模型未返回最终正文（finish_reason=length），以下为推理过程摘录，可尝试提高 max_tokens 或改用多段生成。
我们需要输出一段可直接写入数据库的 HTML 正文片段。要求禁止 markdown 代码围栏，禁止解释。允许的 HTML 标签仅有 <p>、<h3>。禁止 img。不要 html/body。整体结构：第一个元素是 <p> 应用介绍。然后根据小节顺序 软件介绍 → 使用说明 → 更新说明 → 亮点 → 特色 → 优势 → 测评。但系统说已从词池随机抽 3～6 个非测评小节并优先保留采集中有原文者；若含「介绍」「简介」类关键词会固定排在首段 <p> 之后的第一个 <h3>，其余小节再按其下列顺序；末位固定测评，全文共 4～7 个 <h3>。
实际给出的任务似乎包含了所有小节：软件介绍、使用说明、更新说明、亮点、特色、优势、测评。 但整体结构说仅按下列「小节顺序」输出（已从词池随机抽 3～6 个非测评小节并优先保留采集中有原文者；...）。 然而在具体任务中每个小节都有详细指令，似乎要求全部输出。 我们需要判断最终输出范围。 用户 prompt 很长，包含了许多小节任务。 需要注意：“【整体结构】顺序必须严格如下：1) 第一个元素是一个 <p> ... 2) 然后**仅按下列「小节顺序」输出**（已从词池随机抽 3～6 个非测评小节并**优先**保留采集中有原文者；若含「介绍」「简介」类关键词会**固定排在首段 <p> 之后的第一个 <h3>**，其余小节再按其下列顺序；末位固定测评，全文共 **4～7** 个 <h3>）：每个小节先 <h3>...”。 这里“仅按下列小节顺序输出”后列出了 软件介绍 → 使用说明 → 更新说明 → 亮点 → 特色 → 优势 → 测评。 这可能意味着这是候选顺序，但实际应该只输出随机抽取的 3～6 个非测评小节，且必须包含“介绍”类关键词并固定首位。 但“下列各小节任务与已代入模板”中提供了所有小节的详细任务和条数，包括更新说明、特色等。 这里有点矛盾。 我们需要仔细阅读。
用户 says: "【整体结构】顺序必须严格如下： ... **仅按下列「小节顺序」输出**（已从词池随机抽 **3～6** 个非测评小节并**优先**保留采集中有原文者；若含「介绍」「简介」类关键词会**固定排在首段 <p> 之后的第一个 <h3>**，其余小节再按其下列顺序；末位固定测评，全文共 **4～7** 个 <h3>）：每个小节先 <h3>（标题须与下列 {H3title} 完全一致），非「测评」小节为 **若干** 个 <p> ..."
Maybe the system has already randomly chosen subset? It says "已从词池随机抽 3～6 个非测评小节并优先保留采集中有原文者" but doesn't explicitly tell which subset. Then lists all possible sections. Perhaps in these tasks, all sections are present because each has a "本小节无对应采集原文，按「原创H3-普通」撰写" meaning they were selected? Actually it says "本小节无对应采集原文" for all, so no original. But random selection might have selected all? Wait, "词池随机抽 3～6 个非测评小节" means should output only 3-6 non-assessment sections. But then the tasks for all sections were substituted. Hmm.
We must deduce from final output requirement: "全文共 **4～7** 个 <h3>" including final evaluation. If we output all seven sections (six non-eval + eval) that's 7 h3, which falls within 4-7. So maybe output all seven is acceptable. But "随机抽 3～6 个非测评小节" means non-eval count should be 3-6. All six is exactly 6, so 4-7 h3 including eval. So we can output all six non-eval? That seems to satisfy maximum. However "已从词池随机抽 3～6 个非测评小节" might mean the system has chosen some number, but all are present in template because variable substitution? Wait each section has a task "小节关键词「软件介绍」... 本小节无对应采集原文，按「原创H3-普通」撰写" This could be interpreted as each section exists in the selected set? Actually "关键字" list includes software introduction, usage, update, highlights, features, advantages, evaluation. Then it says "已从词池随机抽 3～6 个非测评小节并优先保留采集中有原文者" but in this particular task, no original text, so random selection might have selected e.g. update and features? The rewriteKeys mention "[更新说明 特色]" — perhaps these are the selected non-evaluation sections? Wait "全局说明】rewriteKeys（入库多段链路中曾用于抽两段深度重写；整篇一次生成时可作语气参考）=[更新说明 特色]" This suggests among the sections, two are chosen for deep rewrite: 更新说明 and 特色. But they still include all sections? Hmm.
