Codex EchoInk

by Akin-Lvyf
5
4
3
2
1
Score: 55/100

Description

Codex for Obsidian plugin

Reviews

No reviews yet.

Stats

48
stars
3,601
downloads
6
forks
125
days
13
days
13
days
4
total PRs
1
open PRs
2
closed PRs
1
merged PRs
6
total issues
0
open issues
6
closed issues
1,015
commits

Latest Version

13 days ago

Changelog

2.8.10 · 一人公司开工

EchoInk 2.8.10:一人公司开工

中文

类型:新功能

更新内容

  • 总助可以带着角色 Agent 一起做项目。 小事直接派给员工,多人配合的事先给你开工卡,列出分工、里程碑和验收标准,点「开工」后再执行。需要回答的问题与写入审批集中在「等你定」,项目页可以查看进度、员工记录与产出。一人公司在电脑、iPhone 和 iPad 提供入口;移动端仍为 Beta,项目和员工会话暂不跨设备查看、接续,手机只在 App 前台推进项目。
  • 每个模块有明确的负责人。 在 Agent 设置中给日记、待办、财务等模块指定角色 Agent;无人负责时先选人,再启动任务。总助负责读取资料、派活和收汇报,内容交给员工执行。总助编辑页可选「我来审批」或「总助审批」,默认由你审批,删除始终需要你确认。负责人和一人公司本身不要求 PRO,财务、专家聘请等原有 PRO 限制仍然适用。
  • 电脑端可以梳理知识体系、给知识库做体检。 按领域查看地图、选择知识树方案,确认后写入;新增知识可归入已有树,保留手写内容。体检分快查与深查,可检查链接、分类、重复或过时内容,再预览、确认适用的修复。深查和体系梳理需要调用你配置的模型,模型判断可能遗漏或误判,修改前请核对。
  • 电脑端新增记忆星图,记忆也有了活跃与淡出。 从「设置 → 记忆」打开记忆页,查看画像、记忆及来源,修正内容或固定重要记忆。固定的记忆保留在核心,其他记忆按类型、时间和使用情况调整活跃度;淡出只退出自动召回,不会删除,仍可搜索、恢复。桌面与移动共用相关性和生命周期规则,桌面自动召回增加按当前问题挑选相关记忆;星图暂只在电脑端提供。
  • 手机和平板可以边收边显示回答与思考。 在模型服务参数中设置「流式输出」;有思考内容时,先展开显示,正文开始后自动收起,也可手动开合。流式是否可用取决于服务商,连接被拦时可关闭流式,改为整段接收。本版修复 iOS 上 HTML 扩展名冲突可能导致插件加载失败的问题,并调整输入框对悬浮底栏和键盘的避让;这些改动未经 iPhone/iPad 真机验证。
  • 长时间循环调用工具会自动停止。 到达上限时保留中断提示,可发送「继续」接着做。命令行未启用时会正确提示不可用,不再误报连错笔记库。记忆设置集中到独立分组,导航图标在点击时播放动效。

如何使用

打开 EchoInk 设置中的 Agent,为需要使用的模块指定负责人,再回到总助交代任务。复杂任务先核对开工卡,确认后点「开工」;到工作台的「等你定」处理问题与审批。想调整员工审批方式,进入总助编辑页;选择总助审批后,拿不准的事项仍会交给你。

电脑端从知识库页进入体系梳理或体检,先看建议与修复预览,再确认写入。从「设置 → 记忆」打开记忆页查看星图。记忆星图、知识库体检和体系梳理不要求 PRO;调用模型使用你自己的额度。手机流式输出在对应模型服务的参数中设置,不支持的服务可保留整段回答。

下载与升级

需要 Obsidian 1.11.4 及以上。升级前完整备份 Vault、插件设置和数据,多设备请统一升级到 2.8.10。通过 Obsidian 社区插件检查更新,或从本次 Release 下载 main.js、manifest.json、styles.css,只替换 .obsidian/plugins/codex-echoink/ 中这三个文件,保留原有设置和数据,再重新启用插件。移动端仍为 Beta,本版新增与修复未经 iPhone/iPad 真机验证;支持范围见 README。模型与搜索服务的费用由相应服务商另行收取。

