InContext — 划词释义（基于全文理解）

划选网页中的任意一段，基于大模型对整篇文章的理解给出解释，并可就这段多轮追问。Explain any selection — grounded in the whole page, not just the snippet.

Most "explain" or "translate selection" extensions send only the highlighted snippet to the AI, so the explanation is out of context. InContext is different: it first extracts the WHOLE article (via Mozilla Readability) and feeds it to the model as grounding. So it explains what your selection actually means IN THIS article — the author's intent, the implied meaning, and what each term refers to here. ★ Whole‑page grounding — the entire article is the context, not the isolated snippet. ★ Three‑step explanations — term definition → meaning in context → example when helpful. ★ Multi‑turn — keep asking follow‑up questions about the same passage. ★ Any language — explain in Chinese, English, Japanese, and more. ★ Cited web search (optional) — a search engine fetches live facts and your model writes the answer with [n] cited sources. Context‑aware, bilingual queries that disambiguate homonyms. ★ Works with any OpenAI-compatible model service, plus native Anthropic — Bring Your Own Key. Keys are stored locally in your browser and never sent to any third party. ★ Per‑tab conversations — each tab keeps its own thread; switching tabs never interrupts a running explanation. Privacy: InContext has no backend. Your API keys and conversations stay in your browser. Page text is sent only to the model endpoint you configure yourself. How to use: select text → click the floating Explain button (or right‑click / Ctrl+Shift+E) → read the grounded explanation in the side panel → ask follow‑ups. Toggle Web for cited, up‑to‑date answers. Open source (MIT): https://github.com/qucooln/incontext --- 普通划词翻译/解释只把选中的一小段发给 AI，解释脱离语境。 InContext 不一样：先用 Readability 抽取【整篇文章】喂给模型做背景，所以它解释的是这段话【在这篇文章里】到底什么意思——作者意图、隐含信息、专有名词的真实所指。 ★ 基于全文，而非孤立片段 ★ 三步式解释：术语释义 → 全文语境 → 必要时举例 ★ 多轮追问、多语种输出 ★ 可选联网：搜索引擎取数、你的模型作答并标注 [n] 来源；检索词结合上下文、中英双语、自动消歧 ★ 兼容任意 OpenAI 格式的大模型服务，并支持 Anthropic 原生接口，自带 key（BYOK）；key 仅存本地，不传第三方 ★ 按标签页独立对话，切换不打断 隐私：无后端，key 与对话只存在你浏览器，正文只发往你自己配置的模型接口。 开源(MIT)：https://github.com/qucooln/incontext