@itradingai/aiwiki 0.2.21 → 0.2.22

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/docs/FAQ.md CHANGED
@@ -1,100 +1,124 @@
1
1
  # AIWiki FAQ
2
2
 
3
- ## AIWiki 是什么?
3
+ ## What is AIWiki?
4
4
 
5
- AIWiki 是一个开源的 Agent-first 本地 LLM-wiki CLI。宿主 Agent 负责读取和理解内容,AIWiki 负责把结果稳定写成本地 Markdown 知识库。
5
+ AIWiki is a local Markdown knowledge base for AI assistants. Your assistant reads and understands sources; AIWiki writes structured, traceable, reusable Markdown knowledge files.
6
6
 
7
- ## AIWiki 是网页爬虫吗?
7
+ ## Is AIWiki a web crawler?
8
8
 
9
- 不是。网页读取主要交给宿主 Agent,AIWiki 专注于校验结构化输入、写本地文件、生成 Wiki Entry、提供 context lint。
9
+ No. Web reading belongs to the host assistant. AIWiki receives content the assistant already read and turns it into a local knowledge base.
10
10
 
11
- ## AIWiki 会自动高质量总结吗?
11
+ ## Does AIWiki call an LLM?
12
12
 
13
- 不会。AIWiki CLI 本身不调用 LLM。高质量 Wiki Entry 来自宿主 Agent payload 中提供的 `analysis` `wiki_entry`。
13
+ No. AIWiki CLI does not call an LLM. High-quality summaries and analysis come from the host assistant through `analysis` or `wiki_entry` fields.
14
14
 
15
- ## 为什么会生成 05-wiki?
15
+ ## Why does installation fail?
16
16
 
17
- 因为 AIWiki 的目标不是只保存资料,而是让资料进入可查询、可维护的 Wiki 知识层。成功入库会生成 `05-wiki/source-knowledge/<slug>.md`。
17
+ Check Node.js first:
18
18
 
19
- ## Wiki Entry 的两种质量模式是什么?
19
+ ```bash
20
+ node --version
21
+ ```
22
+
23
+ AIWiki requires Node.js 20 or newer. If `node` is missing or older than 20, upgrade Node.js before running `npm install -g @itradingai/aiwiki@latest`.
20
24
 
21
- - `agent_enriched` / `enriched`:宿主 Agent 提供了总结、核心观点、知识点等分析结果。
22
- - `deterministic_fallback` / `scaffold`:AIWiki 只生成标题、来源、正文预览、反链和待补全区。
25
+ ## Why does AIWiki create `05-wiki`?
23
26
 
24
- ## 05-wiki 是否代表我的观点?
27
+ Because the goal is not just to save source material. AIWiki creates reusable Wiki Entries that assistants can query later.
25
28
 
26
- 不一定。外部资料生成的 Wiki Entry 默认代表外部资料的结构化整理,不代表用户个人观点。
29
+ ## What are the two Wiki Entry quality modes?
27
30
 
28
- ## 一定要用 Obsidian 吗?
31
+ - `agent_enriched` / `enriched`: the assistant provided analysis or wiki content.
32
+ - `deterministic_fallback` / `scaffold`: AIWiki created a traceable shell from source content only.
29
33
 
30
- 不一定。AIWiki 生成的是标准 Markdown frontmatter,Obsidian 只是最适合它的审阅界面之一。
34
+ ## Does a Wiki Entry represent my personal opinion?
31
35
 
32
- ## 一定要装 Dataview 吗?
36
+ Not by default. External sources use `source_role: input` and `represents_user_view: false`. Use `source_role: output` and `represents_user_view: true` only for your own published writing, scripts, talks, newsletters, or similar authored output.
33
37
 
34
- 不一定。没有 Dataview,也可以用 Obsidian 原生的 Properties、Backlinks、Search 和 Graph View。
38
+ ## Do I need Obsidian?
35
39
 
36
- ## 宿主 Agent AIWiki 各负责什么?
40
+ No. AIWiki writes plain Markdown and frontmatter. Obsidian is a useful viewing surface, not a requirement.
37
41
 
