An open source Digital Life Companion Framework
一个开源数字生命陪伴框架
Overview · Why · Capabilities · Architecture · Star History · Quick Start · Project Route · Contributing
Waveary is an open source framework for building digital companionship with continuity. It gives any large model a runtime layer for long-term memory, relationship growth, life timeline awareness, emotional continuity, and voice interaction.
Waveary 是一个面向长期陪伴的开源框架。 它为任意大模型补上长期记忆、关系成长、人生时间轴、情绪连续性与语音交互这一层运行时能力。
Waveary is not trying to create a slightly smarter chatbot. It is trying to build the infrastructure for an AI that can remember, understand, grow, and stay with a person over time.
Waveary 不是在做一个“更会回答问题”的聊天机器人，而是在构建一种基础设施，让 AI 可以长期记住、理解、成长，并持续陪伴一个人的生活。
Most AI products can answer. Very few can continue.
多数 AI 产品会回答问题，但很少真的会延续关系。
What users miss is not only intelligence. It is:
- being remembered
- being understood in context
- being accompanied across real life events
- being cared for with emotional continuity
用户真正缺少的不只是“更聪明的回答”，而是：
- 被记住
- 被持续理解
- 被陪伴着经历真实生活
- 被带着情绪温度地关心
Waveary is designed to fill that gap with systems, not roleplay tricks.
Waveary 想补上的，正是这层“连续性系统”，而不是只靠提示词维持的人设幻觉。
- a Digital Life Companion Framework
- a continuity layer for companion products
- a reusable runtime for memory, relationship, timeline, emotion, and voice
- a multi-provider foundation rather than a single-model wrapper
- 数字生命陪伴框架
- 面向陪伴产品的连续性基础层
- 一套可复用的记忆、关系、时间轴、情绪、语音运行时
- 面向多模型兼容的基础设施，而不是单模型外壳
AI GirlfriendAI Boyfriend- a prompt-only roleplay shell
- a generic chatbot skin
- “AI 女友项目”
- “AI 男友项目”
- 只靠提示词撑起来的角色扮演壳
- 普通聊天机器人换皮
| Thesis | 中文 |
|---|---|
| Memory comes before model. | 记忆优先于模型。 |
| Relationship comes before features. | 关系优先于功能。 |
| Companionship comes before intelligence. | 陪伴优先于智能。 |
These three rules define the product direction and the architecture boundary of Waveary.
这三条不是口号，而是 Waveary 的产品方向与架构边界。
| Area | Status | Notes |
|---|---|---|
| Chat runtime | Available | Browser-first chat shell is usable |
| Multi-provider access | Available | OpenAI-compatible provider abstraction exists |
| Model discovery | Available | Provider /models discovery is supported |
| Long-term memory | Available baseline | Extraction, storage, retrieval scaffold exists |
| Relationship continuity | Available baseline | Stateful continuity layer exists |
| Timeline continuity | Available baseline | Timeline-aware session structure exists |
| Identity summary | Available baseline | Concept-level user and bond understanding exists |
| Emotion continuity | In progress | Companion-side emotion state is expanding |
| Voice routing | Available baseline | Dedicated voice path and provider routing exist |
| Realtime voice | In progress | Interruption-safe browser loop is underway |
| Proactive care | In progress | Decision layer exists and is being extended |
| 模块 | 状态 | 说明 |
|---|---|---|
| 对话运行时 | 可用 | 浏览器优先的对话外壳已可使用 |
| 多供应商接入 | 可用 | 已有 OpenAI-compatible 抽象层 |
| 模型发现 | 可用 | 已支持供应商 /models 检索 |
| 长期记忆 | 基线可用 | 已具备提取、存储、检索骨架 |
| 关系连续性 | 基线可用 | 已有有状态关系层 |
| 时间轴连续性 | 基线可用 | 已有面向时间轴的会话结构 |
