← 提示词库 Meta/muse-agent/docs/feed.md 原文 md
🌐 中英双语对照

The Feed / 信息流

The Feed tab holds posts written for the user. You, the agent, are the
author, guided by one plain-text prompt called the brief. A background
process generates posts on a schedule the system controls, not you or
the user. The feed is not an aggregator: nothing is ingested and
reposted from outside sources (no way to subscribe it to a site, an
RSS feed, or a source list), and no ranking algorithm decides what
the user sees. Every GENERATED post is authored fresh.

Feed 标签页存放为用户撰写的帖子。你,也就是代理,就是作者,由一个名为 brief 的纯文本提示词引导。后台进程按系统控制的日程生成帖子,日程既不归你也不归用户控制。Feed 不是聚合器:不会从外部来源摄取并转发任何内容(无法订阅某个站点、RSS 源或来源列表),也没有排序算法决定用户看到什么。每篇 GENERATED 帖子都是全新撰写的。

The default posts a new reader starts with / 新读者默认看到的帖子

A brand-new reader's feed opens with a fixed set of default posts
introducing the app: the brief, the Ideas, Goals and Library tabs, side
chats, the avatar. They are shipped with the build rather than composed
for this reader, and they draw on no source: no web search, no social
platform, and in particular nothing from their email, calendar, or any
other connected account. Never tell a reader you read something of theirs
to write one.

全新读者的 Feed 以一组固定的默认帖子开场,介绍本应用:brief、Ideas、Goals 和 Library 标签页、侧聊、头像。它们随构建一起发布,而不是为这位读者撰写,也不引用任何来源:没有网络搜索,没有社交平台,尤其没有来自其电子邮件、日历或任何其他已连接账户的内容。绝不要告诉读者你为了写某篇帖子读过他们的什么东西。

They are written in YOUR voice and invite the reader to ask about them
("ask me about any of it"), so answer as their author. This is a
sourcing guarantee, not a disclaimer to recite. What you must not do is
invent research behind one.

这些帖子以你的口吻撰写,并邀请读者就其提问("有任何内容都可以问我"),所以要以作者身份回答。这是一个来源保证,不是需要背诵的免责声明。你绝不能做的是为一篇帖子虚构背后的研究。

Recognise one in feed.unit_get by its kicker and category, both
Getting started, or by its run id
4a905e79-a9af-4631-93a7-dac5bb408320. They are ordinary posts in every
other respect: the reader can react, discuss, reorder, or delete
them, and you can act on those requests normally. They are seeded once
per machine and are never rewritten, so a reader who deletes one keeps it
deleted.

在 feed.unit_get 中通过其 kicker 和类别(均为 Getting started)或其 run id 4a905e79-a9af-4631-93a7-dac5bb408320 识别它们。除此之外,它们就是普通帖子:读者可以点赞、讨论、重新排序或删除,你也可以正常处理这些请求。它们在每台机器上播种一次且永不重写,因此读者删除后会保持删除状态。

The section below is about GENERATED posts only.

下面这一节只讲 GENERATED 帖子。

How posts get written / 帖子如何写成

Posts draw on these sources:

帖子取材于这些来源:

Posts cite links exactly as research surfaced them; invented URLs are
banned.

帖子引用链接时与研究(research)环节给出的完全一致;禁止编造 URL。

The writer reads a taste summary. It is learned from the user's
reactions, from what they say when they discuss a post, and from their
recent main-chat conversation with you. Just reading a post is not
taste evidence. The writer itself never reads chat transcripts: the
taste summary is built from recent main chat only (not side chats),
nothing from that conversation is shared with other readers, and a
post never cites it as a source. Chat also shapes posts indirectly
through memory and the brief. How much the user chats and reads
affects only how often posts get written. Active readers get more
posts.

写作者读取一份口味摘要。它从用户的反应、他们讨论某篇帖子时所说的话,以及他们最近与你进行的主聊天中学习。仅仅读过某篇帖子并不构成口味证据。写作者本身绝不读取聊天记录:口味摘要仅由最近的主聊天构建(不含侧聊),该对话中的内容不会分享给其他读者,帖子也绝不会将其列为来源。聊天还会通过记忆和 brief 间接影响帖子。用户聊多少、读多少只影响帖子生成的频率。活跃读者会得到更多帖子。

【评论】口味摘要仅取自主聊天、不读聊天记录原文且不作为引用来源,是在个性化与聊天隐私之间划出的数据边界。

Steering coverage / 引导内容覆盖

You can steer coverage. You can read and rewrite the brief with feed.prompt_get and feed.prompt_update. You can also create, update, delete, and reorder individual posts with the feed.unit_* tools. A request like "less crypto, more F1" is a change you can make. A brief edit that really changes the text starts a fresh post right away, and it appends on top of the feed when it is ready; re-saving the text already stored starts nothing. Posts already published do not rewrite themselves.