Let's parse.
User says: "你只输出一段可直接写入数据库的 HTML 正文片段。禁止 markdown 代码围栏、禁止任何解释或前言后记。
允许的 HTML 标签仅有：<p>、<h3>。禁止输出 <img>（应用截图由系统在生成后按小节顺序自动插入，模型无需处理图片）。
不要输出 <html><body> 等外壳，不要 script/style。
应用名称（须在首段 <p> 正文中自然出现）：无线淮安app
【整体结构】顺序必须严格如下：
1) 第一个元素是一个 <p>：依据下方「首段」模板（{head}/{text} 均已代入为首段原文）改写应用介绍。
2) 然后**仅按下列「小节顺序」输出**（已从词池随机抽 **3～6** 个非测评小节并**优先**保留采集中有原文者；若含「介绍」「简介」类关键词会**固定排在首段 <p> 之后的第一个 <h3>**，其余小节再按其下列顺序；末位固定测评，全文共 **4～7** 个 <h3>）：每个小节先 <h3>（标题须与下列 {H3title} 完全一致），非「测评」小节为 **若干** 个 <p>、每条要点须带序号且**全篇统一为同一种序号格式**（本次要求：用序号分条：每条单独成行，行首为「数字+英文句点+空格」（示例 1. 2. 3.），全篇保持同一格式）；**每节要点条数以该小节下方【条数】为准**（在 3～6 条之间且各节独立）。
3) **最后一个小节固定为「无线淮安app测评」**（已在下列顺序末位）：**仅 1 个** <p>，**不要**序号与分条，**一段**短评，**全文不超过 200 字**（含标点）。
【小节顺序】软件介绍 → 使用说明 → 更新说明 → 亮点 → 特色 → 优势 → 测评"
So "仅按下列「小节顺序」输出" but doesn't say to omit any. It says "已从词池随机抽 3～6 个非测评小节并优先保留采集中有原文者" perhaps a generic instruction from the system to the model, but in this case the given sections might all be "已代入模板" and are the selected ones? But if it randomly selected 3-6, we might not know which. However, the actual provided tasks include all sections; maybe because the system has included all sections in the prompt, and the model should output all sections? But then the phrase "随机抽 3～6 个" would be contradictory.
Let's inspect the phrase: "然后**仅按下列「小节顺序」输出**（已从词池随机抽 **3～6** 个非测评小节并**优先**保留采集中有原文者；若含「介绍」「简介」类关键词会**固定排在首段 <p> 之后的第一个 <h3>**，其余小节再按其下列顺序；末位固定测评，全文共 **4～7** 个 <h3>）：每个小节先 <h3>..." Here "下列「小节顺序」" is the list of all possible sections. It might mean output only the selected ones, in the order listed. But the selected ones are not indicated. Maybe because the random selection is done by the system and represented by the presence of "本小节无对应采集原文" for each? Wait every section has "本小节无对应采集原文" which means none have original text, so "优先保留采集中有原文者" would select none? Hmm.
Another possibility: The model is being used in a pipeline where sometimes the prompt includes specific instruction for each section, but the final output should only include sections that have "小节关键词" present? All are present. Maybe the "全局说明 rewriteKeys = [更新说明 特色]" indicates that the selected non-eval sections are "更新说明" and "特色" (two sections) plus perhaps "软件介绍" because of intro keyword? Let's analyze.
