@itradingai/aiwiki 0.2.20 → 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.
@@ -0,0 +1,65 @@
1
+ # AIWiki 路线图
2
+
3
+ AIWiki 基础版只有一个公开产品定位:给 AI 助手使用的本地 Markdown 知识库。
4
+
5
+ 路线图会保持基础 CLI 聚焦。高级自动化、爬取、多知识库、向量检索和团队控制,不应混入基础版承诺,除非后续单独规划。
6
+
7
+ ## 当前重点
8
+
9
+ ### 1. 首次使用成功
10
+
11
+ - 更清楚的 AI 助手安装提示词
12
+ - 更安全的 `agent sync` 和知识库根指导
13
+ - 更清楚的 `agent check` 诊断
14
+ - 从 setup 到第一次入库、第一次查询的完整路径
15
+
16
+ ### 2. 本地知识复用
17
+
18
+ - 更清楚的 Wiki Entry 质量信号
19
+ - 更好解释 query/context 的结果
20
+ - 更多写作、研究、决策、回顾场景案例
21
+ - 更明确区分外部资料和用户个人输出
22
+
23
+ ### 3. 知识库健康
24
+
25
+ - 给人和 AI 助手都能读懂的 lint 输出
26
+ - 只在窄范围、可逆时提供自动安全修复
27
+ - 更好的 status 和 doctor 引导
28
+ - 新知识库减少空洞可选产物
29
+
30
+ ### 4. 公开试用资产
31
+
32
+ - 5-10 分钟能完成的小任务
33
+ - 与当前 CLI 行为一致的样例知识库
34
+ - 微信群反馈模板
35
+ - 区分用户痛点和功能蔓延的队列规则
36
+
37
+ ## 不进入基础版队列
38
+
39
+ 基础版 AIWiki 当前不规划:
40
+
41
+ - 网页爬虫
42
+ - 微信公众号读取
43
+ - 浏览器插件
44
+ - 向量检索
45
+ - RAG-over-wiki
46
+ - 多知识库
47
+ - RBAC
48
+ - RSS 或定时采集
49
+ - 默认人工审核流程
50
+ - 自动安装 Dataview 或 Obsidian 插件
51
+
52
+ 这些能力可以进入服务层、集成项目或 Pro 相邻项目,但不应模糊基础版 README 的承诺。
53
+
54
+ ## 运营原则
55
+
56
+ AIWiki 应该在关键地方保持稳定朴素:
57
+
58
+ ```text
59
+ 助手负责阅读
60
+ AIWiki 负责写入
61
+ Markdown 保持本地
62
+ 上下文以后还能复用
63
+ ```
64
+
65
+ 产品进步的重点,是让第一次入库、第一次查询和第一次维护更可靠。
package/docs/SHOWCASE.md CHANGED
@@ -1,25 +1,51 @@
1
- # AIWiki 示例展示
1
+ # AIWiki Showcase
2
2
 
3
- 这页不是讲概念,直接看一次完整跑通会发生什么。
3
+ This page shows what a real AIWiki run creates.
4
4
 
5
- ## 示例 1:一篇文章进知识库
5
+ ## 10-minute Trial Walkthrough
6
6
 
7
- ### 用户输入
7
+ This is the public-trial happy path:
8
8
 
9
9
  ```text
10
- 入库 https://example.com/article
10
+ 1. Create a temporary knowledge base.
11
+ 2. Ingest one readable source.
12
+ 3. Inspect the generated run summary, Source Card, and Wiki Entry.
13
+ 4. Ask the knowledge base one question.
14
+ 5. Run lint/doctor for workspace health.
15
+ 6. Send feedback with the trial template.
11
16
  ```
12
17
 
13
- ### 宿主 Agent 做的事
18
+ Commands for a local-file trial:
14
19
 
15
- 1. 读取网页正文或可访问内容。
16
- 2. 组装 `aiwiki.agent_payload.v1`,并尽量提供 `analysis` 或 `wiki_entry`。
17
- 3. 调用 `aiwiki ingest-agent --stdin`。
18
- 4. AIWiki 的处理结果回复给用户。
20
+ ```bash
21
+ aiwiki setup --path ./aiwiki-trial --yes
22
+ aiwiki doctor --path ./aiwiki-trial
23
+ aiwiki ingest-file --file ./my-first-source.md --path ./aiwiki-trial
24
+ aiwiki query "my topic" --path ./aiwiki-trial
25
+ aiwiki context "my topic" --path ./aiwiki-trial
26
+ aiwiki lint --json --path ./aiwiki-trial
27
+ ```
28
+
29
+ For a URL trial, the assistant reads the URL first, then calls `aiwiki ingest-agent` with the content it read.
30
+
31
+ ## Scenario 1: Ingest an Article
32
+
33
+ User message:
34
+
35
+ ```text
36
+ Ingest this into AIWiki:
37
+ https://example.com/article
38
+ ```
39
+
40
+ Assistant work:
19
41
 
