Mio 给你 25 个可以连接的角色。Lumi 给你一个完全理解你的存在。它们满足不同的情感需求，但共享同一套记忆管线、情绪引擎和语音链。一个 monorepo 怎么装下两个陪伴产品——以及 agentId?: string 这个模式教会了我什么。

Mio gives you 25 personas to connect with. Lumi gives you one presence that knows you completely. They serve different emotional needs but share the same memory pipeline, emotion engine, and voice chain. Here's how a monorepo houses two companion products — and what the agentId?: string pattern teaches about building parallel products.

识川本来只做珠宝。一个问题——'为什么不是全品类全链路？'——把它变成了 7 个品类的电商内容平台：AI 图片、小红书文字卡、平台适配裁剪、客户端视频生成，99 元/套，95%+ 毛利。

Shichuan started as a jewelry content generator. One question — 'Why not full pipeline for all categories?' — turned it into a 7-domain e-commerce content platform with AI images, XHS text cards, platform-aware formatting, client-side video, and 95%+ margins at 99 RMB per suite.

大多数 TTS 给你一个声音。豆包说它能给你情感。我跑了一轮系统性实验——30 个音频样本，跨越预置音色、克隆音色、3 种情感控制方法——来验证到底什么管用。自己听。

Most TTS gives you a voice. Doubao claims to give you emotions. I ran a systematic experiment — 30 audio samples across stock voices, cloned voices, and 3 emotion control methods — to find out what actually works. Listen for yourself.

火山引擎 Seed ASR 大模型完整接入指南——提交-轮询 REST API、中文语音转文字、链式降级模式、成本追踪，以及 8 个 v2/v3 API 差异的坑。

Complete integration guide for Volcengine Seed ASR bigmodel — the submit-then-poll REST API for Chinese speech-to-text. Console setup, two-step async flow, fallback chain pattern, cost tracking, and 8 gotchas about v2 vs v3 APIs that will save you a day of debugging.

火山引擎豆包 Seed-ICL 2.0 完整接入指南——声音克隆、context_texts 自然语言情感控制、逐句多次调用合成、NDJSON 响应解析，以及 7 个能帮你省几小时 debug 的坑。

Complete integration guide for Volcengine Doubao Seed-ICL 2.0 — voice cloning, natural language emotion control via context_texts, per-sentence multi-call synthesis, NDJSON response parsing, and 7 gotchas that will save you hours of debugging.

拆解 ÉLAN 的 10 段式 prompt 系统、让奢侈品「不经意」出现的 VANITY_DESIGN_INSTRUCTIONS、SSE 流式推送架构、三种风格的配文生成，以及真实成本结构——输出图片的 token 占了绝大部分开销。

A deep dive into ÉLAN's 10-section prompt system, the VANITY_DESIGN_INSTRUCTIONS that make luxury look incidental, SSE streaming architecture, caption generation with 3 style templates, and the real cost breakdown — output image tokens dominate virtually all of the per-session cost.

ÉLAN 的目标用户拍照在手机上，发帖在手机上。纯网页版错过了核心场景。这篇记录了把 Next.js 应用搬到 React Native (Expo) 的真实过程——哪些能复用，哪些不能，以及一个 SSE 流式传输的坑是怎么填的。

ÉLAN's target user takes photos on her phone and posts from her phone. A web-only product misses the core use case. Here's what actually happened when I brought a Next.js app to React Native via Expo — what transferred, what didn't, and the SSE streaming hack that saved the project.

朋友说'做云佩戴'。调研 20 分钟就否掉了。但同一套 Gemini 管线换个 prompt，就能给珠宝商生成编辑级产品内容——棚拍主图、宝石星图、色彩基因页——成本是请摄影师的百分之一。一天后，识川上线了。

A friend pitched 'cloud try-on for jewelry.' Deep research killed the plan in 20 minutes. But the same Gemini pipeline from ÉLAN could generate editorial-quality product content — studio shots, material constellations, color DNA pages — for a fraction of what merchants pay photographers. One day later, Shichuan was live.

重建不等于全部重写。Mio 核心的 60-70%——记忆系统、媒体管线、成本追踪——直接复用。数据库从 10 张表砍到 4 张。哪些活了，哪些死了，为什么。

Rebuilding doesn't mean rewriting everything. 60-70% of Mio's core — memory, media pipeline, cost tracking — carries over. The database drops from 10 tables to 4. Here's what lives, what dies, and why.