Skip to main content

Discovery enrichment beyond tagline + top offer

Detailed plan for surfacing the fields we already fetch in the same Sections() call that powers wolt feed and wolt venues. No extra HTTP — only widening what we render from the existing payload.

See docs/roadmap.md ("Discovery enrichment beyond tagline + top offer") for the high-level entry.

What we already pull​

  • tagline (from upstream short_description / short_description_v2.value).
  • top_offer (preferring discount-variant entries from promotions[]).

Both are rendered as columns on wolt venues rows and wolt feed sections.

What we still drop on the floor​

1. venue_preview_items​

Flagship dish + price per sponsored/featured row, e.g. "Cheeseburger pizza 19,90 €".

  • Add a menu_highlights[] field to venue rows in BuildDiscoveryFeed and BuildVenueSearchResult.
  • Default table: append one column Highlights (truncated).
  • Cleanest visual win, ships first.

2. badges_v2​

Newer badge format that carries an icon name (coupon-fill, bike, wolt-plus) and a variant. Today we only read the legacy badges array.

  • Surface as ASCII glyphs in the Venue cell, e.g. ◯ Wolt+, % 20% off, ⚡ Fast.
  • Falls back gracefully when the field is empty — no extra column.

3. Brand carousels​

feed currently skips sections whose items have no venue block — that drops "Popular stores", "Brands", and similar carousels.

  • Render them as a single one-line row per section: Brands: K-Market · Musti ja Mirri · ….
  • Same one-line treatment for any future non-venue carousel.

4. Grocery deals section — deferred (no upstream sample)​

Speculative when written; the live /v1/pages/front endpoint that wolt feed consumes does not carry a product-card grocery-deals section in any sampled city (Helsinki, Warsaw, Krakow, Berlin, Prague, Athens, Tokyo, May 2026). The payload-wide template enum is {hero, medium, square-title-bottom, venue, quicklink} and the keys old_price / new_price / original_price / discounted_price don't appear anywhere in the response.

Hypotheses for where the section actually lives:

  • A different endpoint — possibly /v1/pages/grocery-deals or a Wolt-Market-specific page (the brand carousel at woltmarket-popular-brands:helsinki hints at this).
  • Behind authentication — segmented to users with grocery order history.
  • Removed / not yet shipped.

Not implementing until a real sample payload exists. To revive: capture a populated payload, drop it under internal/service/observability/testdata/feed-grocery-deals.json, and revisit the design here. #2.1 (venue_preview_items + badges_v2) and #2.2 (brand carousels) ship without it and stand on their own.

Scope and commit slicing​

  • #2.1 — venue_preview_items + badges_v2 (shipped, commit 4d5cafa).
  • #2.2 — Brands one-liner (shipped, commit ce4a041).
  • #2.3 — Grocery deals — deferred, see section 4 above.

Tests​

  • Each enrichment field gets a unit test with a synthetic payload.
  • Brand-section test asserts the one-line render.
  • Grocery deals get an e2e snapshot of the new mini-table.

Risks​

  • Width. Adding columns to an already-wide venues table starts to wrap on narrow terminals. May need to make the Highlights column opt-in via a --show-highlights flag rather than default.
  • Icon rendering. ASCII glyphs (◯, %, ⚡) render inconsistently across terminals. If users hit boxes, fall back to plain text prefixes ([Wolt+], [20% off], [Fast]).

Output-contract impact​

If shipped, the JSON/YAML schemas in docs/output-contract.md grow new optional fields on the venue row:

menu_highlights[]: { name, formatted_price }
badges[]: { icon, variant, text }     # was string[] in the legacy shape

And a new section type for grocery deals on wolt feed:

sections[].kind: "venues" | "brands" | "grocery_deals"
sections[].brands[]: { name, slug }              # kind = "brands"
sections[].deals[]: { item_name, old_price, new_price, venue_name, venue_slug }
                                                  # kind = "grocery_deals"

These are additive — existing consumers keep working.