Nearby: variety for smaller metros (press revived, topics, spread)

docs/design/morning.md

@@ -16,6 +16,8 @@ Three things hold it together:
 - **The format is enforced in code, with one redraft** (`_is_structured`, same remedy as `_NAMES_THE_LINK`), and only when there is something to bullet (`need = min(2, len(required))`) — a user with nothing loaded still gets a plain greeting. A redraft only replaces the draft if it is better formed, so a fix for one rule cannot break another.
 - **Length is capped at a line boundary** (`_tidy_morning_text`, `MORNING_LINE_MAX` 560). Whitespace collapses within a line but newlines survive; the old single-space join would have flattened the bullets, and a mid-line cut leaves half a fact.
 
+**Weather and markets are in every update.** WEATHER is required whenever the payload has weather (it needs a city). MARKETS is required whenever the payload has a price row, and a user who follows no ticker gets one anyway: `home._fetch_prices` falls back to `tickers.DEFAULT_MARKETS` (S&P 500 and Dow) only when *nothing* in their topics resolved to a symbol. Tickers that resolve but fail to fetch do not trigger it — those are the user's chosen list, and a rate-limit blip must not swap it for someone else's. A user who hid the Markets section gets neither the fallback nor the requirement. The same payload feeds the page, so their Markets card shows the indices too.
+
 Two properties are load-bearing:
 - **The URL is last and alone.** Message apps only draw the rich link preview when the message carries exactly one URL at a boundary, and that preview is most of the value. Nothing may follow it — not a period, not a sign-off.
 - **`carries_link` gates the status callback.** A link message is sent with `add_status_callback=False`, because the `/sms-status` shorten-and-retry would truncate the URL into garbage.

docs/design/opening.md

@@ -37,6 +37,35 @@ Three things are load-bearing:
   on Haiku, once per city, cached — deliberately *not* a lookup table, which is
   the mistake `tickers.py` made twice with `PRIVATE_COMPANIES`. Ticketmaster
   sidesteps the problem entirely by taking `latlong` + radius instead of a name.
+- **A smaller metro came back all concerts, and four things caused it.** St.
+  Louis showed Black Country New Road, D.L. Hughley and Journey, week after week.
+  (1) `_local_candidates` was `trusted_only`, and stltoday, stlpr, stlmag, fox2now
+  and ksdk are the real newsrooms and none is on the allowlist — it returned
+  nothing, so the whole press half was silently empty. It now runs untrusted at
+  `LOCAL_MIN_SCORE` 0.60 (`_search_raw` still sorts trusted first) over four
+  queries — openings, "festivals fairs events {Month Year}", "festival this
+  weekend", "family events … this weekend" — on a 21-day window, because the
+  14-day one made the same query swing between zero and eight hits. (2)
+  Ticketmaster was Music and Arts only; Miscellaneous (fairs, festivals, family
+  shows) is added, Sports deliberately is not (regular-season fixtures belong to
+  people who follow the team, which is a per-user question). (3) One shared
+  prompt cut at 12k characters meant ~8.6k of concerts and a clipped sliver of
+  press; `_fit` now holds each family to its own budget and drops whole items
+  rather than cutting one mid-JSON. (4) With both families in one call the model
+  took the first good candidates it met, which were the concerts, whatever the
+  prompt said about spread — so events and press are **curated in separate
+  calls** (`EVENT_ROWS`/`PRESS_ROWS` 5 each, `_interleave`d), and the per-topic
+  cap (`MAX_PER_TOPIC` 2) is enforced in code. Rows carry a `topic`
+  (music, comedy, theatre, food, festival, family, film, other) and `_spread`
+  takes one of each topic before a second at read time.
+- **A press-sourced event must carry a date or it is dropped.** The curator lifts
+  individual events out of weekend roundups, and a date there is the model's
+  reading of prose. Undated, a row can never expire, and "Oct 9-11" was served on
+  the 14th. The prompt asks for the LAST day of a multi-day event; Ticketmaster
+  rows have a real date upstream and are exempt. Cost: Tavily goes from 2 to 4
+  searches per metro per week and Haiku from 1 to 2 calls per metro per day.
+  Preference-driven ordering (family, teams, what the user has said) is the
+  next layer and is not built; today every user in a metro shares one ranking.
 - **The local outlets had to be added to `trusted_sources.json`.** Palmer Home
   passes `trusted_only=True`, and Eater, Time Out, LAist, Thrillist and the city
   dailies were all tier 3 — so before they were added the section returned