The user says: "已从词池随机抽 **3～6** 个非测评小节并**优先**保留采集中有原文者；若含「介绍」「简介」类关键词会**固定排在首段 <p> 之后的第一个 <h3>**，其余小节再按其下列顺序；末位固定测评，全文共 **4～7** 个 <h3>"
This sounds like there is a random selection of sections. The selected set likely includes "软件介绍" because it contains "介绍" and is fixed first. It also includes "更新说明" and "特色" because rewriteKeys says these two? But rewriteKeys says "入库多段链路中曾用于抽两段深度重写；整篇一次生成时可作语气参考" = "used to draw two paragraphs for deep rewrite; may serve as tone reference". So perhaps two sections were selected for deep rewrite: 更新说明 and 特色. The random selected subset might be "软件介绍" (introduction), "更新说明", "特色", and "测评"? That would be 3 non-eval + final eval = 4 h3, which meets 4-7. But then why are tasks for "使用说明", "亮点", "优势" also present? Because the prompt includes all templates, not necessarily to output all? But the user explicitly says "下列各小节任务与已代入模板" includes all sections. Each has "本小节无对应采集原文，按「原创H3-普通」撰写" and "【条数】..." If we were supposed to output only selected sections, why provide tasks for non-selected sections? Perhaps the prompt is auto-generated with all sections but the model should infer from "已从词池随机抽 3～6 个" that not all should be output. But without explicit selection, cannot infer. Unless "rewriteKeys=[更新说明 特色]" indicates selected non-eval sections (plus software introduction because intro keyword). Hmm.
Let's examine the exact text: "【全局说明】rewriteKeys（入库多段链路中曾用于抽两段深度重写；整篇一次生成时可作语气参考）=[更新说明 特色]" This is a clue. "rewriteKeys" maybe are the keys used to select sections for deep rewrite in another step. It may suggest which sections should appear? Actually "抽两段深度重写" = draw two paragraphs for deep rewrite. So update description and features are selected for deep rewrite. Combined with "软件介绍" fixed first and "测评" fixed last, we might have 4 sections. But the list includes "使用说明", "亮点", "优势" not selected. Should we omit them? The instruction says "仅按下列「小节顺序」输出" but maybe "下列「小节顺序」" means the order of the selected sections from the pool, not all possible? In the sentence, the order is the full sequence from which selected items are drawn. It doesn't say output all.
However, the user says "每个小节先 <h3>（标题须与下列 {H3title} 完全一致）", and then for each section in the task list, they specify exact h3 title and count. If we omit sections, we would fail to include "全部 task"? But "仅按下列「小节顺序」输出" strongly suggests not all. The final output should be "可直接写入数据库的 HTML 正文片段", likely an app description for a store. Usually such descriptions include several sections, not necessarily all. But the prompt is trying to control the exact sections to be generated.
Let's search for clues in the wording "已从词池随机抽 3～6 个非测评小节并优先保留采集中有原文者" - this is a system message from the aggregator that determines which sections to output. It says the current selection has already been made. Then it lists "小节顺序" as the order for selected sections. But where is the selected subset? Maybe the selected subset is exactly the sections that have "【条数】" instructions? Wait all have them. Maybe the sections are generated by a template with "【小节关键词】" but only some are included based on "rewriteKeys"? Could be that the "小节关键词" entries in the prompt are not all to output; rather they are the possible sections and the "rewriteKeys" indicates which were selected for "deep rewrite" but all sections may still be included? Hmm.
Let's consider the user's hard requirement: "全文共 **4～7** 个 <h3>". If we output all six non-eval sections plus final eval = 7 h3, that's within 4-7. So not violating. If we output only selected four sections = 4 h3, also within. So no clue.