20
- ### AIWiki 会写什么
42
+ 1. read the source
43
+ 2. create an `aiwiki.agent_payload.v1` payload
44
+ 3. provide `analysis` or `wiki_entry` when possible
45
+ 4. call `aiwiki ingest-agent --stdin`
46
+ 5. report the generated files
21
47
 
22
- 核心产物优先看这些:
48
+ Core output:
23
49
 
24
50
  ```text
25
51
  09-runs/<run-id>/payload.json
@@ -32,7 +58,7 @@
32
58
  05-wiki/source-knowledge/<slug>.md
33
59
  ```
34
60
 
35
- 只有在宿主 Agent 提供了对应内容,或 payload 明确请求时,才会出现可选增强产物:
61
+ Optional output appears only when the assistant provides matching content:
36
62
 
37
63
  ```text
38
64
  09-runs/<run-id>/creative-assets.md
@@ -44,50 +70,38 @@
44
70
  08-outputs/outlines/
45
71
  ```
46
72
 
47
- 仓库里的 `examples/obsidian-vault-sample/` 同时包含一个普通本地文件入库和一个 enriched Agent payload 入库。前者只生成核心产物,后者因为包含 claims、topic candidates、creative assets 和 outline 请求,所以生成可选增强目录。
48
-
49
- ### 用户在 Obsidian 里怎么看
50
-
51
- - 打开 `dashboards/AIWiki Home.md`
52
- - 查看 `05-wiki/source-knowledge/<slug>.md`
53
- - 从 Wiki 条目回到资料卡、原文和处理记录;如果这次入库生成了选题、大纲或 Claim,再继续查看对应可选产物
54
- - 需要结构检查时运行 `aiwiki lint`
73
+ ## Scenario 2: The Source Cannot Be Read
55
74
 
56
- ## 示例 2:网页没读到正文
75
+ Some pages require login or cannot be accessed by the assistant.
57
76
 
58
- 有些页面需要登录,或者宿主 Agent 暂时拿不到正文。这个时候 AIWiki 也不会装死,它会把失败原因记录下来。
59
-
60
- ### 用户输入
77
+ AIWiki should still record the attempt:
61
78
 
62
79
  ```text
63
- 入库 https://example.com/locked-page
80
+ 09-runs/<run-id>/payload.json
81
+ 09-runs/<run-id>/processing-summary.md
64
82
  ```
65
83
 
66
- ### 结果
84
+ The failure reason is preserved, and the user can retry later when the assistant can access the source.
85
+
86
+ ## Scenario 3: Reuse the Knowledge Later
87
+
88
+ User message:
67
89
 
68
90
  ```text
69
- 09-runs/<run-id>/payload.json
70
- 09-runs/<run-id>/processing-summary.md
91
+ What does AIWiki know about AI agents?
71
92
  ```
72
93
 
73
- ### 你会看到什么
94
+ Assistant command:
74
95
 
75
- - 失败原因会写进 `processing-summary.md`
76
- - 用户仍然能在 Obsidian 里看到这次处理记录
77
- - 下次只要宿主 Agent 能读到正文,就可以重新入库
78
-
79
- ## 这个示例页想说明什么
96
+ ```bash
97
+ aiwiki context "AI agents"
98
+ ```
80
99
 
81
- - AIWiki 不是网页爬虫本体
82
- - 宿主 Agent 负责读取内容
83
- - 宿主 Agent 负责高质量总结和知识提炼
84
- - AIWiki 负责稳定写入本地知识库
85
- - AIWiki 会创建 Wiki Entry;没有 Agent 分析字段时只生成可追溯脚手架
86
- - 结果始终可追踪、可复盘、可继续写作
100
+ The assistant should use the returned JSON, including match reasons and quality signals, before answering.
87
101
 
88
- ## 本仓库样例
102
+ ## Sample Files
89
103
 
90
- - `examples/demo-run/`:输入、命令和 CLI 输出。
91
- - `examples/obsidian-vault-sample/`:已经生成好的样例知识库。
104
+ - [`../examples/demo-run/`](../examples/demo-run/) records input files, commands, and CLI outputs.
105
+ - [`../examples/obsidian-vault-sample/`](../examples/obsidian-vault-sample/) is a generated sample vault.
92
106
 