38
- - 宿主 Agent 负责读取网页、正文或附件,并尽量生成 `analysis` / `wiki_entry`
39
- - AIWiki 负责校验 payload、写本地知识库、生成 Wiki Entry、输出处理记录
40
- - 用户负责提供链接或正文,并在需要时让 Agent 继续查询或补全
42
+ ## Do I need Dataview?
41
43
 
42
- ## 为什么要先让 Agent 学会 AIWiki
44
+ No. Dataview is optional. AIWiki does not install Dataview and does not edit `.obsidian`.
43
45
 
44
- 因为大多数用户不想每次手动跑命令。更顺的方式是:把链接发给宿主 Agent,Agent 自动把内容交给 AIWiki,AIWiki 再把结果写进本地知识库。
46
+ ## Why should I sync the assistant first?
45
47
 
46
- ## 我能不能把它当成一个通用爬虫用?
48
+ Because the intended workflow is assistant-driven. You send a source or question to the assistant; the assistant calls AIWiki commands. Without sync, the assistant may fall back to generic file search and skip AIWiki's ingest, query, status, and lint surfaces.
47
49
 
48
- 不建议。AIWiki 的边界不是“抓全网”,而是“把宿主 Agent 已经读到的内容稳定沉淀下来”。
50
+ ## Where should I start?
49
51
 
50
- ## 我从哪里开始最好?
52
+ Ask your assistant to run the Quick Start prompt in the main [README](../README.md). Then read the [Usage Guide](USAGE.md).
51
53
 
52
- 先跑:
54
+ ## What should my first 10-minute trial include?
55
+
56
+ Use one temporary knowledge base, one source, one query, and one health check:
57
+
58
+ ```bash
59
+ aiwiki setup --path <workspace> --yes
60
+ aiwiki doctor --path <workspace>
61
+ aiwiki ingest-file --file <file> --path <workspace>
62
+ aiwiki query "<topic>" --path <workspace>
63
+ aiwiki context "<topic>" --path <workspace>
64
+ aiwiki lint --json --path <workspace>
65
+ ```
66
+
67
+ If the source is a URL, your assistant should read the URL and then call `aiwiki ingest-agent`; AIWiki does not crawl the URL itself.
68
+
69
+ ## Why is my assistant still reading files directly?
70
+
71
+ Usually the assistant has not loaded the AIWiki skill, or the workspace root is missing AIWiki guidance. Run:
53
72
 
54
73
  ```bash
55
- npx @itradingai/aiwiki@latest setup
56
74
  aiwiki agent sync --yes
57
- aiwiki agent check --json
58
75
  aiwiki agent sync --path <workspace> --yes
59
76
  aiwiki agent check --path <workspace> --json
60
77
  ```
61
78
 
62
- 然后去看:
79
+ Then restart or reload the assistant if needed.
63
80
 
64
- - [USAGE.md](USAGE.md)
65
- - [SHOWCASE.md](SHOWCASE.md)
66
- - [AGENT_HANDOFF.md](AGENT_HANDOFF.md)
81
+ ## How do I query the knowledge base?
67
82
 
68
- ## 有没有可以直接看的样例?
83
+ Ask your assistant:
69
84
 
70
- 有。`examples/demo-run/` 记录输入、命令和 CLI 输出;`examples/obsidian-vault-sample/` 是已经生成好的样例知识库。它展示了核心产物优先,以及可选增强目录只在有内容时出现。
85
+ ```text
86
+ What does AIWiki know about <topic>?
87
+ ```
71
88
 
72
- ## 怎么从知识库里查询?
89
+ The assistant should call:
73
90
 
74
- 让宿主 Agent 调用:
91
+ ```bash
92
+ aiwiki context "<topic>"
93
+ ```
94
+
95
+ For human terminal output:
75
96
 
76
97
  ```bash
77
- aiwiki context "主题"
98
+ aiwiki query "<topic>"
78
99
  ```
79
100
 
