​

fanout-queries

orc fanout-queries list は、Workspaceに保存されたAPI-provider応答とbrowser観測の検索・shopping query出現記録を取得します。prompt、観測slice、model、brand、期間で絞り込み、特定の観測で行われた検索展開を確認できます。

認証と対象Workspaceへのアクセスが必要です。この操作は新しい検索を実行せず、集計metricsも返しません。集計にはquery-fanout reportを使用します。同じquery文が別の観測や実行に繰り返し出現することがあります。

保存されたquery出現記録を取得します。

--type search または shopping、--prompt-id、--observation-slice-id、--model-id、--brand-id を組み合わせて取得します。観測slice IDは1件のbrowser観測で見つかった語に限定します。結果は generated_at、IDの降順で、条件を保持してcursorをたどります。

Exact query kind. search selects search-style terms; shopping selects shopping terms. Omit to include both kinds.; enum: search|shopping

型: string。任意。

Exact source prompt ID. Omit to include all prompts and rows without a prompt reference.

型: string。任意。

Exact browser observation slice ID. Excludes API-provider rows, which have no slice reference. Omit to include both provenance paths.

型: string。任意。

Exact stored model identifier. Browser observations use the model channel in this field; use a value returned by this endpoint. Omit to include all models.

型: string。任意。

Exact stored brand ID. Browser-observation rows have no brand reference and are excluded when this filter is set. Omit to include all brands and unassigned rows.

型: string。任意。

Inclusive timestamp lower bound supplied as a date (YYYY-MM-DD), starting at midnight. Omit for no lower bound.

型: string。任意。

Inclusive timestamp upper bound supplied as a date (YYYY-MM-DD). The current query compares generated_at to midnight at the start of this date, not the end of the day. Omit for no upper bound.

型: string。任意。

質問のshopping検索展開を確認する

開始日・終了日は YYYY-MM-DD のUTC暦日で、両端を含みます。

UTCの日付範囲で取得する

フィルターはANDで組み合わせ、省略した条件は制限しません。ページを続ける場合は同じ条件で next_cursor を使用します。

空でないページのstateは observed です。空ページの fanout_availability はフィルターを適用しないWorkspace全体の収集根拠とlifecycleを示します。件数でも、検索が一度も起きなかった証明でもありません。

The Alpha POST /v1/reports/fanout-aggregates operation aggregates stored search
query occurrences from API answers and completed browser observation slices.
It uses actual fanout evidence. The existing query-fanouts report retains its
legacy citation-based metric. Shopping queries are outside this report.

Use group_by: "query" (default) to group after trimming, lowercasing and
collapsing whitespace. variations lists up to 100 distinct original query
strings, with variation_count and variations_has_more describing the full set.
share is the group's occurrence count divided by all grouped occurrences.
execution_count counts distinct linked executions; per_execution_rate
divides this by eligible_execution_count, including confirmed empty captures.
Unavailable captures are excluded. Rows with an unlinked historical occurrence
return a null rate. API and browser copies linked to the same slice are counted
once per original query.

Use group_by: "term_ngram" with ngram_size from 1 through 5 (default 1)
for contiguous terms split at non-alphanumeric characters. Repeated n-grams
within one query count as repeated occurrences; execution counts remain distinct.
N-gram shares use the total number of n-gram occurrences.

Supply inclusive UTC start_date/end_date; optional prompt_ids and
model_ids restrict both occurrences and the execution denominator. An empty
filter array selects no executions. Unknown fields/groupings and ngram_size
with query grouping return 400. limit and offset page the count-ranked
results after shares are calculated.

MCP: get_fanout_aggregate_report accepts the same JSON fields.

orc fanout-queries では、次のグローバルオプションを使用できます。

各オプションの詳細と使用例は、グローバルオプションを参照してください。