fix(docs): 中文搜不到英文注释的 API —— 补关键词层

## 问题(实测)

SDK 里 100 个有摘要的符号中 **66 个是英文注释**,例如:

    RegisterTool registers a tool that the LLM can call.

于是搜「注册工具」——中文受众最自然的问法——**RegisterTool 得分 0,一条都搜不到**。
更糟的是逐字匹配把噪声顶上来了:搜「注册工具」返回 24 条,排第一的是
`SetToolBlocks`(描述里有「工具」二字),`RegisterTool` 根本不在列表里。

## 修法

不改源码注释(那会让代码与文档脱节),而是在检索索引上加一层**人工标注的
中文功能词**:`tools/apidoc/keywords.json`。

- `rules`:按符号名前缀/子串批量覆盖(`Register*` 全带「注册」,`*Memory*` 带「记忆」)
- `symbols`:逐符号补充(重点 API、或规则覆盖不到的)

词只进 `api-index.json` 的 `g` 字段,**不影响页面展示**;检索结果里会显示
(「为何命中」),读者能理解排序依据。

同时把中文逐字匹配从主信号降为**弱信号**(要求 60% 以上字符命中)——
它正是噪声来源:凡是含「工具」二字的说明都会被「注册工具」匹上。

## 结果

| 查询 | 修改前 | 修改后 |
|---|---|---|
| 注册工具 | 24 条,RegisterTool 缺席 | **7 条,RegisterTool 第一** |
| 崩溃 | 1 条 | 2 条(SetAutoRestart + AutoRestart)|
| InjectText | 7 条(含重复)| 6 条 |
| memory.recall | 1 条 | 1 条(不变)|

顺带修掉索引重复:接口会同时作为 `type` 符号与接口本身被加两次
(`MemoryAPI` 等 11 个各重复一条)。现在接口只走接口那条路径,索引 200 → 189 条。

keywords.json 是**可选**的:读不到只警告不中断,检索退化为原行为。
This commit is contained in:
JianFeeeee
2026-09-24 13:00:32 +08:00
parent 0a6e2b7dc4
commit 9b6abe1b73
12 changed files with 1368 additions and 486 deletions

View File

@ -268,18 +268,6 @@ const ContextPolicyPrune
<small>`plugin.go:45`</small>
### `IOInjector`
```go
type IOInjector interface { InjectInterruptText(source, channel, text string) InjectText(source, channel, text string) InjectTextNoMemory(source, channel, text string) // I …
```
IOInjector provides methods for injecting input and interrupts into the agent pipeline.
All methods accept (source, channel) where channel is the target output channel
for routing the agent's response.
<small>`plugin.go:220`</small>
### `PluginSDK.InjectInputMedia`
```go

View File

@ -44,18 +44,6 @@ EventHandler processes a system event.
<small>`plugin.go:273`</small>
### `EventSubscriber`
```go
type EventSubscriber interface { Subscribe(eventType EventType, handler EventHandler) func() }
```
EventSubscriber allows plugins to subscribe to kernel events.
This is a restricted interface: plugins can subscribe but the kernel
controls which events are delivered.
<small>`plugin.go:278`</small>
### `EventType`
```go

View File

@ -89,16 +89,6 @@ AutoRestart 返回插件是否允许自动重启。
<small>`plugin.go:791`</small>
### `Plugin`
```go
type Plugin interface { Name() string Start(sdk *PluginSDK) error Stop() error }
```
Plugin is the interface every plugin must implement.
<small>`plugin.go:13`</small>
### `PluginSDK.PluginMgr`
```go
@ -110,17 +100,6 @@ May be nil if the host did not wire it.
<small>`plugin.go:653`</small>
### `PluginMgrAPI`
```go
type PluginMgrAPI interface { // ReloadOne 重载单个插件(停止后重新加载)。 ReloadOne(name string) error // ListLoadedPlugins 列出已加载插件。 ListLoa …
```
PluginMgrAPI 提供插件管理能力(外部插件可调用)。
由 bridge 注入 dispatch 实现,走 C ABI CORE_PLUGIN_RELOAD_ONE 等。
<small>`plugin.go:284`</small>
### `PluginSDK.PluginName`
```go

View File

@ -48,13 +48,3 @@ LLM returns the LLM provider API (may be nil if not available).
<small>`plugin.go:432`</small>
### `LLMAPI`
```go
type LLMAPI interface { ListSources() []string SetSource(name string) error CurrentSource() string }
```
LLMAPI provides access to the LLM provider manager.
<small>`llm.go:4`</small>