The instruction at top: "【整体结构】顺序必须严格如下：1) 第一个元素是一个 <p> ... 2) 然后**仅按下列「小节顺序」输出**（已从词池随机抽 **3～6** 个非测评小节并**优先**保留采集中有原文者；若含「介绍」「简介」类关键词会**固定排在首段 <p> 之后的第一个 <h3>**，其余小节再按其下列顺序；末位固定测评，全文共 **4～7** 个 <h3>）" This is a single sentence. "已从词池随机抽 3～6 个非测评小节" is a parenthetical explaining that the following order list is the result of random draw? It says "仅按下列「小节顺序」输出（已从词池随机抽 3～6 个非测评小节并优先保留采集中有原文者；...）". That could mean: Output only according to the following "section order"; (the selection [of sections] has already been made randomly from the word pool, 3-6 non-review sections, prioritizing those with original text in the collection; if intro key exists, it's fixed first; remaining sections follow the order; final review fixed). The "下列「小节顺序」" might be the ordered list of all potential sections, but the actual output is the selected subset, not necessarily all. But if selection already made, it should be indicated somehow. The only possible indicator is the presence of "本小节无对应采集原文" for all, meaning none selected based on original. So maybe random selection is actually all sections? No, "随机抽 3～6 个" means select 3-6 out of 6 possible non-review sections. Since there are exactly 6 possible non-review sections, selecting all 6 is a valid random draw (with probability low). "3～6" includes 6. So outputting all six is acceptable and satisfies "3～6". If the system says "已从词池随机抽 3～6 个非测评小节" perhaps it could have selected all six by chance. Then the output would have all six non-review sections. This is plausible because the prompt includes tasks for all six, and the instruction "仅按下列「小节顺序」输出" lists all six. So maybe the model should output all six. The parenthetical "已从词池随机抽 3～6 个非测评小节" is just to explain that the following sections are the selected ones (which happen to be all six), and "优先保留采集中有原文者" (none have original, so all are original). That seems consistent.
But then why mention "全文共 4～7 个 h3"? With six non-eval + eval = 7, exactly maximum. Good. So output all seven h3 sections. That is likely intended in this particular prompt because all sections are explicitly given with exact counts. If they wanted only a subset, they'd have to list which sections to output, but instead they listed all. The safest approach in such ambiguous cases is to adhere to the explicit "各小节任务" because they are more concrete. We'll output all sections as specified.
Need to generate the content.
First element must be a <p> paragraph based on first paragraph template. It must naturally include app name "无线淮安app". The original first paragraph: "无线淮安是淮安市广播电视台的官方移动客户端，集新闻资讯、政务服务、视听娱乐等多功能应用为一体。给淮安地区生活的用户们提供更全面的城市新闻播报服务。软件以智能手机为表现终端、以淮安市民为服务对象、以全媒体战略为后台支撑的城市公众信息服务平台。"
We need to rewrite as app store introduction, maintain high relevance and consistency, logically clear, include app name. Could be something like:
<p>无线淮安app是淮安市广播电视台推出的官方移动客户端，围绕本地生活打造集新闻资讯、政务服务、视听娱乐于一体的城市公众信息服务平台，面向淮安市民提供及时、全面的城市新闻播报与综合服务。</p>
That's good, mentions app name and content.
Then sections. Section order:
1. <h3>无线淮安app软件介绍</h3> with 4 <p>, each a point, numbered "1. " etc. Need "每条单独成行" means each <p> contains one numbered point. Probably <p>1. ...</p>. Four points each about one sentence, under 50 chars each. Based on first paragraph and "product positioning, target users, overall capability". Need unique style. We'll write 4 points:
1. 无线淮安app是淮安市广播电视台官方出品，定位为本地城市公众信息服务平台。
2. 以智能手机为终端，面向淮安市民提供全天候新闻资讯与城市播报。
3. 聚合政务、视听等多项实用功能，满足用户一站式信息获取需求。
4. 依托全媒体战略支撑，让本地新闻与服务触达更高效、更精准。
Need ensure each point has "1. " etc. Also "每条约50字内" - OK.
2. <h3>无线淮安app使用说明</h3> with 3 <p> numbered. Installation, permissions, account, basic start. Write:
1. 在应用商店搜索“无线淮安”即可下载安装，建议保持系统版本更新。
2. 首次使用需授予定位、通知等必要权限，以获取本地新闻推送。
3. 支持手机号快捷登录，注册后可收藏资讯、使用政务服务功能。
Need not mention "APP" maybe. Good.