80
- ## 怎么检查知识库结构?
101
+ ## How do I check the workspace?
81
102
 
82
- 运行:
103
+ Ask your assistant to run:
83
104
 
84
105
  ```bash
85
- aiwiki lint
106
+ aiwiki lint --json
86
107
  ```
87
108
 
88
- 如果只是清理旧版本留下的空可选增强目录,可以让 Agent 先运行 `aiwiki lint --json`,确认只有 safe fix 后再运行 `aiwiki lint --fix-empty-dirs --json`。
109
+ `aiwiki doctor --path <workspace>` checks whether the workspace is ready and reports actionable setup problems.
89
110
 
90
- ## 为什么我的 Agent 还是在直接翻文件?
91
-
92
- 通常是它没有加载 AIWiki skill,或者知识库根目录缺少 AIWiki 的 `AGENTS.md` 指导。先运行:
111
+ When safe fixes are allowed:
93
112
 
94
113
  ```bash
95
- aiwiki agent sync --yes
96
- aiwiki agent sync --path <workspace> --yes
97
- aiwiki agent check --path <workspace> --json
114
+ aiwiki lint --fix-empty-dirs --json
115
+ aiwiki lint --json
98
116
  ```
99
117
 
100
- 之后再让 Agent 处理 AIWiki 任务。它应该优先调用 `aiwiki lint --json`、`aiwiki status`、`aiwiki query`、`aiwiki context`、`aiwiki ingest-file` 或 `aiwiki ingest-agent`。只有这些命令不够用时,才退回到通用文件搜索。
118
+ ## How do I send useful trial feedback?
119
+
120
+ Use [TRIAL_FEEDBACK_TEMPLATE.md](TRIAL_FEEDBACK_TEMPLATE.md). Focus on whether setup worked, whether the assistant called AIWiki commands, whether generated files were easy to inspect, whether query/context helped, and what real scenario you wanted to use.
121
+
122
+ ## What is out of scope?
123
+
124
+ AIWiki is not a web crawler, WeChat reader, browser extension, built-in LLM, vector database, RAG replacement, Obsidian plugin, default manual review queue, multi-knowledge-base manager, RSS tool, or scheduled collection system.
@@ -0,0 +1,103 @@
1
+ # AIWiki 常见问题
2
+
3
+ ## AIWiki 是什么?
4
+
5
+ AIWiki 是给 AI 助手使用的本地 Markdown 知识库。AI 助手负责读取和理解资料;AIWiki 负责写入结构化、可追踪、可复用的 Markdown 知识文件。
6
+
7
+ ## AIWiki 是网页爬虫吗?
8
+
9
+ 不是。网页读取属于宿主 AI 助手。AIWiki 接收助手已经读到的内容,并把它变成本地知识库。
10
+
11
+ ## AIWiki 会调用 LLM 吗?
12
+
13
+ 不会。AIWiki CLI 本身不调用 LLM。高质量总结和分析来自宿主助手提供的 `analysis` 或 `wiki_entry`。
14
+
15
+ ## 为什么安装失败?
16
+
17
+ 先检查 Node.js:
18
+
19
+ ```bash
20
+ node --version
21
+ ```
22
+
23
+ AIWiki 需要 Node.js 20 或更新版本。如果没有 `node` 命令,或版本低于 20,请先升级 Node.js,再运行 `npm install -g @itradingai/aiwiki@latest`。
24
+
25
+ ## 为什么会生成 `05-wiki`?
26
+
27
+ 因为 AIWiki 的目标不是只保存资料,而是创建以后能被助手查询和复用的 Wiki Entry。
28
+
29
+ ## Wiki Entry 的两种质量模式是什么?
30
+
31
+ - `agent_enriched` / `enriched`:助手提供了分析或 Wiki 内容。
32
+ - `deterministic_fallback` / `scaffold`:AIWiki 只根据来源内容生成可追踪脚手架。
33
+
34
+ ## Wiki Entry 代表我的观点吗?
35
+
36
+ 默认不代表。外部资料使用 `source_role: input` 和 `represents_user_view: false`。只有导入自己的已发布文章、演讲稿、公众号文章等个人输出时,才使用 `source_role: output` 和 `represents_user_view: true`。
37
+
38
+ ## 一定要用 Obsidian 吗?
39
+
40
+ 不一定。AIWiki 写的是普通 Markdown 和 frontmatter。Obsidian 只是好用的查看界面。
41
+
42
+ ## 一定要安装 Dataview 吗?
43
+
44
+ 不需要。Dataview 是可选增强。AIWiki 不会安装 Dataview,也不会修改 `.obsidian`。
45
+
46
+ ## 为什么要先同步助手?
47
+
48
+ 因为 AIWiki 的主路径是助手驱动。你把资料或问题交给助手,助手调用 AIWiki 命令。如果没有同步,助手可能只会直接翻文件,绕过 AIWiki 的入库、查询、状态和 lint 能力。
49
+
50
+ ## 从哪里开始?
51
+
52
+ 先把主 [README](../README.zh-CN.md) 里的 Quick Start 提示词复制给 AI 助手,再阅读 [使用指南](USAGE.zh-CN.md)。
53
+
54
+ ## 为什么我的助手还是直接翻文件?
55
+
56
+ 通常是助手没有加载 AIWiki skill,或者知识库根目录缺少 AIWiki 指导。运行:
57
+
58
+ ```bash
59
+ aiwiki agent sync --yes
60
+ aiwiki agent sync --path <workspace> --yes
61
+ aiwiki agent check --path <workspace> --json
62
+ ```
63
+
64
+ 必要时重启或重新加载助手。
65
+
66
+ ## 怎么查询知识库?
67
+
68
+ 对 AI 助手说:
69
+
70
+ ```text
71
+ AIWiki 里关于 <主题> 有什么?
72
+ ```
73
+
74
+ 助手应该调用:
75
+
76
+ ```bash
77
+ aiwiki context "<主题>"
78
+ ```
79
+
80
+ 人在终端里查询可以用:
81
+
82
+ ```bash
83
+ aiwiki query "<主题>"
84
+ ```
85
+
86
+ ## 怎么检查知识库?
87
+
88
+ 让助手运行:
89
+
90
+ ```bash
91
+ aiwiki lint --json
92
+ ```
93
+
94
+ 允许安全修复时:
95
+
96
+ ```bash
97
+ aiwiki lint --fix-empty-dirs --json
98
+ aiwiki lint --json
99
+ ```
100
+
101
+ ## 哪些能力不在范围内?
102
+
103
+ AIWiki 不是网页爬虫、微信读取器、浏览器插件、内置 LLM、向量数据库、RAG 替代品、Obsidian 插件、默认人工审核队列、多知识库管理器、RSS 工具或定时采集系统。
package/docs/README.md CHANGED
@@ -1,36 +1,47 @@
1
1
  # AIWiki Docs
2
2
 
