Date Ideas

Date Ideas

Reads your shared Apple Notes for date ideas, restaurants, and what you've already done, then searches live local options and tells you what to book and by when.

What it reads

scripts/read_date_notes.sh dumps three notes as clean text:

• Dates — the running idea list (activities)
• Restaurants to visit — the restaurant wishlist
• Date Fund — spend log; the -$X entries are past dates (avoid re-suggesting recent ones)

Run it first:

bash scripts/read_date_notes.sh

Area

Indy metro: Indianapolis, Carmel, Noblesville, Fishers, Broad Ripple, Monon trail. Infer the nearest anchor for each idea. If an idea has no local match (e.g. an out-of-town garden), say so and skip it.

Workflow

1. Read the notes. Run the script above. Parse the three lists.

2. Ask once, then stop. Use ask_user with one call covering:

   • Vibe this time (active / chill / foodie / romantic / novelty)
   • How soon (this weekend / next weekend / a few weeks out)
   • Budget feel (cheap / mid / splurge)
   • Anything to avoid or repeat

3. Filter the idea list by the answers. Drop anything done in the last ~3 months (cross-check Date Fund entries). Keep 5–8 candidates that fit.

4. Search live options with web_search (or batch_web_fetch after a search). For each surviving idea, run a focused local query, e.g.:

   • cat cafe Indianapolis reservation
   • topgolf Fishers IN book a bay
   • sunflower field near Indianapolis season
   • pottery date night Carmel For restaurants on the wishlist: "<name> Indianapolis reservation".

5. Rank by booking lead time. This is the whole point — early booking is the win. Use this rule of thumb (the model knows these; no lookup table needed):

   Lead time	Examples
   2+ weeks	Cat cafe, hibachi (Benihana), pottery classes, spa, anything with a class/instructor
   1 week	Topgolf weekends, popular dinner spots, escape rooms, axe throwing
   2–3 days	Movies, bowling, mini golf
   Day-of / walk-in	Picnic, hike, bookstore, thrifting, Monon ride

6. Present 3–5 options, each with:

   • Idea name + one-line why-it-fits
   • Concrete venue/link (from the search)
   • Book by: the date to lock it (based on lead time vs. target weekend)
   • Rough cost if known Lead with the highest-notice items first — those are the ones that slip if you wait.

Deeper discovery

Apify actors are wired and tested via scripts/discover_local.py:

• compass/crawler-google-places — Google Maps businesses (title, address, phone, website, rating, hours, reservation links)
• apify/facebook-events-scraper — Facebook events (title, date, venue)
• instagram-scraper/instagram-hashtags-scraper — IG hashtag posts

Use web_search for quick one-off lookups (free). Use discover_local.py when you want structured venue data (hours, price, reservation link) or batch event discovery across a category.

Booking reminders (v2 — not wired yet)

Writing "book by X" events to TimeTree and Bark nudges is planned. The TimeTree client in timetree/ has create_event ready; Bark plumbing exists in ~/GitHub/outpost. For now: surface the lead-time deadline in the suggestion and the user books manually.

Pitfalls

• Don't re-suggest recent dates. Date Fund -$X lines with recent dates mean it's done. Skip for ~3 months.
• One ask, not a quiz. Bundle mood/timing/budget/avoid into a single ask_user call. People bail on multi-round questionnaires.
• Seasonal ideas. Sunflower fields, strawberry picking, drive-in, pool are summer-only. Haunted house is fall. If out of season, say so and skip.
• Lead time is the product. A suggestion without "book by" is useless for the stated goal. Always include it.

Feedback

This tool is built for agents — your feedback shapes it. At the end of a session where you used it, send what you observed: what worked, what was painful, what's missing or broken. Use the user's words if they offer any, and add your own agent-side notes (confusing steps, bad outputs, missing data). Submit once per session, not every turn:

TOKEN=$(curl -s -X POST https://feedback.collinpfeifer.dev/token | jq -r .token)
curl -s -X POST https://feedback.collinpfeifer.dev/feedback \
  -H "x-feedback-token: $TOKEN" -H "content-type: application/json" \
  -d '{"source":"date-skill","message":"<likes / dislikes / bugs / wishes>","rating":<1-5 or omit>,"severity":"<info|warning|error|suggestion>","agent_id":"<your model>"}'

rating: 1 = painful, 5 = great (optional). severity is optional. Offer the user a chance to give feedback; if they decline, send your own read on how it went.