English

2.8.10 · Start your one-person company

Type: New

What's new

  • Your assistant can coordinate role agents on projects. Small jobs go directly to an employee; work involving several roles starts with a proposal showing assignments, milestones, and acceptance criteria. Execution begins only after you click Start. Questions and write approvals appear together under Pending decisions, while the project page shows progress, employee records, and outputs. The workflow is available on desktop, iPhone, and iPad. Mobile remains in Beta: projects and employee conversations cannot yet be viewed or resumed across devices, and mobile projects advance only while the app is in the foreground.
  • Assign an owner to each module. Choose role agents for Journal, To-dos, Finance, and other modules in Agent settings. If a module has no owner, choose one before starting its agent task. The assistant reads, delegates, and receives reports; employees carry out content work. In the assistant editor, choose manual approval or assistant approval. Manual is the default, and deletions always require your confirmation. Module ownership and the company workflow do not require PRO; existing PRO requirements for Finance, expert hiring, and other paid features still apply.
  • Organize knowledge systems and run checkups on desktop. View domain maps, choose a knowledge-tree proposal, and confirm before writing it. New knowledge can join existing trees while handwritten content stays intact. Quick and deep checkups inspect issues such as links, classification, duplicate content, or outdated material, with previews and confirmation for supported repairs. Deep checkups and system organization use your configured model. Model judgments can miss issues or be wrong; review proposed changes.
  • A desktop memory constellation shows what is remembered. Open the memory page from Settings → Memory to inspect your profile, memories, and sources, correct content, or pin important memories. Pinned memories stay at the core; others change activity according to type, time, and use. Faded memories leave automatic recall without being deleted and remain searchable and restorable. Desktop and mobile share relevance and lifecycle rules, and desktop automatic recall now selects memories for the current question. The constellation is desktop-only for now.
  • Stream answers and reasoning on phones and tablets. Set Streaming output in the model service parameters. Reasoning opens while it arrives and collapses when the answer begins; you can also toggle it yourself. Streaming depends on provider support. If the connection is blocked, turn streaming off to receive complete answers instead. This release also fixes an iOS HTML-extension conflict that could prevent plugin loading and adjusts the input area around floating navigation and the keyboard. These changes have not been verified on physical iPhone or iPad devices.
  • Repeated tool calls now stop at a limit. An interruption message remains, and you can send Continue to resume. A disabled command-line interface is correctly reported as unavailable instead of being mistaken for the wrong vault. Memory settings have their own group, and navigation icons animate when clicked.

How to use

Open Agent in EchoInk settings, assign owners to the modules you use, and give the assistant a task. For a complex task, review the proposal and click Start. Handle questions and approvals under Pending decisions on the workbench. Change employee approval mode in the assistant editor; uncertain decisions still return to you when assistant approval is enabled.

On desktop, open system organization or checkups from Knowledge, review the suggestions and repair preview, and confirm changes. Open Settings → Memory for the constellation. Memory constellation, knowledge checkups, and system organization do not require PRO; model calls use your own credits. Configure mobile streaming in each model service's parameters, or keep complete answers for unsupported services.

Download and upgrade

Requires Obsidian 1.11.4 or later. Back up your complete vault, plugin settings, and data before upgrading, and upgrade every device to 2.8.10. Check for updates in Obsidian's community plugins, or download main.js, manifest.json, and styles.css from this release. Replace only those three files in .obsidian/plugins/codex-echoink/, keep existing settings and data, and re-enable the plugin. Mobile remains in Beta; this version's additions and fixes have not been verified on physical iPhone or iPad devices. See the README for supported features. Model and search providers bill their services separately.