3
- AIWiki 是一个 Agent-first 的本地知识库工具,用来把宿主 Agent 已经读取到的文章、网页正文或本地文本写入当前配置的知识库。
3
+ AIWiki is a local Markdown knowledge base for AI assistants.
4
4
 
5
- - 用户使用说明:[USAGE.md](USAGE.md)
6
- - Agent 对接协议:[AGENT_HANDOFF.md](AGENT_HANDOFF.md)
7
- - Obsidian + Dataview 方案:[OBSIDIAN_DATAVIEW_PLAN.md](OBSIDIAN_DATAVIEW_PLAN.md)
8
- - 架构图:[architecture.svg](architecture.svg)
9
- - 示例展示:[SHOWCASE.md](SHOWCASE.md)
10
- - 常见问题:[FAQ.md](FAQ.md)
11
- - 路线图:[ROADMAP.md](ROADMAP.md)
12
- - 开发记录:[development-log.md](development-log.md)
5
+ This docs folder keeps the public guides short and task-oriented. The main README is the product entry point; these files explain daily use, assistant handoff, examples, release checks, and roadmap boundaries.
13
6
 
14
- ## Examples
7
+ ## Start Here
15
8
 
16
- - `../examples/demo-run/` records the input files, commands, and CLI outputs from a regenerated demo.
17
- - `../examples/obsidian-vault-sample/` is a sample vault showing the current core-first artifact contract.
9
+ - [Main README](../README.md)
10
+ - [Chinese README](../README.zh-CN.md)
11
+ - [Usage Guide](USAGE.md)
12
+ - [FAQ](FAQ.md)
13
+ - [Agent Handoff](AGENT_HANDOFF.md)
14
+ - [Showcase](SHOWCASE.md)
15
+ - [Trial Feedback Template](TRIAL_FEEDBACK_TEMPLATE.md)
16
+ - [Roadmap](ROADMAP.md)
17
+ - [Release Notes](RELEASE.md)
18
18
 