palmer/agent.py

@@ -228,7 +228,14 @@ def _build_system(phone: str, include_recent: bool = False, is_new_user: bool =
         # once — userprofile._update_profile marks score_offer_sent under this
         # same condition after the turn, so this never becomes a nag.
         teams = profile.get("sports_teams")
-        teams = ", ".join(teams) if isinstance(teams, list) else str(teams)
+        # sports_teams is meant to be prose, but a profile can hold follow-list
+        # dicts there (the old `teams` alias migrated them in). Join names, or
+        # this block raises and takes the whole turn with it.
+        if isinstance(teams, list):
+            teams = ", ".join(t.get("name", "") if isinstance(t, dict) else str(t)
+                              for t in teams)
+        else:
+            teams = str(teams)
         system += (
             "\n\nLIVE SCORES OFFER\n"
             f"They've mentioned a team they follow ({teams}) and nothing is set up for it. "

palmer/home.py

@@ -192,14 +192,22 @@ def _fetch_prices(profile: dict, previous: list[dict] | None = None) -> list[dic
     exactly like Palmer forgetting, and is far worse than a slightly stale
     number sitting under a visible "N min ago" stamp."""
     from palmer.datafeeds import price_snapshot
-    from palmer.tickers import resolve_topic_asset
+    from palmer.tickers import resolve_topic_asset, DEFAULT_MARKETS
     stale = {(p.get("label") or "").lower(): p for p in (previous or [])}
     assets, seen = [], set()
     for topic in (profile.get("morning_topics") or []):
         got = resolve_topic_asset(topic)
         if got and got[0].lower() not in seen:
             seen.add(got[0].lower())
             assets.append(got)
+    # Nothing resolved to a symbol: show the broad market rather than an empty
+    # section, so every user's morning has a markets line. Only when NOTHING
+    # resolved — someone tracking tickers whose fetches all failed chose those
+    # tickers, and indices would be somebody else's list. A user who hid the
+    # section asked not to see it.
+    hidden = (profile.get("morning_prefs") or {}).get("hidden_sections") or []
+    if not assets and "markets" not in hidden:
+        assets = list(DEFAULT_MARKETS)
     out = []
     for symbol, label in assets[:MAX_PRICES]:
         snapshot = price_snapshot(symbol, label) or stale.get(label.lower())

palmer/main.py

@@ -388,7 +388,7 @@ async def artifact_page(token: str):
 async def icon_png():
     """The page's favicon and apple-touch-icon. No token: it is the same mark
     for everyone and carries no one's data."""
-    from cards import render_icon
+    from palmer.cards import render_icon
     return Response(content=render_icon(), media_type="image/png",
                     headers={"Cache-Control": "public, max-age=86400"})
 

palmer/morning.py

@@ -572,6 +572,10 @@ def generate_morning_line(phone: str, payload: dict) -> str:
         required.append("the commute")
     if score_lines(payload):
         required.append("their team — yesterday's result and/or today's game, with the score or the time")
+    hidden = (payload.get("page_prefs") or {}).get("hidden_sections") or []
+    if payload.get("prices") and "markets" not in hidden:
+        required.append("the markets — how each price row below moved, with its percent "
+                        "(for most people that is just the S&P 500 and the Dow)")
     if payload.get("opening"):
         required.append("1-2 things newly open or worth catching near them this week, named specifically")
     required_block = (
@@ -591,7 +595,7 @@ def generate_morning_line(phone: str, payload: dict) -> str:
 Rules:
 - Cover every REQUIRED item above, in your own words, with real specifics — the actual temperature, the actual commute time, the actual name of the place or event. Never gesture at a category ("some stuff opened nearby") instead of naming it.
 - FORMAT: one opener line, then one bullet per item. The opener is a single short sentence in Palmer's voice (the day's character, a greeting, or the one thing that stands out) carrying no data the bullets also carry. Each bullet is its own line, exactly "- LABEL: text", where LABEL is ALL CAPS and one of: WEATHER, COMMUTE, NEARBY, SCORES, MARKETS, NEWS. Use WEATHER for weather, COMMUTE for the drive, NEARBY for openings/events, SCORES for a followed game. Only include a bullet for an item the data actually has. Each bullet is one tight line of plain words, not a clipped fragment. No blank lines between bullets.
-- Beyond the required items, you may add ONE more bullet (MARKETS or NEWS) about something else on the page ONLY if it is genuinely notable — a real change, not routine. Skip it rather than padding; this text is basics plus link, not the briefing.
+- Beyond the required items, you may add ONE more NEWS bullet about something else on the page ONLY if it is genuinely notable — a real change, not routine. Skip it rather than padding; this text is basics plus link, not the briefing.
 - Under {MORNING_LINE_MAX} characters total, counting every line. Tight and scannable.
 - Never say the word "link", "page", "dashboard", "site", or "click" - the link speaks for itself and naming it makes this sound like a product notification.
 - Write ONLY the text. The link is attached automatically after it. Do not write a URL, and do not leave a placeholder like [link] or (url) where you think one goes - anything like that ships to them as literal text.