AI 分析
痛点:在旧合同、发票、笔记等本地文件里定位某句话极其费时,而云端聊天机器人要求上传全部文件且无法显示答案出处,隐私与可信度都无法接受。
目标用户:处理大量敏感本地文档的 Mac 用户:律师、财务/会计、咨询顾问、独立研究者、产品与法务岗
证据:I kept losing time hunting for one line across old contracts, invoices and notes. Chatbots could answer, but I had to upload everything and then just trust them, because I couldn't see where an answer came from / The hardest part was making the answers point at the right sentence, not just the right file / It's free for 30 days with every feature, no card. After that it's a subscription from $9.99 a month.
变现:订阅制($9.99/月,可加年付折扣、Pro/团队版),或一次性买断 + 大版本升级收费;后续可扩展为企业私有部署许可
MVP:2 周内用 Swift/SwiftUI + MLX 或 Ollama 做最小版:只索引指定文件夹的 PDF 和 Markdown,本地嵌入检索 + 本地小模型生成答案,强制返回可点击的原文句子高亮(句级而非文件级溯源),仅支持 Apple Silicon;先不做 Apple Notes/Notion/YouTube 等连接器。
竞品:NotebookLM、Rewind、Msty、AnythingLLM、DEVONthink、Pieces、Apple Intelligence / Siri、Recall
摘要:开发者发布 Citenook,一款 macOS 原生本地文件问答应用,可索引 PDF、Word、表格、Apple Notes、Obsidian、Notion、网页和 YouTube,每个答案带编号并可点击跳转到原文句子高亮,用本地模型可断网运行。免费 30 天,之后订阅 $9.99/月起,目前仅支持 macOS 15+ 与 Apple 芯片。痛点源自作者在旧合同、发票和笔记中找一句话耗时,且不愿把文件上传云端。