19
- ## Quick Start
19
+ ## Chinese Docs
20
20
 
21
- ```bash
22
- npx @itradingai/aiwiki@latest setup
23
- aiwiki agent sync --yes
24
- aiwiki agent check --json
25
- aiwiki doctor
26
- ```
21
+ - [Chinese Docs Home](README.zh-CN.md)
22
+ - [Chinese Usage Guide](USAGE.zh-CN.md)
23
+ - [Chinese Agent Handoff](AGENT_HANDOFF.zh-CN.md)
24
+ - [Chinese FAQ](FAQ.zh-CN.md)
25
+ - [Chinese Showcase](SHOWCASE.zh-CN.md)
26
+ - [Chinese Roadmap](ROADMAP.zh-CN.md)
27
+ - [Chinese Release Notes](RELEASE.zh-CN.md)
28
+
29
+ ## Examples
30
+
31
+ - [`../examples/demo-run/`](../examples/demo-run/) records input files, CLI commands, and outputs from a regenerated run.
32
+ - [`../examples/obsidian-vault-sample/`](../examples/obsidian-vault-sample/) is a sample Markdown vault showing the current core-first artifact contract.
27
33
 
28
- 优先运行 `aiwiki agent sync --yes`,让 CLI 幂等同步本机 Codex、QClaw、OpenClaw、Claude Code 等宿主 Agent 对接文件。其他宿主 Agent 可使用 `aiwiki prompt agent` 输出通用协议,再安装成 skill 或粘贴到项目/会话说明里。之后把链接发给宿主 Agent:
34
+ ## Core Workflow
29
35
 
30
36
  ```text
31
- 入库 https://example.com/article
37
+ AI assistant reads a source
38
+ -> AIWiki writes local Markdown artifacts
39
+ -> assistant retrieves context with aiwiki context
40
+ -> aiwiki lint checks structure and consistency
32
41
  ```
33
42
 
34
- ## 重要边界
43
+ For public trials, keep the first loop to one source and one question: setup, ingest, inspect the run artifacts, query/context reuse, lint/doctor, and a short feedback note.
44
+
45
+ ## Important Boundaries
35
46
 
36
- CLI 不保证网页抓取成功。网页读取属于宿主 Agent。
47
+ AIWiki does not fetch webpages, call an LLM, install Obsidian plugins, run vector search, or manage multiple knowledge bases. The host assistant reads sources; AIWiki validates, writes, links, queries, and checks the local Markdown knowledge base.
@@ -0,0 +1,34 @@
1
+ # AIWiki 中文文档
2
+
3
+ AIWiki 是给 AI 助手使用的本地 Markdown 知识库。
4
+
5
+ 主 README 面向 GitHub / npm 的英文入口;中文用户可以从这里进入完整说明。
6
+
7
+ ## 从这里开始
8
+
9
+ - [中文 README](../README.zh-CN.md)
10
+ - [English README](../README.md)
11
+ - [使用指南](USAGE.zh-CN.md)
12
+ - [常见问题](FAQ.zh-CN.md)
13
+ - [Agent 接入说明](AGENT_HANDOFF.zh-CN.md)
14
+ - [案例展示](SHOWCASE.zh-CN.md)
15
+ - [路线图](ROADMAP.zh-CN.md)
16
+ - [发布说明](RELEASE.zh-CN.md)
17
+
18
+ ## 示例
19
+
20
+ - [`../examples/demo-run/`](../examples/demo-run/):记录输入、命令和 CLI 输出。
21
+ - [`../examples/obsidian-vault-sample/`](../examples/obsidian-vault-sample/):已经生成好的样例知识库。
22
+
23
+ ## 核心流程
24
+
25
+ ```text
26
+ AI 助手读取资料
27
+ -> AIWiki 写入本地 Markdown 知识文件
28
+ -> 助手用 aiwiki context 取回上下文
29
+ -> aiwiki lint 检查结构和一致性
30
+ ```
31
+
32
+ ## 重要边界
33
+
34
+ AIWiki 不抓网页、不调用 LLM、不自动安装 Obsidian 插件、不做向量检索、不管理多个知识库。宿主 AI 助手负责读取和理解资料;AIWiki 负责写入、链接、查询和检查本地 Markdown 知识库。
package/docs/RELEASE.md CHANGED
@@ -1,54 +1,84 @@
1
- # AIWiki 发布检查
1
+ # AIWiki Release Notes
2
2
 
