Files
HomeAgent/third_party/homeagent-sdk/example/rss/plugin.go
JianFeeeee d1959cbe80 feat(core): 注入行为的记忆/裁剪标志位落地 + jieba 词库内嵌 + Windows 改走 WSL
配套 SDK 提交:homeagent-sdk ba49dfd(公开 API 纯追加,无签名变更)。
本仓第三方的库镜像同步至该版本,以保证全新 clone 能编译。

## 1. 注入标志位(内核侧)

- 7 条注入路径(排队/中断/同步 × 纯文本/带媒体 + 旧 NoMem 变体)解析并转发
  no_memory / context_policy / cleaner_name;策略在入口**校验**,
  非法值报错而不是静默降级成 none(降级会让调用方以为自己声明的裁剪在生效)。
- 新增 validateContextPolicy(与 tool.register 同一套规则)与 pubSdkInjectOpts。
- input.register 不再手写字段白名单重建 ChannelDef,改为整体传递 + 补 ContextPolicy。
- io 层:applyInjectOpts 把标志位写进事件 payload,仅非零时写
  (零值与旧 payload 逐字节一致,事件订阅方与旧内核都不受影响)。
- ioAdapter / procCore / internal-sdk 别名补齐六个 *Opts 实现。

## 2. 修掉「输入无条件裁剪」这个真缺陷

eventloop 此前对**每条非中断输入**都调 `context.Prune(...)`:破坏性(低相关事件被
归档移出上下文)且无法从调用点看出是谁触发的。改为 pruneOnInput/pruneDeclared:

  优先级:注入点声明(payload.context_policy)> 通道声明(ChannelDef.ContextPolicy)
          > 默认**不裁剪**

查询向量仍取清洗后的内容;新增 cleanInputFor 解析清洗文本,优先级为
注入点声明的 cleaner(cleaner_name)> 按 source 查到的通道 cleaner > 原文,
名字查不到时**记日志再回退**(注入是 fire-and-forget,插件看不到错误,
至少要在内核日志留下「你声明的清洗没生效」的痕迹)。

## 3. jieba 词库内嵌(修「猜 GOMODCACHE → 静默失效」)

原 jiebaDictDir() 去猜 GOMODCACHE/GOPATH/~/go/pkg/mod,部署机上通常没有 Go 模块
缓存 → GetJieba() 返回 nil → 分词/关键词提取/NLP 依存解析(进而 doc→graph 三元组
抽取)/静态词向量 tokenizer **一律静默返回空列表**,只有一行日志。本机看起来正常
只因开发机与生产机重合、恰好有那份缓存。

现在词库随二进制分发:internal/memory/jiebadict/ 5 文件约 11.6MB + go:embed,
按**内容哈希**命名缓存目录落盘(词库升级不复用旧文件),已齐全则跳过写入。
模块缓存降为兜底。homed 体积 32MB。

顺带确认(并有测试佐证):gojieba 的 Tag() 不需要 pos_dict/ 目录——
cppjieba 的 PosTagger 从主词典每行的词性列取 tag。

## 4. homed 放弃 Windows 原生,改走 WSL2

插件体系依赖「继承的 fd」+「统一共享内存区的段内偏移解引用」,Windows 既无 fd
继承语义,其句柄模型也无法表达后者;强行适配等于再维护一套平台专属 ABI
(C ABI 时代三套 ABI 并存曾导致改写型插件在某平台静默失效)。