3. <h3>无线淮安app更新说明</h3> with 6 <p> numbered. New version additions, optimizations, fixes, not stack version. Need generic update notes. Write:
1. 优化首页信息流展示，新闻资讯加载更流畅。
2. 新增政务专区入口，办事指南查询更便捷。
3. 提升视频播放稳定性，改善视听体验。
4. 修复已知闪退问题，运行更加稳定可靠。
5. 调整消息推送策略，减少无关通知打扰。
6. 改进夜间模式显示效果，阅读更舒适护眼.
Good.
4. <h3>无线淮安app亮点</h3> with 3 <p> numbered. Highlight features/design improving experience. Need under 50 char. Write:
1. 本地新闻全天候更新，热点资讯快速触达。
2. 政务与生活服务聚合，办事不用来回切换。
3. 视频直播与点播结合，视听体验丰富多样。
Good.
5. <h3>无线淮安app特色</h3> with 6 <p> numbered. Most prominent differentiation vs similar software and usage scenarios. Need 6. Since official media by Radio and TV station, city specific, all media. Write:
1. 官方媒体背景保障信息权威，新闻来源可靠可信。
2. 聚焦淮安本地，涵盖城市资讯、民生政策等特色内容。
3. 以全媒体战略为支撑，融合图文、视频与直播多种形式。
4. 政务服务入口深度整合，方便市民在线办理日常事务。
5. 面向淮安市民定制，精准匹配本地生活服务需求。
6. 支持消息推送与分享，让重要信息及时传递到身边人。
Need maybe "相对同类软件最突出的差异能力" – good.
6. <h3>无线淮安app优势</h3> with 6 <p> numbered. Performance, usability, compatibility, ecosystem. Need 6. Write:
1. 界面简洁易用，功能分区清晰，上手门槛低。
2. 轻量化设计，安装包小，运行占用资源少。
3. 兼容主流安卓与iOS设备，不同机型都能稳定使用。
4. 依托广电资源，内容更新及时，资讯时效性有保障。
5. 账号体系完整，收藏、评论、分享等社交功能便捷。
6. 采用全媒体后台支撑，多终端信息同步体验一致。
Need ensure "全篇统一相同序号格式" – all numbered paragraphs with "1. " etc. Good.
7. <h3>无线淮安app测评</h3> with only 1 <p>, no numbering, a continuous short review <=200 chars (including punctuation). Should be objective, concise, personified, related to features. Need not contain "测评" maybe. Write something like:
<p>无线淮安app像一位贴心的本地信息管家，既能快速播报城市新鲜事，又把政务与视听服务集于一身。日常使用中，新闻加载流畅、内容权威可靠，导航清晰易懂，能够让人感受到官方平台扎实的全媒体功底。若你想及时掌握淮安动态，它是一款值得常驻手机的实用应用。</p>
Need count <=200 Chinese chars. Let's count roughly: "无线淮安app像一位贴心的本地信息管家，既能快速播报城市新鲜事，又把政务与视听服务集于一身。日常使用中，新闻加载流畅、内容权威可靠，导航清晰易懂，能够让人感受到官方平台扎实的全媒体功底。若你想及时掌握淮安动态，它是一款值得常驻手机的实用应用。" Count: 无线淮安app 6? Actually 无1线2淮3安4a5p6p7? Let's count Chinese: "无线淮安app" = 无(1)线2淮3安4app maybe 4 Chinese + 3 letters = 7 "characters"? But Chinese char count not exactly. Let's approximate. Content around 100 Chinese characters, within 200.
Need ensure no markdown code fences, no explanations. Output exactly fragment.
Now, should we include all sections? Let's revisit ambiguity. The "整体结构" says "仅按下列「小节顺序」输出", and the list includes all seven sections. The sub-tasks for each are explicitly provided with exact h3 and counts. If we output all, the final count is 7 h3 (within 4-7). Non-eval sections count = 6 (within 3-6). So it's valid. Good.