3
- 发布前先确认本地工作区只包含本次发布需要的改动:
3
+ This document describes the release gate for maintainers.
4
+
5
+ AIWiki release work must prove the package locally, on the remote test server, through GitHub, and after npm publication.
6
+
7
+ ## Local Checks
8
+
9
+ Start from a clean, intentional worktree:
4
10
 
5
11
  ```bash
6
12
  git status --short --branch
13
+ npm test
7
14
  npm run release:check
8
15
  ```
9
16
 
10
- `release:check` 会执行测试、构建、`npm pack --dry-run`、CLI 版本一致性检查,以及一次临时知识库入库检查。
17
+ When package contents, docs, examples, or skill files changed, inspect:
18
+
19
+ ```bash
20
+ npm pack --dry-run
21
+ ```
11
22
 
12
- ## 版本号
23
+ The package should contain CLI runtime files, user documentation, examples, and packaged skill files only.
13
24
 
14
- 版本号以 `package.json` 为唯一来源。CLI 的 `aiwiki --version` 会运行时读取 `package.json`,不要在源码里另写一份版本常量。
25
+ ## Version
15
26
 
16
- 升级版本:
27
+ `package.json` is the version source. `aiwiki --version` reads it at runtime.
28
+
29
+ Patch bump by default:
17
30
 
18
31
  ```bash
19
32
  npm version patch --no-git-tag-version
20
33
  ```
21
34
 
22
- ## npm 发布
35
+ ## Pre-delivery Remote Test
36
+
37
+ Before pushing to GitHub or publishing to npm, build the exact local tarball and test it on the remote server.
38
+
39
+ Standard order:
40
+
41
+ ```text
42
+ local verification
43
+ -> version bump
44
+ -> local commit
45
+ -> npm pack
46
+ -> install local tarball on remote test server
47
+ -> run task-specific CLI smoke
48
+ -> GitHub push
49
+ -> GitHub Actions publish workflow
50
+ -> npm registry verification
51
+ -> post-publish remote sanity
52
+ ```
23
53
 
24
- 发布前必须先让同一个本地 tarball 在测试服务器通过 smoke test。顺序是:本地验证、版本号、提交、本地 `npm pack`、测试服务器安装 tarball 并跑 CLI smoke,然后才能 `git push` `npm publish`。
54
+ If the remote smoke fails, do not push and do not publish. Fix locally, rebuild, repack, and rerun the remote smoke.
25
55
 
26
- AIWiki 使用 npm Trusted Publishing。npm 发布应由 GitHub Actions 的 `.github/workflows/publish.yml` 完成,不依赖本机 `npm login`、OTP 或长期 `NPM_TOKEN`。
56
+ ## Publishing
27
57
 
28
- 远程测试和 GitHub push 都通过后,触发发布 workflow:
58
+ AIWiki uses npm Trusted Publishing. Publication should be done through GitHub Actions:
29
59
 
30
60
  ```bash
31
61
  gh workflow run publish.yml --repo iTradingAI/aiwiki
32
62
  gh run watch --repo iTradingAI/aiwiki
33
63
  ```
34
64
 
35
- 如需查看最近一次发布任务:
65
+ Check recent publish runs:
36
66
 
37
67
  ```bash
38
68
  gh run list --workflow publish.yml --repo iTradingAI/aiwiki --limit 5
39
69
  ```
40
70
 
41
- 发布后验证:
71
+ Verify the registry:
42
72
 
43
73
  ```bash
44
74
  npm view @itradingai/aiwiki version
45
75
  npm view @itradingai/aiwiki versions --json
46
76
  ```
47
77
 