| 概念级身份摘要 | 基线可用 | 已能沉淀“这个人是谁、这段关系像什么” |
| 情绪连续性 | 持续建设中 | 陪伴侧情绪状态仍在扩展 |
| 语音路由 | 基线可用 | 已有独立语音路径与供应商路由 |
| 实时语音 | 持续建设中 | 正在推进可打断的浏览器实时对话 |
| 主动关怀 | 持续建设中 | 已有决策层，仍在继续完善 |
| Engine | Name | Responsibility |
|---|---|---|
WME |
Waveary Memory Engine | Long-term memory |
WRE |
Waveary Relationship Engine | Relationship growth |
WTE |
Waveary Timeline Engine | Timeline and recall |
WEE |
Waveary Emotion Engine | Emotional continuity |
WVE |
Waveary Voice Engine | Voice interaction |
| 引擎 | 名称 | 职责 |
|---|---|---|
WME |
Waveary Memory Engine | 长期记忆 |
WRE |
Waveary Relationship Engine | 关系成长 |
WTE |
Waveary Timeline Engine | 时间轴与回忆 |
WEE |
Waveary Emotion Engine | 情绪连续性 |
WVE |
Waveary Voice Engine | 语音交互 |
flowchart TD
UI["Web / Mobile UI"] --> Runtime["Waveary Runtime"]
Runtime --> Memory["WME"]
Runtime --> Relationship["WRE"]
Runtime --> Timeline["WTE"]
Runtime --> Emotion["WEE"]
Runtime --> Voice["WVE"]
Runtime --> Providers["Model / Voice Providers"]
Waveary sits between product interfaces and model vendors. It turns short-term model interaction into long-term companion continuity.
Waveary 位于产品界面与模型供应商之间，负责把短期模型交互提升为长期陪伴连续性。
This graph tracks how public GitHub attention grows over time.
这张图反映了项目在 GitHub 上的公开关注趋势。
waveary/
waveary-core
waveary-memory
waveary-voice
waveary-web
waveary-dataset
docs
Current module roles:
waveary-core: runtime orchestration, provider abstraction, stateful continuitywaveary-memory: memory extraction, storage, retrievalwaveary-voice: voice adapters and voice runtime pathwaveary-web: official web surfacewaveary-dataset: markdown-first companion soul and conversation rules
当前模块分工：
waveary-core：运行时编排、供应商抽象、连续性状态层waveary-memory：记忆提取、存储与检索waveary-voice：语音适配器与语音运行时链路waveary-web：官方 Web 产品界面waveary-dataset：以 Markdown 为主的陪伴灵魂与对话规则
Node.js 20+npm 10+
git clone https://github.com/K2st0r/Waveary.git
cd Waveary
npm installnpm run web:devnpm run web:buildnpm run testnpm run verify:provider-
Choose a provider.
-
Enter the base URL and API key.
-
Discover available models from the provider.
-
Select the model you want to use.
-
Start a conversation with memory, relationship, timeline, emotion, and voice layered on top.
-
选择供应商。
-
填写 Base URL 与 API Key。
-
从供应商检索可用模型。
-
选择你要使用的模型。
-
开始一段具备记忆、关系、时间轴、情绪与语音能力的对话。
Waveary is being built as a layered system, not as a one-shot app shell.
Waveary 的推进方式不是一次性堆满功能，而是一层一层把长期陪伴系统做出来。
- Chat
- Long-term memory
- Timeline
- Relationship growth
- Voice
- Emotion analysis
- Proactive care
- Realtime voice
- Interruption handling
- Full duplex conversation
Long term, Waveary aims to become the continuity and life-memory layer for large models.
从更长远看，Waveary 想成为所有大模型的“人格连续性层”和“人生记忆层”。
Issues, architecture discussion, feature proposals, and focused pull requests are welcome.
欢迎提交 issue、架构讨论、功能提案与聚焦明确的 pull request。
If you want to contribute:
- Fork the repository.
- Create a branch.
- Make one focused change.
- Open a pull request.
如果你想参与贡献：
- Fork 仓库。
- 新建分支。
- 完成一个聚焦明确的改动。
- 发起 pull request。
This project is released under the MIT License.
本项目基于 MIT License 开源。