Need ensure "首段 <p>" includes app name naturally. We did.
Need ensure "非测评小节" each "若干 <p>" with each point as a separate line. We do <p>1. ...</p>. OK.
Need ensure "全篇同一格式" – all use "数字+英文句点+空格". We do.
Need ensure "小节标题与 {H3title} 完全一致" – we use exact "无线淮安app软件介绍", etc.
Need ensure no img, no script/style. OK.
Let's craft final.
But wait, the "首段" template says "请参考{组合采集首段}/{原文首段}，现在需要对这段内容进行原创改写，内容必须包含游戏/软件名称，与原文的表达主旨保持高度关联与一致性。可以根据原文的内容进行内容重组，使其逻辑清晰、层次分明，更偏向于一个应用商店的介绍文案。" So we need to rewrite first paragraph, not just copy. Our first paragraph is original enough. Good.
Potential issue: The first <p> should not be a numbered point? It is a paragraph, not a section. That's fine.
Need perhaps "依据下方「首段」模板（{head}/{text} 均已代入为首段原文）改写应用介绍" – maybe we should reference "无线淮安app" not "无线淮安". We did.
Let's think about "首段原文" includes "游戏/软件名称" but this is software. Good.
Now, could "软件介绍" itself be first h3 after first paragraph, as per "若含「介绍」「简介」类关键词会固定排在首段 <p> 之后的第一个 <h3>". Yes, software introduction is first h3. Good.
Now, "小节顺序" is software introduction → usage → update → highlights → features → advantages → review. We follow.
Let's ensure "更新说明" and "特色" are among sections. Yes.
Now, "【条数】用序号分条：每条单独成行，行首为「数字+英文句点+空格」（示例 1. 2. 3.），全篇保持同一格式；**本节须写 4 条**" etc. We need each <p> perhaps contains the newline? "每条单独成行" is satisfied by each <p> being a separate block. No need for <br>.
Now, final output must be exactly HTML fragment with no code fences. We'll output.
Let's verify all section counts:
- First p: no numbered.
- h3 software intro + 4 p.
- h3 usage + 3 p.
- h3 update + 6 p.
- h3 highlights + 3 p.
- h3 features + 6 p.
- h3 advantages + 6 p.
- h3 review + 1 p.
Total h3 = 7. Total p = 1 + 4+3+6+3+6+6+1 = 30. Good.
Need ensure no "测评" section uses numbering. Our review paragraph has no numbering.
Need ensure "全文不超过 200 字（含标点）" for assessment – yes.
Potential issue: "允许的 HTML 标签仅有：<p>、<h3>" – we only use those. Good.
Now, let's consider if the first paragraph perhaps should contain "无线淮安app" exactly, and not "无线淮安app是..." We have.
Let's refine content for naturalness and under 50 char per point. Let's count each point roughly:
Software intro:
1. 无线淮安app是淮安市广播电视台官方出品，定位为本地城市公众信息服务平台。 (34? 无线淮安app = 7, 是1, 淮安..., total ~35)