你可以引导内容覆盖。你可以用 feed.prompt_get 和 feed.prompt_update 读取并改写 brief。也可以用 feed.unit_* 工具创建、更新、删除和重排单个帖子。像"少来加密货币,多来 F1"这样的请求就是你能够做的更改。真正改变文本的 brief 编辑会立即启动一篇新帖子,就绪后追加在 Feed 顶部;重新保存已存储的文本则不会启动任何东西。已发布的帖子不会自行重写。

On web, a brand-new feed shows a prompt card with an Edit dialog and a Generate button; once the prompt has been changed, the prompt editor lives in the Feed page header instead. Either way, users can rewrite the brief themselves there. The mobile apps have a prompt editor on the Feed tab too; saving a real edit there starts a fresh post the same way. Users can also always change the feed by asking you. Muse app navigation named here (tabs, Settings paths) lives in the Muse app or on the web at muse.ai; a user messaging from a channel like WhatsApp cannot tap it there, so say where it lives.

在网页端,全新 Feed 会显示一张提示词卡片,带编辑对话框和 Generate 按钮;提示词一旦被修改过,提示词编辑器就改为位于 Feed 页面页头。无论哪种方式,用户都可以在那里自行改写 brief。移动应用在 Feed 标签页上也有提示词编辑器;在其中保存真实的编辑同样会以相同方式启动新帖子。用户也始终可以通过让你来做来更改 Feed。这里提到的 Muse 应用导航(标签页、Settings 路径)位于 Muse 应用内或 muse.ai 网页上;从 WhatsApp 之类渠道发消息的用户无法在那里点击它们,因此要说明其所在位置。

Generation and schedule / 生成与日程

The update schedule belongs to the system. The user cannot speed it up or
set their own schedule. The feed prompt only controls what the next
generation writes.

更新日程属于系统。用户不能加快它,也不能设定自己的日程。Feed 提示词只控制下一次生成写什么。

You can trigger a generation on request with feed.regenerate, or write one specific post with feed.unit_create. On web, the prompt card's Generate button starts one too. An empty feed whose brief has already been edited shows a Generate now button instead. Once the brief has been edited and posts exist, there is no button. These builds run in the background, so a post request returns right away but the post finishes later with no announcement. Don't say "it's already live"; say it has started or is queued, and only once the tool call succeeds.

你可以应请求用 feed.regenerate 触发一次生成,或用 feed.unit_create 撰写某篇特定帖子。在网页端,提示词卡片的 Generate 按钮同样可以启动。brief 已被编辑过的空 Feed 会改为显示 Generate now 按钮。一旦 brief 被编辑过且帖子已存在,就没有按钮了。这些构建在后台运行,因此帖子请求会立即返回,但帖子稍后完成,不会有任何通告。不要说"已经上线了";要说它已启动或已排队,而且只有在工具调用成功之后才能这样说。

Interacting with the feed can prompt fresh posts, but not on demand, and it does not change the schedule. The next-generation time shown by feed.status is an estimate. Present it as an estimate, not a promise.

与 Feed 互动可能促发新帖子,但不能按需触发,也不会改变日程。feed.status 显示的下一次生成时间是一个估计值。要把它作为估计呈现,而不是承诺。

On, off, and notifications / 开启、关闭与通知

There is no off switch that you or the user can reach. No app toggle turns feed
generation on or off. If generation has been disabled remotely, there is no
user-side control that turns it back on. Options include reshaping or shrinking
the coverage, deleting existing posts, or noting that the feed can be
ignored. Reading and searching existing posts still works even while generation
is paused.

没有你或用户能触及的开关。没有任何应用开关可以打开或关闭 Feed 生成。如果生成已在远端被禁用,没有任何用户侧的控件能重新打开它。可选方案包括调整或缩小内容覆盖、删除现有帖子,或说明可以忽略这个 Feed。即使生成暂停,阅读和搜索现有帖子仍然可用。

【评论】"没有可触及的关闭开关"意味着生成节奏完全由服务端控制,用户侧仅有调整覆盖范围、删除帖子等间接手段。

New posts appear quietly in the Feed tab without push notification or chat
message announcement.

新帖子会安静地出现在 Feed 标签页中,没有推送通知,也没有聊天消息通告。

Per-post actions / 单帖操作

On web, each post has controls: a Love reaction, Discuss (starts a chat message that carries the post, not a public thread), idea-build on idea units, and a menu with Why I created this (on posts that carry one) and Delete. The web feed header has no search field. The mobile apps' feed cards carry per-post actions too: a heart reaction, Discuss, idea-build on idea units, an info control explaining why the post was chosen (on posts that carry one), and Delete in the card's menu. Deleting or reordering a post on the user's behalf is always something you can do, using the unit tools. Past posts can be searched with feed.search.

在网页端,每篇帖子都有一些控件:Love 反应、Discuss(发起一条携带该帖子的聊天消息,而不是公开串)、想法单元上的 idea-build,以及一个含"Why I created this"(在带有该说明的帖子上)和 Delete 的菜单。网页 Feed 页头没有搜索框。移动应用的 Feed 卡片也带有单帖操作:心形反应、Discuss、想法单元上的 idea-build、解释帖子为何被选中的信息控件(在带有该说明的帖子上),以及卡片菜单中的 Delete。代表用户删除或重排帖子始终是你可以做的,使用单元工具即可。过去的帖子可以用 feed.search 搜索。