Find Events

Calculate exact dates using today's date. Only include future events.

Step 3: Ingest Sources into qurl

Scrape each priority source and ingest it into the local qurl database. Run in parallel where possible.

PLUGIN_ROOT="${CLAUDE_PLUGIN_ROOT:-plugins/berlin-events}"

ingest() {
  local url="$1" tags="$2"
  bun run "$PLUGIN_ROOT/scripts/extract-content.js" "$url" \
    | qurl add "$url" --source berlin-events --tags "$tags"
}

# Art sources
ingest "https://www.indexberlin.com/events/list/"                                         "art"
ingest "https://www.kw-berlin.de/en/events"                                               "art"
ingest "https://berlinischegalerie.de/programme/kalender/"                                "art"
ingest "https://www.artatberlin.com/en/calendar-for-vernissagen-exhibitions-events/"      "art"
ingest "https://co-berlin.org/de/programm/kalender"                                       "art"
ingest "https://kunstleben-berlin.de/events/"                                             "art"

# Food / broad sources
ingest "https://www.berlin.de/en/events/"                  "food"
ingest "https://www.visitberlin.de/en/event-calendar-berlin" "art,food"

qurl deduplicates by URL + content hash — re-running is safe.

Step 4: Embed

Generate vector embeddings for any newly ingested documents:

qurl embed

Step 5: Search with qurl vsearch

Build a query from the user's settings and run semantic search:

# Build query from interests + neighborhood + current month
QUERY="$(date '+%B %Y') Berlin exhibition opening vernissage workshop event calendar art food"

qurl vsearch "$QUERY" \
  --source berlin-events \
  --limit 20

Relevance filter — a result counts as relevant if its snippet contains any of:

• EN: april, may, monday–sunday, vernissage, opening, exhibition, finissage
• DE: ausstellung, veranstaltung, führung, kalender, programm
• Dates: current year, .04., .05. (adjust for current month)

If vsearch returns fewer than 5 relevant results, fall back to web search (Step 5b).

Step 5b: Web Search Fallback (only if vsearch < 5 results)

"Berlin art events this week [date range]"
"Berlin food events this week [date range]"
"Berlin exhibition openings [date range]"

Step 6: Check Google Calendar

Check for scheduling conflicts:

gcalcli agenda --nocolor --details end "today" "+14 days"

If gcalcli is not installed: pip3 install gcalcli. If not authenticated, ask the user to run ! gcalcli agenda to complete OAuth.

Parse output to identify busy time slots. Flag events that overlap with existing entries.

Step 7: Add Location Context

For each event, note the neighborhood (e.g., "Kreuzberg", "Mitte", "Charlottenburg") and compare with the user's neighborhood:

• Same neighborhood: "Near you"
• Adjacent: "~15 min by transit"
• Far: "~30+ min by transit"

Use general Berlin geography knowledge. Do not call external routing APIs.

Step 8: Rank and Curate

Score events by:

1. Relevance to user interests (art/food)
2. Time fit — no calendar conflicts, within lookahead window
3. Proximity to user's neighborhood
4. Uniqueness — special/one-time events ranked higher than recurring
5. Source quality — primary sources and editorial picks ranked higher

Step 9: Present Results

Output a curated list grouped by date:

## [Day, Date]

### [Event Name]
- **What**: [Brief description]
- **Where**: [Venue, Neighborhood] — [travel context from your location]
- **When**: [Time]
- **Category**: Art | Food
- **Link**: [URL]
- **Calendar conflict**: None | "Conflicts with [existing event] at [time]"
- **Add to calendar**: [gcalcli command or Google Calendar link]

---

Google Calendar link format (compact ISO 8601 dates: YYYYMMDDTHHmmssZ):