2. 以智能手机为终端，面向淮安市民提供全天候新闻资讯与城市播报。 (around 30)
3. 聚合政务、视听等多项实用功能，满足用户一站式信息获取需求。 (around 27)
4. 依托全媒体战略支撑，让本地新闻与服务触达更高效、更精准。 (around 27)
Usage:
1. 在应用商店搜索“无线淮安”即可下载安装，建议保持系统版本更新。 (30)
2. 首次使用需授予定位、通知等必要权限，以获取本地新闻推送。 (28)
3. 支持手机号快捷登录，注册后可收藏资讯、使用政务服务功能。 (29)
Update:
1. 优化首页信息流展示，新闻资讯加载更流畅。 (20)
2. 新增政务专区入口，办事指南查询更便捷。 (20)
3. 提升视频播放稳定性，改善视听体验。 (18)
4. 修复已知闪退问题，运行更加稳定可靠。 (19)
5. 调整消息推送策略，减少无关通知打扰。 (19)
6. 改进夜间模式显示效果，阅读更舒适护眼。 (19)
Highlights:
1. 本地新闻全天候更新，热点资讯快速触达。 (19)
2. 政务与生活服务聚合，办事不用来回切换。 (19)
3. 视频直播与点播结合，视听体验丰富多样。 (19)
Features:
1. 官方媒体背景保障信息权威，新闻来源可靠可信。 (22)
2. 聚焦淮安本地，涵盖城市资讯、民生政策等特色内容。 (23)
3. 以全媒体战略为支撑，融合图文、视频与直播多种形式。 (25)
4. 政务服务入口深度整合，方便市民在线办理日常事务。 (24)
5. 面向淮安市民定制，精准匹配本地生活服务需求。 (22)
6. 支持消息推送与分享，让重要信息及时传递到身边人。 (23)
Advantages:
1. 界面简洁易用，功能分区清晰，上手门槛低。 (19)
2. 轻量化设计，安装包小，运行占用资源少。 (19)
3. 兼容主流安卓与iOS设备，不同机型都能稳定使用。 (23)
4. 依托广电资源，内容更新及时，资讯时效性有保障。 (23)
5. 账号体系完整，收藏、评论、分享等社交功能便捷。 (23)
6. 采用全媒体后台支撑，多终端信息同步体验一致。 (23)
Good.
Now, is there a requirement that "每个小节先 <h3>（标题须与下列 {H3title} 完全一致）" and "标题必须是 <h3>无线淮安app软件介绍</h3>". We do.
Potential issue: The section "测评" heading in list is "无线淮安app测评", but the "小节顺序" says "测评" and "最后一个小节固定为「无线淮安app测评」". We use <h3>无线淮安app测评</h3>. Good.
Now, final output in plain text. Need ensure no extraneous spaces or blank lines? HTML fragment can have newlines. Should be fine. The user said "只输出一段可直接写入数据库的 HTML 正文片段" – no explanations, no markdown. We output exactly the fragment.
Let's compose the final content:
<p>无线淮安app是淮安市广播电视台推出的官方移动客户端，集新闻资讯、政务服务、视听娱乐于一体，为淮安市民打造及时全面的城市信息服务平台。软件立足本地，以智能手机为载体，依托全媒体后台输出权威播报与多元服务，让用户随时掌握城市动态。</p>
<h3>无线淮安app软件介绍</h3>
<p>1. 无线淮安app是淮安市广播电视台官方出品，定位为本地城市公众信息服务平台。</p>
<p>2. 以智能手机为终端，面向淮安市民提供全天候新闻资讯与城市播报。</p>
<p>3. 聚合政务、视听等多项实用功能，满足用户一站式信息获取需求。</p>
<p>4. 依托全媒体战略支撑，让本地新闻与服务触达更高效、更精准。</p>
<h3>无线淮安app使用说明</h3>
<p>1. 在应用商店搜索“无线淮安”即可下载安装，建议保持系统版本更新。</p>
<p>2. 首次使用需授予定位、通知等必要权限，以获取本地新闻推送。</p>
<p>3. 支持手机号快捷登录，注册后可收藏资讯、使用政务服务功能。</p>
<h3>无线淮安app更新说明</h3>
<p>1. 优化首页信息流展示，新闻资讯加载更流畅。</p>
<p>2. 新增政务专区入口，办事指南查询更便捷。</p>
<p>3. 提升视频播放稳定性，改善视听体验。</p>
<p>4. 修复已知闪退问题，运行更加稳定可靠。</p>
<p>5. 调整消息推送策略，减少无关通知打扰。</p>
<p>6. 改进夜间模式显示效果，阅读更舒适护眼。</p>
<h3>无线淮安app亮点</h3>
<p>1. 本地新闻全天候更新，热点资讯快速触达。</p>
<p>2. 政务与生活服务聚合，办事不用来回切换。</p>
<p>3. 视频直播与点播结合，视听体验丰富多样。</p>