View File

@ -238,16 +238,6 @@ DocMemory returns the document memory API (may be nil if not available).
<small>`plugin.go:418`</small>
### `DocMemoryAPI`
```go
type DocMemoryAPI interface { Query(text string, topK int) []*Doc Insert(doc *Doc) error // InsertWithMedia 写入文档并关联媒体。attachments 里带 Data 的会落进 …
```
DocMemoryAPI provides access to the document vector store.
<small>`memory.go:74`</small>
### `Entity`
```go
@ -258,22 +248,6 @@ Entity represents a named entity in the knowledge graph.
<small>`memory.go:13`</small>
### `PluginSDK.Knowledge`
```go
func (s *PluginSDK) Knowledge() KnowledgeAPI
```
Knowledge returns the knowledge store API (may be nil if not available).
**示例插件里的真实用法**
| 插件 | 位置 | 代码 |
|---|---|---|
| [`recoverydiag`](../examples/index.md#recoverydiag) | `example/recoverydiag/plugin.go:978` | `if p.sdk != nil && p.sdk.Knowledge() != nil {` |
<small>`plugin.go:425`</small>
### `Knowledge`
```go
@ -290,15 +264,21 @@ Knowledge represents a knowledge entry.
<small>`knowledge.go:11`</small>
### `KnowledgeAPI`
### `PluginSDK.Knowledge`
```go
type KnowledgeAPI interface { Search(query string, topK int) ([]*Knowledge, error) Add(name, content string) error List() ([]string, error) }
func (s *PluginSDK) Knowledge() KnowledgeAPI
```
KnowledgeAPI provides access to the knowledge store.
Knowledge returns the knowledge store API (may be nil if not available).
<small>`knowledge.go:4`</small>
**示例插件里的真实用法**
| 插件 | 位置 | 代码 |
|---|---|---|
| [`recoverydiag`](../examples/index.md#recoverydiag) | `example/recoverydiag/plugin.go:978` | `if p.sdk != nil && p.sdk.Knowledge() != nil {` |
<small>`plugin.go:425`</small>
### `MediaAttachment`
@ -330,16 +310,6 @@ Memory returns the graph memory API (may be nil if not available).
<small>`plugin.go:404`</small>
### `MemoryAPI`
```go
type MemoryAPI interface { Recall(query []string, depth int) ([]Entity, []Relation, error) Commit(triples []Triple) error Introspect() (map[string]interface{}, error) Merg …
```
MemoryAPI provides access to the graph memory (entity-relation store).
<small>`memory.go:4`</small>
### `PersonProfile`
```go
@ -370,17 +340,6 @@ Social returns the social graph API (may be nil if not available).
<small>`plugin.go:439`</small>
### `SocialAPI`
```go
type SocialAPI interface { GetPerson(name string) (*PersonProfile, error) GetTrait(name, trait string) (string, bool) GetRelations(name string) ([]SocialRelation, error) G …
```
SocialAPI provides read-only access to the social graph (person profiles and relationships).
External plugins can query person traits and social networks but cannot modify them.
<small>`memory.go:100`</small>
### `SocialRelation`
```go
@ -409,16 +368,6 @@ TextMemory returns the text memory API (may be nil if not available).
<small>`plugin.go:411`</small>
### `TextMemoryAPI`
```go
type TextMemoryAPI interface { Append(evt TextEvent) error }
```
TextMemoryAPI provides access to chronological text event storage.
<small>`memory.go:43`</small>
### `Triple`
```go

View File

@ -195,11 +195,3 @@ sett 在 New 时一次性写入且无 setter,故不需要加锁。
<small>`plugin.go:401`</small>
### `SettingsAPI`
```go
type SettingsAPI interface { // Get reads the plugin's own config value (config_<name> table). Get(key string) (interface{}, error) // Set writes a config value to the plugi …
```
<small>`settings.go:3`</small>

File diff suppressed because it is too large Load Diff

View File

@ -65,6 +65,17 @@
else if (name.indexOf(ql) === 0) s += 600;
else if (name.indexOf(ql) > 0) s += 350;
// 中文检索关键词(keywords.json 产出,字段 g)。
// 为什么需要:SDK 里 66/100 个符号是英文注释(`RegisterTool registers a
// tool that the LLM can call.`),懂中文的人搜「注册工具」会一条都找不到。
// 关键词命中给较高权重(仅次于名称精确命中),因为它就是为「按功能找」准备的。
var kws = item.g || [];
for (var ki = 0; ki < kws.length; ki++) {
var kw = String(kws[ki]).toLowerCase();
if (kw === ql) { s += 480; break; }
if (kw.indexOf(ql) >= 0) { s += 300; break; }
}
// 限定符:PluginSDK.RegisterTool / IOInjector.InjectText。
// 额外支持「去掉 API/SDK 后缀」与「去掉点号」两种写法,
// 因为读者习惯写 `memory.recall`(RPC 名),而 Go 名是 `MemoryAPI.Recall`。
@ -94,13 +105,17 @@
}
// 中文按字匹配:中文没有词边界,逐字命中比整串更实用。
// 注意只把它当作**弱信号**:光靠逐字会把「注册工具」匹到凡是含「工具」
// 字样的任何东西(实测 ContextPolicyNone 的说明里有「工具调用」也会命中)。
// 所以阈值卡在 60% 以上才算有效命中。
if (/[\u4e00-\u9fa5]/.test(q)) {
var hit = 0;
var hay = desc + " " + kws.join(" ");
for (var i = 0; i < q.length; i++) {
if (desc.indexOf(q[i]) >= 0) hit++;
if (hay.indexOf(q[i]) >= 0) hit++;
}
if (hit === q.length) s += 100; // 全部字都出现
else s += hit * 8;
var ratio = hit / q.length;
if (ratio >= 0.6) s += Math.round(hit * 6);
}
// 公开 API 略优先于「仅内置」——后者通常是噪声。
@ -173,6 +188,10 @@
var d = el("span", "api-desc", it.d);
li.appendChild(d);
}
// 关键词是给检索用的;显示出来能让读者明白“为什么这条被匹配到”。
if (it.g && it.g.length) {
li.appendChild(el("span", "api-kw", it.g.slice(0, 6).join(" · ")));
}
if (it.f) {
li.appendChild(el("span", "api-loc", it.f + (it.l ? ":" + it.l : "")));
}

View File

@ -71,6 +71,14 @@
color: var(--md-default-fg-color--light);
}
/* 「为何命中」的中文关键词(keywords.json 产出)。 */
.api-kw {
flex-basis: 100%;
font-size: 0.7rem;
color: var(--md-accent-fg-color);
opacity: 0.9;
}
.api-loc {
flex-basis: 100%;
font-size: 0.68rem;