Function handler

packages/tools/src/definitions/keyword-frequency.ts:23–74

(params: Record<string, unknown>, context: ToolExecutionContext)

}

async function handler(params: Record<string, unknown>, context: ToolExecutionContext): Promise<ToolResult> {
  const { locale, segmentText } = context
  const isZh = isChineseLocale(locale)
  const days = (params.days as number) || 30
  const topN = (params.top_n as number) || 50

  if (!segmentText) {
    const text = isZh ? '当前环境不支持分词功能' : 'Text segmentation is not available in this environment'
    return { content: text, data: null }
  }

  const sql = `
    SELECT content FROM message
    WHERE type = 0 AND content IS NOT NULL AND LENGTH(content) > 1
      AND ts > unixepoch('now', '-' || @days || ' days')
    LIMIT 50000
  `
  const rows = await context.dataProvider!.executeParameterizedSql<TextRow>(sql, { days })
  if (!rows || rows.length === 0) {
    const text = isZh ? '该时间范围内没有文本消息' : 'No text messages in this time range'
    return { content: text, data: null }
  }

  const texts = rows.map((r) => r.content)
  const segLocale: string = locale?.startsWith('ja') ? 'ja-JP' : locale?.startsWith('zh') ? 'zh-CN' : 'en-US'
  const freqResult = segmentText(texts, segLocale, {
    minCount: 2,
    topN,
    posFilterMode: 'meaningful',
    enableStopwords: true,
  })

  if (freqResult.words.size === 0) {
    const text = isZh ? '分词后没有有意义的高频词' : 'No meaningful high-frequency words found after segmentation'
    return { content: text, data: null }
  }

  const ranking = [...freqResult.words.entries()].map(([word, count], i) => ({
    rank: i + 1,
    word,
    count,
  }))

  const data = {
    period: isZh ? `近${days}天` : `Last ${days} days`,
    totalMessages: rows.length,
    totalKeywords: ranking.length,
    keywords: ranking.map((r) => `${r.rank}. ${r.word} (${r.count}${isZh ? '次' : ''})`),
  }

  return { content: JSON.stringify(data), data }
}

export const keywordFrequencyTool: ToolDefinition = {
  name: 'keyword_frequency',