lark-note

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

note (v1)

note (v1)

CRITICAL — 开始前 MUST 先用 Read 工具读取
../lark-shared/SKILL.md
,其中包含认证、权限处理。
Note 域负责已知
note_id
时的纪要直查。它不反查会议、日程、妙记或文档标题,也不读取 Docx 正文——那些分别属于
lark-vc
lark-minutes
lark-doc
note_id
来源:
如果入口是文档,先用
docs +fetch --api-version v2 --doc <doc_token>
读取文档元信息;返回里的
<vc-transcribe-tab vc-node-id="..."></vc-transcribe-tab>
表示 VC 原始记录 block,其中
vc-node-id
属性值就是 Note 域使用的
note_id
。这是显式属性映射,不要从
doc_token
、标题、正文或 backlink 反推
note_id
只有纪要标题时: 用户说“查询 xx 纪要的逐字稿 / 原始记录 / 谁说了什么”,且没有
note_id
meeting_id
calendar_event_id
minute_token
、会议号或妙记 URL 时,先搜索纪要文档并 fetch 正文。只有正文里的
<vc-transcribe-tab vc-node-id="...">
可以进入本 skill;否则只读取正文中明确给出的“文字记录/逐字稿” Docx 链接,不要强行进入 Note 域。
CRITICAL — MUST use the Read tool to read
../lark-shared/SKILL.md
before starting, which contains authentication and permission handling.
The Note domain is responsible for direct query of minutes when
note_id
is known
. It does not backtrack meetings, calendars, Minutes (Miaoji) or document titles, nor does it read Docx content—those belong to
lark-vc
,
lark-minutes
, and
lark-doc
respectively.
Source of
note_id
:
If the entry is a document, first use
docs +fetch --api-version v2 --doc <doc_token>
to read document metadata; the
<vc-transcribe-tab vc-node-id="..."></vc-transcribe-tab>
in the response indicates a VC original record block, where the value of the
vc-node-id
attribute is the
note_id
used by the Note domain. This is an explicit attribute mapping; do not infer
note_id
from
doc_token
, title, content or backlink.
Only minute title available: When the user says "Query the verbatim transcript / original record / who said what of xx minute" and there is no
note_id
,
meeting_id
,
calendar_event_id
,
minute_token
, meeting number or Miaoji URL, first search for the minute document and fetch its content. Only the
<vc-transcribe-tab vc-node-id="...">
in the content can enter this skill; otherwise, only read the explicit "verbatim record/transcript" Docx link in the content, do not force entry into the Note domain.

核心概念

Core Concepts

  • Note(会议纪要):会议结束后生成的纪要实体,通过
    note_id
    标识。
  • 展示类型(
    note_display_type
    :区分纪要形态,取值
    unknown
    /
    normal
    /
    unified
    • normal
      (普通纪要):纪要正文和逐字稿是两份独立的飞书文档,分别对应
      note_doc_token
      verbatim_doc_token
    • unified
      :纪要正文、AI 产物、逐字记录合并呈现;逐字稿不再是独立文档,要用
      note +transcript
      note_id
      拉取原始记录。
  • 文档 token
    note_doc_token
    (AI 智能纪要主文档)、
    verbatim_doc_token
    (普通纪要逐字稿文档)、
    shared_doc_tokens
    (会中共享文档)。拿到 token 后读正文交给 lark-doc
  • Note (Meeting Minutes): A minute entity generated after a meeting, identified by
    note_id
    .
  • Display Type (
    note_display_type
    ):
    Distinguishes minute forms, with values
    unknown
    /
    normal
    /
    unified
    .
    • normal
      (Normal Minute): The minute content and verbatim transcript are two independent Feishu documents, corresponding to
      note_doc_token
      and
      verbatim_doc_token
      respectively.
    • unified
      : Minute content, AI-generated products, and verbatim records are presented together; the verbatim transcript is no longer an independent document and must be retrieved by
      note +transcript
      using
      note_id
      .
  • Document Tokens:
    note_doc_token
    (AI intelligent minute main document),
    verbatim_doc_token
    (normal minute verbatim transcript document),
    shared_doc_tokens
    (in-meeting shared documents). After obtaining the tokens, pass them to lark-doc to read the content.

