Daily News Digest

# Lobster workflow: daily-news-digest
# Aggregates news from HN front page + web search into a morning briefing
#
# Usage:
#   lobster run --file workflow.yaml --args-json '{"topics":"AI, startups, technology"}'
#
# Requires: jq, curl
# LLM: Uses prompt step (llm:host by default — uses your agent's model)

name: daily-news-digest
description: Aggregate and summarize news from multiple sources into a morning briefing

args:
  topics:
    description: "Topics to track (comma-separated)"
    default: "AI, startups, technology"
  max_articles:
    description: "Maximum articles to include"
    default: "20"

steps:
  - id: fetch-hn
    command: |
      curl -sf "https://hacker-news.firebaseio.com/v0/topstories.json" \
        | jq '.[0:10]' > /tmp/lb_hn_ids.json
      echo '[]' > /tmp/lb_hn_articles.json
      for id in $(jq -r '.[]' /tmp/lb_hn_ids.json); do
        curl -sf "https://hacker-news.firebaseio.com/v0/item/$id.json" \
          | jq -c '{title:.title, url:(.url//"https://news.ycombinator.com/item?id=\(.id)"), source:"HN", score:.score}' \
          >> /tmp/lb_hn_each.jsonl 2>/dev/null || true
      done
      if [ -f /tmp/lb_hn_each.jsonl ]; then
        jq -sc '.' /tmp/lb_hn_each.jsonl
      else
        echo '[]'
      fi
      rm -f /tmp/lb_hn_ids.json /tmp/lb_hn_each.jsonl /tmp/lb_hn_articles.json

  - id: search-topics
    command: |
      # Use Brave Search API if BRAVE_API_KEY is set, otherwise skip
      query=$(echo "${topics}" | tr ',' ' ' | sed 's/  */ OR /g')
      if [ -n "$BRAVE_API_KEY" ]; then
        curl -sf "https://api.search.brave.com/res/v1/web/search?q=$(echo "$query" | jq -sRr @uri)&count=10&freshness=pd" \
          -H "Accept: application/json" \
          -H "X-Subscription-Token: $BRAVE_API_KEY" \
          | jq -c '[.web.results[]? | {title:.title, url:.url, source:"Web", score:0}]' 2>/dev/null || echo '[]'
      else
        echo '[]'
      fi

  - id: combine-and-dedup
    command: |
      hn=$(cat <<'HNEOF'
      $fetch-hn.stdout
      HNEOF
      )
      search=$(cat <<'SEARCHEOF'
      $search-topics.stdout
      SEARCHEOF
      )
      echo "$hn" > /tmp/lb_nd_hn.json
      echo "$search" > /tmp/lb_nd_search.json
      max=${max_articles}
      jq -sc --argjson max "$max" '
        .[0] + .[1] |
        unique_by(.title | ascii_downcase | .[0:50]) |
        sort_by(-.score) |
        .[0:$max]
      ' /tmp/lb_nd_hn.json /tmp/lb_nd_search.json
      rm -f /tmp/lb_nd_hn.json /tmp/lb_nd_search.json

  - id: summarize-digest
    command: |
      llm_task.invoke --prompt "SYSTEM: You are a morning news curator. Write concise, scannable digests that help busy people stay informed. Group by theme, highlight what matters, skip the fluff.\n\nUSER: Create a morning news digest from the articles below.\nTopics of interest: ${topics}.\n\nGroup articles by theme. For each group, write a 2-3 sentence summary\nof why it matters. Include the article title and URL for each item.\n\nFormat as a readable briefing with emoji section headers.\nOpen with today's date and a greeting. Keep it concise but informative.\n\nArticles:\n$(cat /dev/stdin)"
    stdin: $combine-and-dedup.stdout
    env:
      CLAWD_URL: "http://127.0.0.1:3000"