- cmd/homed/platform_{windows,other}.go:原生 Windows 启动即拒绝并打印 WSL2 指引。
- internal/plugin/proc/shmalloc_windows.go:allocShm 直接返回「请用 WSL2」,
  **不返回半可用的段**(与 shmalloc_other.go 同风格:未支持平台显式报错);
  procEnvForShm 返回 nil。顺手修掉两处长期编译错误
  (cryptorand→rand、h.evData→h.unified.evtData),使 GOOS=windows 至少能编译。
  注:homed 本就编不出 Windows——internal/memory 依赖 cgo-only 的 gojieba。
- deploy/packaging/installer.nsi:不再安装 homed.exe/initconfig.exe,改为携带
  **linux payload** 并调用新的 install-via-wsl.ps1;退出码 20/21 表示
  「需先装 WSL/发行版」,走指引而非报错。
- deploy/packaging/windows/install-via-wsl.ps1(新):检测 WSL → 引导安装 →
  确保 WSL2 → 送包进发行版 → 在 WSL 内按 Linux 方式安装。**复用 Linux 包与
  linux/setup.sh**,不另写一套安装逻辑;落点与 deb 布局统一
  (/usr/bin/homed + /usr/lib/homeagent/setup.sh)。
- deploy/packaging/linux/setup.sh:API Key 允许 HOMEAGENT_API_KEY 覆盖
  (否则安装器界面显示一份、config.db 里另一份 → 登录不上)。
- deploy/packaging/build.sh:windows 目标只构建 waiter + gui,并新增
  stage_linux_payload 把 Linux 包暂存给安装器;homed/initconfig 在 windows
  目标下明确拒绝。

## 5. 插件调用点统一写明意图

- webui 的 OpenAI 兼容端点(固定提示词模板)→ InjectTextSyncNoMemory。
- agentcli 的 5 处纯状态通知(已启动/超时/执行结束/进程退出/读取结束)→ NoMemory;
  **带输出**的 2 处(定时反馈、有新输出)刻意保留记忆并注明理由。
- timer 的定时提醒 → NoMemory(中断本来也隐含 NoMemory,这里是写明意图)。

## 6. 版本

meta.Version 仍为 1.2.0(main 是下一个未发布中版本);
SDKCompatibleVersion 1.1.0 → **1.2.0**(本内核已实现 SDK 1.2.0 全部新增方法)。

## 测试

- core:默认不裁剪(无声明/none/空)、通道 opt-in、注入点双向覆盖通道、
  nil context/io 安全、cleaner 优先级与未知名回退。
- io:零值 opts 与历史 payload 逐键相同;text/中断/媒体三类注入标志位都落到
  payload;旧方法仍生效。
- proc:validateContextPolicy 只接受 ""/none/prune,报错含位置与实际值;
  **跨进程** e2e——testdata 插件经 io.injectText 送出三个标志位,断言它们穿过 RPC
  到达内核。
- memory:模块缓存不可见时内嵌词库仍可用(分词与 POS 内容词均非空)、
  落盘幂等、内容哈希稳定。

验证:go build ./... / go vet ./... / go vet -tags onnxruntime ./...
      go test -short ./internal/memory/... ./internal/nlp/... ./internal/plugin/...
      ./internal/agent/{core,io}/... ./pkg/...
2026-09-11 20:31:50 +08:00

