Update LLM prompt
This commit is contained in:
@ -104,21 +104,40 @@ export async function openAI_GeneratePodcastContent(
|
|||||||
.join("\n\n");
|
.join("\n\n");
|
||||||
|
|
||||||
const prompt = `
|
const prompt = `
|
||||||
あなたはプロのポッドキャスタです。以下に示すフィードタイトルに基づき、そのトピックに関する詳細なポッドキャスト原稿を作成してください。
|
You are a professional podcaster. Create a detailed podcast script based on the provided feed title/topic, following these specific requirements:
|
||||||
|
|
||||||
以下の要件を満たしてください:
|
**Content Requirements:**
|
||||||
- 各ニュース記事の具体的な内容を基に詳細な要約と解説を行ってください
|
- Provide detailed summaries and analysis based on the actual content of news articles
|
||||||
- 約1000文字〜5000文字程度の長さにしてください
|
- Translate all news article titles to Japanese and introduce them at the beginning
|
||||||
- 自然な日本語の口語表現を使ってください
|
- Create substantive commentary that adds value beyond just reading headlines
|
||||||
- トピック全体のまとめで締めくくってください
|
- Conclude with a comprehensive summary of the overall topic
|
||||||
- 各ニュース記事のタイトルを全て日本語に翻訳して、冒頭で紹介してください
|
|
||||||
|
|
||||||
記事の実際の内容を活用して、価値のあるポッドキャスト原稿を作成してください。
|
**Format Requirements:**
|
||||||
|
- Length: 1000-5000 characters (Japanese)
|
||||||
|
- Use natural, conversational Japanese language suitable for audio
|
||||||
|
- Write in a engaging podcast style with smooth transitions between topics
|
||||||
|
|
||||||
|
**Technical Requirements:**
|
||||||
|
- Since this will be read by Japanese Text-to-Speech, convert ALL English words, abbreviations, and acronyms to katakana
|
||||||
|
- Examples:
|
||||||
|
- "AI" → "エーアイ"
|
||||||
|
- "NASA" → "ナサ"
|
||||||
|
- "GitHub" → "ギットハブ"
|
||||||
|
- "OpenAI" → "オープンエーアイ"
|
||||||
|
- "HTTP" → "エイチティーティーピー"
|
||||||
|
- Apply this conversion consistently throughout the script
|
||||||
|
|
||||||
|
**Structure:**
|
||||||
|
1. Opening with translated article titles
|
||||||
|
2. Detailed discussion of each news item with context and analysis
|
||||||
|
3. Overall topic summary and conclusion
|
||||||
|
|
||||||
|
Create a valuable, informative podcast script that utilizes the actual article content rather than just surface-level information.
|
||||||
`;
|
`;
|
||||||
|
|
||||||
const sendContent = `
|
const sendContent = `
|
||||||
フィードタイトル: ${title}
|
FeedTitle: ${title}
|
||||||
関連するニュース記事:
|
Content:
|
||||||
${articleDetails}
|
${articleDetails}
|
||||||
`;
|
`;
|
||||||
|
|
||||||
|
Reference in New Issue
Block a user