48
- 如果 workflow 提示 Trusted Publishing / OIDC 权限问题,检查 npm 包设置里的 Trusted Publisher 是否指向 `iTradingAI/aiwiki` workflow 文件名 `publish.yml`,并确认 workflow 顶层包含 `permissions: id-token: write`。
78
+ If Trusted Publishing fails, verify the npm Trusted Publisher settings, repository name, workflow filename, and `id-token: write` permission.
49
79
 
50
- ## 包体积
80
+ ## Package Images
51
81
 
52
- npm 包只应包含 CLI 运行和用户文档所需文件。README 中的图片使用 GitHub raw 链接展示,`docs/assets/` 不进入 npm 包。
82
+ The README uses GitHub raw URLs for public images so GitHub and npm can render them without bundling `docs/assets/` into the npm package.
53
83
 
54
- 如果 `npm pack --dry-run` 输出里出现 `docs/assets/`,说明包内容配置回退了,需要先修复 `package.json.files`。
84
+ If `npm pack --dry-run` includes unexpected assets or private planning files, fix `package.json.files` before publishing.
@@ -0,0 +1,84 @@
1
+ # AIWiki 发布说明
2
+
3
+ 这份文档给维护者使用。
4
+
5
+ AIWiki 发布必须证明:本地通过、测试服务器通过、GitHub 推送通过、npm 发布后也通过。
6
+
7
+ ## 本地检查
8
+
9
+ 先确认工作区只包含本次发布需要的改动:
10
+
11
+ ```bash
12
+ git status --short --branch
13
+ npm test
14
+ npm run release:check
15
+ ```
16
+
17
+ 如果包内容、文档、示例或 skill 有变化,检查:
18
+
19
+ ```bash
20
+ npm pack --dry-run
21
+ ```
22
+
23
+ npm 包只应包含 CLI 运行文件、用户文档、示例和 packaged skill。
24
+
25
+ ## 版本
26
+
27
+ `package.json` 是版本来源。`aiwiki --version` 运行时读取它。
28
+
29
+ 默认 patch bump:
30
+
31
+ ```bash
32
+ npm version patch --no-git-tag-version
33
+ ```
34
+
35
+ ## 发布前远端测试
36
+
37
+ 推 GitHub 或发布 npm 之前,必须把同一个本地 tarball 放到测试服务器安装并跑 smoke。
38
+
39
+ 标准顺序:
40
+
41
+ ```text
42
+ 本地验证
43
+ -> 版本号
44
+ -> 本地提交
45
+ -> npm pack
46
+ -> 测试服务器安装本地 tarball
47
+ -> 跑任务对应 CLI smoke
48
+ -> GitHub push
49
+ -> GitHub Actions 发布 workflow
50
+ -> npm registry 验证
51
+ -> 发布后远端 sanity
52
+ ```
53
+
54
+ 如果远端 smoke 失败,不推 GitHub,不发布 npm。先本地修复,重新构建、打包、远端验证。
55
+
56
+ ## 发布
57
+
58
+ AIWiki 使用 npm Trusted Publishing。发布应通过 GitHub Actions 完成:
59
+
60
+ ```bash
61
+ gh workflow run publish.yml --repo iTradingAI/aiwiki
62
+ gh run watch --repo iTradingAI/aiwiki
63
+ ```
64
+
65
+ 查看最近发布任务:
66
+
67
+ ```bash
68
+ gh run list --workflow publish.yml --repo iTradingAI/aiwiki --limit 5
69
+ ```
70
+
71
+ 验证 npm:
72
+
73
+ ```bash
74
+ npm view @itradingai/aiwiki version
75
+ npm view @itradingai/aiwiki versions --json
76
+ ```
77
+
78
+ 如果 Trusted Publishing 失败,检查 npm Trusted Publisher、仓库名、workflow 文件名和 `id-token: write` 权限。
79
+
80
+ ## README 图片
81
+
82
+ README 使用 GitHub raw 图片链接,让 GitHub 和 npm 都能显示图片,同时避免把 `docs/assets/` 打进 npm 包。
83
+
84
+ 如果 `npm pack --dry-run` 出现意外资产或内部规划文件,先修复 `package.json.files`。
package/docs/ROADMAP.md CHANGED
@@ -1,43 +1,65 @@
1
1
  # AIWiki Roadmap
2
2
 