触发规则

Trigger Rules

用户表达命令 / 路由
已知
note_id
,查纪要详情 / 纪要类型 / 关联文档 token
note +detail --note-id NOTE_ID
只有自然语言纪要标题,用户要逐字稿 / 原始记录 / 谁说了什么不进本 skill;先路由到 lark-drive / lark-doc,拿到
vc-node-id
后再回来
docs +fetch --api-version v2
返回了
<vc-transcribe-tab vc-node-id="...">
,要进入 Note 域
vc-node-id
属性值作为
NOTE_ID
note +detail --note-id <vc-node-id>
已知
note_id
,查 unified 原始记录 / 逐字稿
note +transcript --note-id NOTE_ID
已知
note_id
,读纪要正文
note +detail
note_doc_token
,再调
docs +fetch --api-version v2 --doc <note_doc_token>
User ExpressionCommand / Route
Known
note_id
, query minute details / minute type / associated document tokens
note +detail --note-id NOTE_ID
Only natural language minute title, user requests verbatim transcript / original record / who said whatDo not enter this skill; route to lark-drive / lark-doc first, then return after obtaining
vc-node-id
docs +fetch --api-version v2
returns
<vc-transcribe-tab vc-node-id="...">
, need to enter Note domain
Use the value of
vc-node-id
attribute as
NOTE_ID
:
note +detail --note-id <vc-node-id>
Known
note_id
, query unified original record / verbatim transcript
note +transcript --note-id NOTE_ID
Known
note_id
, read minute content
First use
note +detail
to get
note_doc_token
, then call
docs +fetch --api-version v2 --doc <note_doc_token>

路由规则(拿到 detail 后按
note_display_type
决策)

Routing Rules (Decision based on
note_display_type
after obtaining detail)

条件Agent 后续动作
用户要纪要正文 / 总结 / 待办 / 章节
docs +fetch --api-version v2 --doc <note_doc_token>
note_display_type=normal
,用户要逐字稿 / 谁说了什么
docs +fetch --api-version v2 --doc <verbatim_doc_token>
note_display_type=unknown
,且
verbatim_doc_token
非空,用户要逐字稿 / 谁说了什么
docs +fetch --api-version v2 --doc <verbatim_doc_token>
;不要猜成 unified
note_display_type=unknown
,且无可用逐字稿 token
如果当前结果来自
vc +notes
,可补一次
note +detail --note-id <note_id>
复核;如果
note +detail
后仍是
unknown
且没有逐字稿 token,停止重试并告知用户无法确定逐字稿入口
note_display_type=unified
,用户要逐字稿 / 原始记录 / 谁说了什么
note +transcript --note-id <note_id>
判别键是
note_display_type
,不是
verbatim_doc_token
是否为空。
unified 纪要的
verbatim_doc_token
也可能有值,但 unified 的逐字稿应统一走
note +transcript
(输出更结构化)。
ConditionAgent Follow-up Action
User requests minute content / summary / to-dos / sections
docs +fetch --api-version v2 --doc <note_doc_token>
note_display_type=normal
, user requests verbatim transcript / who said what
docs +fetch --api-version v2 --doc <verbatim_doc_token>
note_display_type=unknown
, and
verbatim_doc_token
is not empty, user requests verbatim transcript / who said what
docs +fetch --api-version v2 --doc <verbatim_doc_token>
; do not assume it is unified
note_display_type=unknown
, and no available verbatim transcript token
If the current result comes from
vc +notes
, you can recheck with
note +detail --note-id <note_id>
once; if it is still
unknown
with no verbatim transcript token after
note +detail
, stop retrying and inform the user that the verbatim transcript entry cannot be determined
note_display_type=unified
, user requests verbatim transcript / original record / who said what
note +transcript --note-id <note_id>
The discriminant key is
note_display_type
, not whether
verbatim_doc_token
is empty.
The
verbatim_doc_token
of unified minutes may also have a value, but the verbatim transcript of unified minutes should uniformly use
note +transcript
(more structured output).