README file from

Github

What you can do

💬 Chat with your notes Ask anything about your vault. Answers cite the notes they actually used.
📚 Turn sources into knowledge Organize new material into a knowledge base, with health checkups and previewed repairs.
🧠 Memory you can see and edit EchoInk learns about you over time. Browse, correct, pin, or forget any memory.
📅 Journal, to-dos, and finances Write your day once. EchoInk can turn journal entries into tasks and records.
🤖 A small AI team An assistant plans and delegates, role agents do the work, and you approve the decisions.
✏️ Create from your sources Generate notes, Canvas layouts, and drawings from the material you already have.
🔌 Your notes as context for other agents A local MCP server lets the agents you already use read the knowledge and memory you authorize.

Your notes stay in your vault. Bring your own model API key, or use the built-in models with AI credits.

Get started

  1. Install. Open EchoInk in Obsidian, or search EchoInk in Community plugins. Requires Obsidian 1.11.4+.
  2. Set up. Open the workbench and pick Automatic, Guided, or Manual setup. Your existing notes stay where they are.
  3. Connect a model. Add an API key for OpenAI or a compatible provider, or sign in to use the built-in models, then start chatting.

Free, PRO, and Max

Free PRO Max
Chat, knowledge Q&A, knowledge base and checkups ✅ ✅ ✅
Long-term memory and memory constellation ✅ ✅ ✅
Journal, to-dos, workbench ✅ ✅ ✅
AI team, module owners, custom Skills ✅ ✅ ✅
@Agent inside notes, PDF reading — ✅ ✅
Finance, English diary — ✅ ✅
Expert hiring, official methodology Skills, EchoInk MCP server — ✅ ✅
Cloud sync — 1 GB 5 GB
AI credits for built-in models, per month — 500 3,000

You can bring your own model API key on any plan. See current plans and prices.

💬 Conversations and knowledge

Feature Description Basic PRO
Agent conversations Ask with notes, continue conversations, or archive them; desktop supports attachments ✅ ✅
Knowledge Q&A (Ask) Follow related sources and cite the sources actually used ✅ ✅
Knowledge systems and checkups Organize domain trees and run quick or deep checkups on desktop; preview and confirm repairs. Review model judgments; deep checkups and organization use model credits ✅ ✅
Knowledge maintenance Organize new or changed sources incrementally, update existing knowledge, and preserve handwritten content ✅ ✅
Generate from sources Generate notes, Canvas layouts, or editable drawings on desktop; drawings require Excalidraw 2.29 or later enabled and a model supporting tool calls ✅ ✅
PDF reading and extraction Extract originals and generate translations and reading notes on desktop; optional MinerU enhanced parsing ❌ ✅
@Agent in notes Ask in a note, review proposed edits, and confirm changes before writing them back; undo is available ❌ ✅
Web search Off by default; choose Tavily or DeepSeek, with separate service charges ✅ ✅

🧠 Memory and methods

Feature Description Basic PRO
Long-term memory Browse, correct, export, forget, restore, and pin; fading does not delete memories, and knowledge retrieval excludes private memory ✅ ✅
Memory constellation Inspect profile, memories, and sources on desktop via Settings → Memory; correct, pin, or restore memories ✅ ✅
Skills (built-in and custom) Built-in basic Skills, custom Skills, and basic knowledge review ✅ ✅
Official methodology Skills and enhanced knowledge review Seven official methodology Skills and enhanced knowledge review download on demand ❌ ✅
EchoInk MCP server On desktop, share authorized knowledge, profile, and memories with external agents on the same computer; off by default and read-only ❌ ✅

📅 Everyday life

Feature Description Basic PRO
Workbench View tasks, reminders, starter cards, and weather for your chosen city in one place on desktop ✅ ✅
Journal Customize naming; preview moves for older journals and undo those moves on desktop ✅ ✅
To-dos Manage categories, due dates and times, and completion; enable automatic journal-to-task organization on desktop ✅ ✅
Finance Manage income, expenses, budgets, bill imports, and analysis ❌ ✅
English diary Generate English drafts on desktop, correct errors or clearly unnatural English, and collect useful expressions ❌ ✅

