "rationale": "From /cities/san-francisco/essentials/ + /best/ramen/san-francisco/: matches user craving 'ramen', #3 in city, A-grade, within Mission anchor radius."
        },
        {
          "slot": "dinner",
          "venue": "aiso",
          "venue_url": "https://spinach.guide/venues/aiso-124450/",
          "name": "Aíso",
          "cuisine": "Vegan Tapas, Oyster Skewers",
          "rating": 5.0,
          "vfi_grade": "A",
          "price": "$$$",
          "fully_vegan": true,
          "tags": ["Date Night", "Dinner", "Drinks", "Small Groups", "Special Occasion"],
          "rationale": "From /cities/san-francisco/dinner-destinations/ + /cities/san-francisco/fully-vegan/: #5 in essentials, #1 for breaded oyster mushrooms, address 4068 18th St (Mission-adjacent)."
        }
      ]
    },
    {
      "day": 2,
      "anchor": "Marina",
      "meals": [
        {
          "slot": "breakfast",
          "venue": "the-butterfly-joint",
          "venue_url": "https://spinach.guide/venues/the-butterfly-joint-2421/",
          "name": "The Butterfly Joint",
          "cuisine": "Coffee Shop, Vegan Donuts",
          "rating": 5.0,
          "vfi_grade": "A",
          "price": "$$",
          "dish_rank": { "dish": "donut", "rank": 2, "scope": "san-francisco" },
          "rationale": "From /cities/san-francisco/breakfast-spots/: #1 in essentials, but 7.3km from centre — flag distance from Marina anchor (~7km) as a stretch; consider Whack Donuts as a closer swap."
        },
        {
          "slot": "lunch",
          "venue": "torraku-ramen-lombard",
          "venue_url": "https://spinach.guide/venues/torraku-ramen-lombard-2525/",
          "name": "Torraku Ramen - Lombard",
          "cuisine": "Japanese Ramen, Vegan Miso",
          "rating": 5.0,
          "vfi_grade": "B",
          "price": "$$",
          "dish_rank": { "dish": "ramen", "rank": 1, "scope": "san-francisco" },
          "rationale": "From /cities/san-francisco/essentials/: #1 ramen in city, Lombard is in Marina/Russian Hill — best anchor fit on day 2."
        },
        {
          "slot": "dinner",
          "venue": "destapas",
          "venue_url": "https://spinach.guide/venues/destapas-2320/",
          "name": "Destapas",
          "cuisine": "Spanish Restaurant, Vegan Paella",
          "rating": 5.0,
          "vfi_grade": "A",
          "price": "$$",
          "rationale": "From /cities/san-francisco/dinner-destinations/: vegan paella standout, varies cuisine from day 1 (Japanese)."
        }
      ]
    },
    {
      "day": 3,
      "anchor": "Downtown",
      "meals": [
        {
          "slot": "breakfast",
          "venue": "salt-straw",
          "venue_url": "https://spinach.guide/venues/salt-straw-2162/",
          "name": "Salt & Straw",
          "cuisine": "Ice Cream, Vegan Flavors",
          "rating": 5.0,
          "vfi_grade": "B",
          "price": "$$$",
          "dish_rank": { "dish": "ice_cream", "rank": 1, "scope": "san-francisco" },
          "rationale": "From /cities/san-francisco/essentials/ + user craving 'ice_cream': #1 ice cream in city, 0.5km from centre = walkable from Downtown anchor. Slotted as 'sweet/coffee' rather than savory breakfast."
        },
        {
          "slot": "lunch",
          "venue": "imperial-garden",
          "venue_url": "https://spinach.guide/venues/imperial-garden-2420/",
          "name": "Imperial Garden",
          "cuisine": "Chinese Dim Sum, Vegan Menu",
          "rating": 5.0,
          "vfi_grade": "B",
          "price": "$$",
          "dish_rank": { "dish": "dumplings", "rank": 2, "scope": "san-francisco" },
          "rationale": "From /cities/san-francisco/essentials/: #2 dumplings, third distinct cuisine in the trip (Chinese)."
        },
        {
          "slot": "dinner",
          "venue": "pena-pachamama",
          "venue_url": "https://spinach.guide/venues/pena-pachamama-2084/",
          "name": "Peña Pachamama",
          "cuisine": "Vegan Latin Restaurant, Raw & Cooked",
          "rating": 5.0,
          "vfi_grade": "A",
          "price": "$$$",
          "fully_vegan": true,
          "tags": ["Special Occasion"],
          "rationale": "From /cities/san-francisco/worth-the-occasion/ + /cities/san-francisco/dinner-destinations/: live Latin music on weekend nights, fully plant-based, end-on-special-occasion satisfied."
        }
      ]
    }
  ],
  "sources_consulted": [
    "https://spinach.guide/cities/san-francisco/",
    "https://spinach.guide/cities/san-francisco/essentials/",
    "https://spinach.guide/cities/san-francisco/breakfast-spots/",
    "https://spinach.guide/cities/san-francisco/lunch-brunch/",
    "https://spinach.guide/cities/san-francisco/dinner-destinations/",
    "https://spinach.guide/cities/san-francisco/worth-the-occasion/",
    "https://spinach.guide/cities/san-francisco/fully-vegan/",
    "https://spinach.guide/best/ramen/san-francisco/"
  ],
  "warnings": [
    "Day 2 breakfast (The Butterfly Joint) is ~7km from the Marina anchor — consider Whack Donuts (3.4km from centre) as a closer swap.",
    "VFI floor 'B' kept Torraku Ramen (B) in; if user tightens to 'A' floor, replace with Menya Kanemaru Golden Ramen (already on day 1)."
  ],
  "data_freshness": "spinach.guide cards are refreshed daily; this itinerary built from pages fetched on 2026-05-19."
}

Outcome shapes