3
- AIWiki 对外只有一个项目,所以这里写的是能力路线,不是版本分层。
3
+ AIWiki has one public base product: a local Markdown knowledge base for AI assistants.
4
4
 
5
- 公开试用和长期运营的完整路线图见:
5
+ The roadmap keeps the base CLI focused. Advanced automation, crawling, multi-knowledge-base workflows, vector search, and team controls belong outside the base scope unless explicitly planned later.
6
6
 
7
- - [AIWiki 长期运营路线图](./20260607-aiwiki-long-term-operating-roadmap.md)
7
+ ## Current Focus
8
8
 
9
- ## 1. 更稳的入库体验
9
+ ### 1. First-use success
10
10
 
11
- - 更顺的 setup 引导
12
- - 更清晰的 doctor 检查
13
- - 更稳定的 Agent handoff
14
- - 更完整的 Obsidian 审阅入口
15
- - 更容易追查失败原因
16
- - 更稳定的公开样例,让新用户先看到真实生成的核心产物
11
+ - clearer assistant-first installation prompt
12
+ - safer `agent sync` and workspace guidance
13
+ - better `agent check` diagnostics
14
+ - a clean path from setup to first ingest to first query
17
15
 
18
- ## 2. 更完整的内容处理能力
16
+ ### 2. Reusable local knowledge
19
17
 
20
- - 更好用的资料卡
21
- - 更清楚的 Claim / Evidence 建议
22
- - 更顺手的选题候选
23
- - 更稳定的大纲生成
24
- - 更好复用的模板和输出结构
18
+ - clearer Wiki Entry quality signals
19
+ - better query/context explanations
20
+ - more examples showing writing, research, decision, and review workflows
21
+ - stronger distinction between external sources and user-authored output
25
22
 
26
- ## 3. 更适合团队和自动化部署
23
+ ### 3. Workspace health
27
24
 
28
- - 更多宿主 Agent 接入方式
29
- - 更适合团队协作的知识库目录
30
- - 更完善的部署说明
31
- - 更清楚的诊断和支持入口
32
- - 更好的服务化落地方式
25
+ - practical lint output for humans and assistants
26
+ - safe automatic fixes only when the fix is narrow and reversible
27
+ - better status and doctor guidance
28
+ - fewer empty optional artifacts in new workspaces
33
29
 
34
- ## 4. 服务层能力
30
+ ### 4. Public trial assets
35
31
 
36
- 公开软件还是 AIWiki 本身。后续如果要做商业化,重点会放在:
32
+ - short use cases users can complete in 5-10 minutes
33
+ - example vaults that match the current CLI behavior
34
+ - WeChat group feedback templates
35
+ - a queue policy that separates user pain from feature creep
37
36
 
38
- - 部署服务
39
- - 陪跑服务
40
- - 团队集成服务
41
- - 技术支持服务
37
+ ## Not in the Base Queue
42
38
 
43
- 这些都属于服务层,不是把 AIWiki 拆成多个软件版本。
39
+ The base AIWiki CLI is not currently planning:
40
+
41
+ - web crawling
42
+ - WeChat Official Account reading
43
+ - browser plugins
44
+ - vector search
45
+ - RAG-over-wiki
46
+ - multiple knowledge bases
47
+ - RBAC
48
+ - RSS or scheduled collection
49
+ - default manual review workflow
50
+ - automatic Dataview or Obsidian plugin installation
51
+
52
+ These may become separate service-layer, integration, or Pro-adjacent projects, but they should not blur the base README promise.
53
+
54
+ ## Operating Principle
55
+
56
+ AIWiki should stay boring in the right places:
57
+
58
+ ```text
59
+ assistant reads
60
+ AIWiki writes
61
+ Markdown stays local
62
+ context can be reused later
63
+ ```
64
+
65
+ The product gets better when the first ingest, first query, and first maintenance pass become more reliable.