🤖 Agent team

Feature Description Basic PRO
One-person company The assistant delegates on desktop, iPhone, and iPad; projects start after proposal confirmation, with questions and approvals in Pending decisions. Mobile is in Beta ✅ ✅
Module owners Assign role agents to modules in Agent settings; choose an owner before starting. Paid modules retain their own entitlement requirements ✅ ✅
Multi-agent tasks View employee tasks and outputs; review interruptions before resuming. Projects and employee conversations cannot yet be viewed or resumed across devices ✅ ✅
Expert hiring Browse and hire experts on desktop, continue from their avatars, and retain conversations after dismissal ❌ ✅
Item Description
Appearance Minimal is free; Highlighter and Manifesto require active PRO. Install the companion theme separately from echoink.cn/theme
Desktop Provides the desktop entries described above; validation in native Windows and Linux environments remains incomplete
Mobile Beta iPhone and iPad provide the workbench, company workflow, module owners, settings, and Journal, To-dos, Finance, Knowledge, and English Diary pages. Handle questions and approvals in Pending decisions. Projects and employee conversations cannot yet be viewed or resumed across devices; projects advance only in the foreground, with interrupted work checked before resuming. Stream answers and reasoning per model service where supported. Memory constellation, knowledge system organization, knowledge checkups, source generation, generation-template settings, and PDF reading are desktop-only. Attachments, CLI, and desktop background knowledge tasks are not connected; Skills use existing mobile tools, and MCP currently saves configuration only. This version's additions and fixes have not been verified on physical iPhone or iPad devices; native Android validation remains incomplete.

Privacy at a glance

  • Your notes stay local by default. Conversations, knowledge, memory, journals, and finances are stored in your vault. Only if you turn on Cloud Sync are files in the selected scope encrypted and uploaded to echoink.cn.
  • AI requests go to the model service you choose, with only the content needed for that request: directly to that service with your own API key, or through echoink.cn using AI credits when you choose xiaoecho.
  • echoink.cn is used for account sign-in and PRO authorization, Cloud Sync, xiaoecho, smart judgment, news, PRO Skill and expert downloads, optional weather, workbench check-in, and problem reports you choose to send after signing in. PRO requires an EchoInk account; chat and knowledge with your own model do not.
  • Optional third-party services: web search (Tavily or DeepSeek, off by default) and MinerU PDF parsing, each with its own account and billing.
  • On desktop, EchoInk can access files outside the vault according to the permission mode you select, and can run the local MCP commands you configure.
  • No client usage telemetry.

Read the full privacy and data details.

Good to know

  • Back up before upgrading. Keep a full copy of your vault and plugin data.
  • Obsidian Sync Standard: the plugin file is larger than its 5 MB per-file limit, so install and update EchoInk on each device separately.
  • Mobile is in Beta. This version's changes have not been verified on physical iPhone or iPad devices.
Step Action Instructions
1 Install Search for EchoInk in Obsidian's community plugins, or download main.js, manifest.json, and styles.css from Releases into .obsidian/plugins/codex-echoink/, then enable it in Community Plugins settings. Requires Obsidian 1.11.4 or later
2 Set up On desktop, open setup from the workbench and choose Automatic, Guided, or Manual. For Automatic or Guided, review the plan before confirming; existing notes stay in place, and a starter guide opens when setup finishes
3 Connect a model Bring your own model API key, or connect later. Directory setup and guide generation do not call a model; connection tests and source extraction you choose to enable use model credits
4 Start Open the workbench to begin a conversation; @Agent in notes requires active PRO

Make a complete backup of your vault, plugin settings, and data before upgrading. For manual upgrades, replace only the official main.js, manifest.json, and styles.css, keep your existing settings and data files, and re-enable the plugin. To roll back, restore the complete pre-upgrade backup together with the matching older plugin; replacing only the plugin files with an older version is not a supported downgrade.

The plugin file exceeds Obsidian Sync Standard's 5 MB per-file limit. Install or update the plugin independently on each device through the official download or Community Plugins; do not rely on Sync Standard to transfer this file.

What's new

2.9.0 (2026-11-19) — Sync your whole vault and pick up chats across devices. See the changelog and releases.

License

The base plugin is open source under the MIT License. Finance, English Diary, membership, and other PRO components use the EchoInk PRO License. See LICENSE for the exact scope and THIRD_PARTY_NOTICES.md for third-party material.