93
- 这个样例不依赖爬虫、向量库、RAG Pro 自动化;它只展示基础 CLI 的真实落盘结果。
107
+ The sample does not rely on crawling, vector search, RAG-over-wiki, or Pro automation. It shows the real base CLI file contract.
@@ -0,0 +1,81 @@
1
+ # AIWiki 案例展示
2
+
3
+ 这页直接展示一次真实 AIWiki 运行会生成什么。
4
+
5
+ ## 场景 1:入库一篇文章
6
+
7
+ 用户说:
8
+
9
+ ```text
10
+ 把这个资料入库到 AIWiki:
11
+ https://example.com/article
12
+ ```
13
+
14
+ 助手会:
15
+
16
+ 1. 读取资料
17
+ 2. 生成 `aiwiki.agent_payload.v1`
18
+ 3. 能理解时提供 `analysis` 或 `wiki_entry`
19
+ 4. 调用 `aiwiki ingest-agent --stdin`
20
+ 5. 汇报生成文件
21
+
22
+ 核心产物:
23
+
24
+ ```text
25
+ 09-runs/<run-id>/payload.json
26
+ 09-runs/<run-id>/raw.md
27
+ 09-runs/<run-id>/source-card.md
28
+ 09-runs/<run-id>/wiki-entry.md
29
+ 09-runs/<run-id>/processing-summary.md
30
+ 02-raw/articles/<slug>.md
31
+ 03-sources/article-cards/<slug>.md
32
+ 05-wiki/source-knowledge/<slug>.md
33
+ ```
34
+
35
+ 可选产物只在助手提供对应内容时出现:
36
+
37
+ ```text
38
+ 09-runs/<run-id>/creative-assets.md
39
+ 09-runs/<run-id>/topics.md
40
+ 09-runs/<run-id>/draft-outline.md
41
+ 04-claims/_suggestions/
42
+ 06-assets/_suggestions/
43
+ 07-topics/ready/
44
+ 08-outputs/outlines/
45
+ ```
46
+
47
+ ## 场景 2:资料暂时读不到
48
+
49
+ 有些页面需要登录,或助手暂时无法访问正文。
50
+
51
+ AIWiki 仍然应该记录这次尝试:
52
+
53
+ ```text
54
+ 09-runs/<run-id>/payload.json
55
+ 09-runs/<run-id>/processing-summary.md
56
+ ```
57
+
58
+ 失败原因会被保留,之后助手能读到正文时可以重新入库。
59
+
60
+ ## 场景 3:以后复用知识
61
+
62
+ 用户问:
63
+
64
+ ```text
65
+ AIWiki 里关于 AI Agent 有什么?
66
+ ```
67
+
68
+ 助手应该调用:
69
+
70
+ ```bash
71
+ aiwiki context "AI Agent"
72
+ ```
73
+
74
+ 助手回答前应读取返回 JSON 里的匹配原因和质量信号。
75
+
76
+ ## 样例文件
77
+
78
+ - [`../examples/demo-run/`](../examples/demo-run/):输入、命令和 CLI 输出。
79
+ - [`../examples/obsidian-vault-sample/`](../examples/obsidian-vault-sample/):已经生成好的样例知识库。
80
+
81
+ 这个样例不依赖爬虫、向量检索、RAG-over-wiki 或 Pro 自动化,只展示基础 CLI 的真实落盘结果。
@@ -0,0 +1,49 @@
1
+ # AIWiki Trial Feedback Template
2
+
3
+ Use this template after a first public trial. It is intentionally small so feedback can be pasted into a WeChat group, issue, note, or planning document without adding a new CLI command.
4
+
5
+ ## Basic Info
6
+
7
+ - AIWiki version:
8
+ - Operating system:
9
+ - Assistant used:
10
+ - Knowledge base path type: temporary / real workspace / project folder
11
+
12
+ ## First-use Result
13
+
14
+ - Setup result: success / blocked / unclear
15
+ - First source type: URL / local file / pasted note / failed fetch
16
+ - Ingest result: success / failed / recorded failure only
17
+ - Query/context result: useful / partially useful / not useful / no match
18
+ - Lint/doctor result: clean / actionable warning / confusing warning / error
19
+
20
+ ## What Worked
21
+
22
+ - The clearest step was:
23
+ - The most useful generated file was:
24
+ - One thing I can reuse later:
25
+
26
+ ## What Was Confusing
27
+
28
+ - Installation or Node.js issue:
29
+ - Assistant did not call AIWiki commands:
30
+ - I could not find generated files:
31
+ - Query/context answer was unclear:
32
+ - Lint/doctor output was unclear:
33
+ - Other:
34
+
35
+ ## One Practical Scenario
36
+
37
+ I would use AIWiki for:
38
+
39
+ ```text
40
+ Example: saving article research before writing a WeChat post.
41
+ ```
42
+
43
+ ## Next Action
44
+
45
+ - Keep using as-is
46
+ - Needs documentation improvement
47
+ - Needs assistant guidance improvement
48
+ - Needs CLI behavior improvement
49
+ - Not a fit for my workflow