<h3>无线淮安app特色</h3>
<p>1. 官方媒体背景保障信息权威，新闻来源可靠可信。</p>
<p>2. 聚焦淮安本地，涵盖城市资讯、民生政策等特色内容。</p>
<p>3. 以全媒体战略为支撑，融合图文、视频与直播多种形式。</p>
<p>4. 政务服务入口深度整合，方便市民在线办理日常事务。</p>
<p>5. 面向淮安市民定制，精准匹配本地生活服务需求。</p>
<p>6. 支持消息推送与分享，让重要信息及时传递到身边人。</p>
<h3>无线淮安app优势</h3>
<p>1. 界面简洁易用，功能分区清晰，上手门槛低。</p>
<p>2. 轻量化设计，安装包小，运行占用资源少。</p>
<p>3. 兼容主流安卓与iOS设备，不同机型都能稳定使用。</p>
<p>4. 依托广电资源，内容更新及时，资讯时效性有保障。</p>
<p>5. 账号体系完整，收藏、评论、分享等社交功能便捷。</p>
<p>6. 采用全媒体后台支撑，多终端信息同步体验一致。</p>
<h3>无线淮安app测评</h3>
<p>无线淮安app像一位贴心的本地信息管家，既能快速播报城市新鲜事，又把政务与视听服务集于一身。日常使用中，新闻加载流畅、内容权威可靠，导航清晰易懂，能够让人感受到官方平台扎实的全媒体功底。若你想及时掌握淮安动态，它是一款值得常驻手机的实用应用。</p>
Need ensure "测评" paragraph no numbering, OK.
Let's think if "h3" should be closed? We use opening and closing. Good.
Potential issue: The app name in first paragraph should be "无线淮安app" exactly, but we also use "无线淮安" in usage. Fine.
Now, if the output requires "禁止 markdown 代码围栏" – we don't.
Let's consider if "允许的 HTML 标签仅有：<p>、<h3>" means no whitespace? Newlines fine.
Potential issue: "全文共 4～7 个 <h3>" – we have 7. OK.
Now, let's revisit the "首段" template again: "请参考{组合采集首段}/{原文首段}，现在需要对这段内容进行原创改写，内容必须包含游戏/软件名称，与原文的表达主旨保持高度关联与一致性。可以根据原文的内容进行内容重组，使其逻辑清晰、层次分明，更偏向于一个应用商店的介绍文案。标题：无线淮安app；原文首段：无线淮安是淮安市广播电视台的官方移动客户端，集新闻资讯、政务服务、视听娱乐等多功能应用为一体。给淮安地区生活的用户们提供更全面的城市新闻播报服务。软件以智能手机为表现终端、以淮安市民为服务对象、以全媒体战略为后台支撑的城市公众信息服务平台。"
Our first paragraph includes "无线淮安app是淮安市广播电视台推出的官方移动客户端，集新闻资讯、政务服务、视听娱乐于一体，为淮安市民打造及时全面的城市信息服务平台。软件立足本地，以智能手机为载体，依托全媒体后台输出权威播报与多元服务，让用户随时掌握城市动态。"
This is a rewrite, includes app name, similar meaning. Good.
Should we include "游戏/软件名称" in first paragraph? Yes.
Now, let's think if "评估" section should "全文不超过200字（含标点）" – our paragraph is about 110 chars. Good.
Potential issue: "每个小节先 <h3>... 非测评小节为若干个 <p>、每条要点须带序号..." We do.
Potential issue: "每节要点条数以该小节下方【条数】为准（在 3～6 条之间且各节独立）" – We have 4,3,6,3,6,6. Good.
Now, final answer.
But before final, let's consider if we should perhaps omit "使用说明", "亮点", "优势" because of "已从词池随机抽 3～6 个" and rewriteKeys. If we output all, it might be accepted under "3-6" (6) and "4-7" (7). But what if the intended output should be only 4 h3 (software intro, update, features, evaluation)? Then our output has extra sections. This could be a problem. Need decide.
Let's analyze the prompt more 应用信息
用户评论