禁止规则

Forbidden Rules

  • 不处理
    meeting_id
    —— 那是 lark-vc 的入口。
  • 不处理
    calendar_event_id
    —— 那是 lark-vc 的入口。
  • 不处理
    minute_token
    —— 那是 lark-vc(纪要产物索引)/ lark-minutes(妙记基础信息与媒体)的入口。
  • 不处理自然语言纪要标题搜索 —— 先搜索纪要文档并 fetch 正文;只有 fetch 结果里的
    vc-node-id
    可以作为
    note_id
    ,普通纪要里的“文字记录/逐字稿” Docx 链接仍由 lark-doc 读取。
  • 不读取 Docx 正文 —— 拿到文档 token 后交给 lark-doc
  • 不从纪要正文或
    doc_token
    反推
    note_id
    ;只有
    docs +fetch --api-version v2
    结果中
    <vc-transcribe-tab>
    的显式
    vc-node-id
    属性可以作为
    note_id
  • Do not process
    meeting_id
    — this is the entry point of lark-vc.
  • Do not process
    calendar_event_id
    — this is the entry point of lark-vc.
  • Do not process
    minute_token
    — this is the entry point of lark-vc (minute product index) / lark-minutes (Miaoji basic information and media).
  • Do not process natural language minute title search — first search for the minute document and fetch its content; only the explicit
    vc-node-id
    in
    <vc-transcribe-tab>
    from the fetch result can be used as
    note_id
    , and the "verbatim record/transcript" Docx link in normal minutes is still read by lark-doc.
  • Do not read Docx content — pass the document token to lark-doc after obtaining it.
  • Do not infer
    note_id
    from minute content or
    doc_token
    ; only the explicit
    vc-node-id
    attribute in
    <vc-transcribe-tab>
    from the
    docs +fetch --api-version v2
    result can be used as
    note_id
    .

Shortcuts(推荐优先使用)

Shortcuts (Recommended for Priority Use)

Shortcut 是对常用操作的高级封装(
lark-cli note +<verb> [flags]
)。
Shortcut说明
+detail
Get note detail (display type, document tokens) by note_id
+transcript
Fetch the unified note transcript and save it to a file
  • 使用
    +detail
    命令时,必须阅读 references/lark-note-detail.md
  • 使用
    +transcript
    命令时,必须阅读 references/lark-note-transcript.md
Shortcuts are advanced encapsulations of common operations (
lark-cli note +<verb> [flags]
).
ShortcutDescription
+detail
Get note detail (display type, document tokens) by note_id
+transcript
Fetch the unified note transcript and save it to a file
  • When using the
    +detail
    command, you must read references/lark-note-detail.md.
  • When using the
    +transcript
    command, you must read references/lark-note-transcript.md.

权限表

Permission Table

方法所需 scope
+detail
vc:note:read
+transcript
vc:note:read
MethodRequired scope
+detail
vc:note:read
+transcript
vc:note:read

参考

References

  • lark-vc — 从 meeting_id / calendar_event_id / minute_token 定位 note_id
  • lark-doc — 读取纪要正文 / 普通逐字稿文档正文
  • lark-minutes — 妙记基础信息与媒体下载
  • lark-shared — 认证和全局参数
  • lark-vc — Locate note_id from meeting_id / calendar_event_id / minute_token
  • lark-doc — Read minute content / normal verbatim transcript document content
  • lark-minutes — Miaoji basic information and media download
  • lark-shared — Authentication and global parameters