We need to produce JSON with fields: seo_title, summary, key_entities (companies, people, locations), expert_faq (question, answer), internal_link_anchor, primary_keyword, secondary_keywords, content_type.

We must write summary 180-250 words, split into 3-4 paragraphs, each 45-65 words. Use bold formatting selectively: 3-5 bolded terms across entire summary. Must not invent facts; only use what is in source or implied. Source: "China kicked off marketing up to €5 billion ($5.7 billion) of sovereign bonds in what could be its largest-ever such deal in euros."

Thus we can elaborate: context: China's sovereign bond issuance in euros, record size, part of diversification, maybe to tap European investor base, etc. But we must not invent facts not in source or implied. We can infer that it's a sovereign bond deal, that it's marketing up to €5bn, could be largest-ever such deal in euros. We can say that it reflects China's efforts to expand foreign currency funding, etc. But we must be careful not to add info not implied. However, we can add generic context about sovereign bond markets, but must not state specifics not in source. The instruction: "If thin on details, expand with relevant context from the source material. Rephrase and elaborate on what is provided, but do NOT add information that is not in or directly implied by the source." So we can elaborate on what is provided: that China is marketing up to €5bn sovereign bonds, that it could be largest-ever such deal in euros. We can discuss implications: impact on euro-denominated debt market, investor demand, pricing, etc. But those are not directly in source; they are implied? Possibly we can say that such a large issuance could affect yields, attract global investors, etc. That's reasonable inference but not explicit. The instruction says "do NOT add information that is not in or directly implied by the source." Implications could be considered directly implied? Might be okay if we stay generic: "The move signals