news headlines derived

{
  "input": {
    "type": "http",
    "method": "GET",
    "queryParams": {
      "lang": "ISO-639-1 language code.",
      "limit": "How many articles to return.",
      "source": "Restrict to one ingestion source. gnews is retired (archived articles only).",
      "category": "Restrict to one topical category."
    }
  },
  "output": {
    "type": "object",
    "required": [
      "updated_at",
      "window_hours",
      "count",
      "articles",
      "freshness"
    ],
    "properties": {
      "count": {
        "type": "integer"
      },
      "articles": {
        "type": "array",
        "items": {
          "type": "object",
          "required": [
            "id",
            "title",
            "url",
            "domain",
            "source",
            "feed",
            "published",
            "categories",
            "seen_at"
          ],
          "properties": {
            "id": {
              "type": "string",
              "description": "sha1(normalized url) hex, first 16 chars — the dedupe key."
            },
            "url": {
              "type": "string"
            },
            "feed": {
              "type": "string",
              "description": "Human feed label, e.g. \"BBC World\", \"Hacker News\", \"GDELT: technology\"."
            },
            "lang": {
              "type": [
                "string",
                "null"
              ],
              "description": "ISO-639-1 language, best-effort."
            },
            "tier": {
              "enum": [
                "official",
                "major",
                "trade",
                "social",
                "enrichment",
                null
              ],
              "type": [
                "string",
                "null"
              ],
              "description": "Reputability tier of the feed: official (Fed/SEC/White House/BLS/ECB press), major (wires + major outlets), trade (quality trade press), social (Hacker News), enrichment (GDELT-discovered, reputation-gated). Absent on records archived before the tiered portfolio."
            },
            "tone": {
              "type": [
                "number",
                "null"
              ],
              "description": "GDELT document tone (negative = adverse), attached by URL match from GDELT's 15-minute files — present on any source GDELT also saw; null otherwise."
            },
            "title": {
              "type": "string"
            },
            "domain": {
              "type": "string"
            },
            "source": {
              "enum": [
                "gdelt",
                "hn",
                "rss",
                "gnews",
                "sitemap"
              ],
              "type": "string"
            },
            "cluster": {
              "type": [
                "string",
                "null"
              ],
              "description": "Story-cluster id — articles sharing one are the SAME story from different outlets. The served window already keeps only one article per cluster, so this is provenance, not a filter you must apply. null when the headline was too short to cluster safely."
            },
            "country": {
              "type": [
                "string",
                "null"
              ]
            },
            "seen_at": {
              "type": "string"
            },
            "snippet": {
              "type": [
                "string",
                "null"
              ],
              "description": "Short plain-text description from the feed (≤240 chars); absent when the feed carries none."
            },
            "published": {
              "type": "string",
              "description": "ISO-8601 publish time (best-effort)."
            },
            "categories": {
              "type": "array",
              "items": {
                "type": "string"
              }
            }
          }
        }
      },
      "freshness": {
        "type": "object",
        "required": [
          "as_of",
          "window_hours",
          "source_count",
          "coverage"
        ],
        "properties": {
          "as_of": {
            "type": "string",
            "description": "updated_at of the corpus this response was served from."
          },
          "coverage": {
            "type": "object",
            "required": [
              "oldest_published",
              "newest_published",
              "hours",
              "articles"
            ],
            "properties": {
              "hours": {
                "type": "number",
                "description": "newest − oldest in hours — the searchable depth, usually far below window_hours."
              },
              "articles": {
                "type": "integer",
                "description": "Articles covering that span."
              },
              "newest_published": {
                "type": [
                  "string",
                  "null"
                ],
                "description": "Newest publish time held."
              },
              "oldest_published": {
                "type": [
                  "string",
                  "null"
                ],
                "description": "Oldest publish time held; the real floor of any date filter."
              }
            }
          },
          "source_count": {
            "type": "integer",
            "description": "Number of sources with ≥1 article in the current pile."
          },
          "window_hours": {
            "type": "integer",
            "description": "Retention POLICY in hours (72). NOT the searchable depth — see coverage.hours."
          }
        }
      },
      "updated_at": {
        "type": "string"
      },
      "window_hours": {
        "type": "integer"
      }
    }
  }
}