501 lines
11 KiB
Go
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

package main
import (
"encoding/json"
"fmt"
"net/http"
"os"
"path/filepath"
"sort"
"strconv"
"strings"
"sync"
"time"
sdk "gitcode.com/JianFeeeee/homeagent-sdk/sdk"
"github.com/mmcdole/gofeed"
)
const injectDedupWindow = 5 * time.Minute
type FeedSub struct {
URL string `json:"url"`
Title string `json:"title"`
AddedAt string `json:"added_at"`
Interval int `json:"interval"`
}
type Plugin struct {
name string
sdk *sdk.PluginSDK
client *http.Client
fp *gofeed.Parser
dataDir string
mu sync.RWMutex
feeds []FeedSub
seenGUIDs map[string]bool
injected map[string]time.Time
stopCh chan struct{}
stopOnce sync.Once
wg sync.WaitGroup
pollTicker *time.Ticker
}
func NewPluginFactory(name string, config map[string]interface{}) (sdk.Plugin, error) {
return &Plugin{name: name}, nil
}
func (p *Plugin) Name() string { return p.name }
func getSetting[T string | int64 | float64](s sdk.SettingsAPI, key string, fallback T) T {
v, err := s.Get(key)
if err != nil || v == nil {
return fallback
}
switch any(fallback).(type) {
case string:
if sv, ok := v.(string); ok {
return any(sv).(T)
}
case int64:
switch val := v.(type) {
case float64:
return any(int64(val)).(T)
case string:
if n, err := strconv.ParseInt(val, 10, 64); err == nil {
return any(n).(T)
}
}
case float64:
switch val := v.(type) {
case float64:
return any(val).(T)
case string:
if n, err := strconv.ParseFloat(val, 64); err == nil {
return any(n).(T)
}
}
}
return fallback
}
func readArg(args map[string]interface{}, key string) string {
if v, ok := args[key]; ok && v != nil {
if s, ok := v.(string); ok {
return s
}
}
return ""
}
func readArgInt(args map[string]interface{}, key string, fallback int) int {
if v, ok := args[key]; ok && v != nil {
switch n := v.(type) {
case float64:
return int(n)
case int64:
return int(n)
}
}
return fallback
}
func (p *Plugin) Start(s *sdk.PluginSDK) error {
s.SetAutoRestart(true)
p.sdk = s
p.client = &http.Client{Timeout: 30 * time.Second}
p.fp = gofeed.NewParser()
p.stopCh = make(chan struct{})
p.seenGUIDs = make(map[string]bool)
p.injected = make(map[string]time.Time)
p.feeds = []FeedSub{}
dataDirVal, err := s.Settings().GetCore("core.daemon.data_dir")
if err != nil || dataDirVal == "" {
dataDirVal = "."
}
p.dataDir = filepath.Join(fmt.Sprint(dataDirVal), "rss")
if err := os.MkdirAll(p.dataDir, 0755); err != nil {
fmt.Printf("[%s] mkdir %s: %v\n", p.name, p.dataDir, err)
}
p.loadData()
// 卸载(删除)时清理订阅数据目录;重载不触发
s.RegisterOnRemoveHandler(p.cleanupData)
s.Settings().RegisterDef(sdk.ConfigDef{
Key: "poll_interval", Default: "30", Type: "string",
DisplayName: "Poll Interval", Description: "Default polling interval in minutes (default: 30)",
Category: "rss",
})
tp := p.name + "_"
s.RegisterTool(tp+"subscribe", sdk.ToolDef{
Name: tp + "subscribe", Description: "Subscribe to an RSS/Atom feed URL",
NoMemory: true,
Parameters: map[string]interface{}{
"type": "object",
"properties": map[string]interface{}{
"url": map[string]interface{}{"type": "string", "description": "Feed URL"},
"interval": map[string]interface{}{"type": "integer", "description": "Poll interval in minutes (default: 30, minimum: 5)"},
},
"required": []string{"url"},
},
}, p.handleSubscribe)
s.RegisterTool(tp+"unsubscribe", sdk.ToolDef{
Name: tp + "unsubscribe", Description: "Unsubscribe from a feed",
NoMemory: true,
Parameters: map[string]interface{}{
"type": "object",
"properties": map[string]interface{}{
"url": map[string]interface{}{"type": "string", "description": "Feed URL to unsubscribe"},
},
"required": []string{"url"},
},
}, p.handleUnsubscribe)
s.RegisterTool(tp+"list", sdk.ToolDef{
Name: tp + "list", Description: "List all subscribed feeds",
Parameters: map[string]interface{}{
"type": "object",
"properties": map[string]interface{}{},
},
}, p.handleList)
s.RegisterTool(tp+"check_now", sdk.ToolDef{
Name: tp + "check_now", Description: "Manually check all feeds for new articles now",
NoMemory: true,
Parameters: map[string]interface{}{
"type": "object",
"properties": map[string]interface{}{},
},
}, p.handleCheckNow)
pollMin := int(getSetting(s.Settings(), "poll_interval", int64(30)))
if pollMin < 5 {
pollMin = 5
}
p.pollTicker = time.NewTicker(time.Duration(pollMin) * time.Minute)
p.wg.Add(1)
go p.pollLoop()
fmt.Printf("[%s] started (%d feeds, poll every %dm)\n", p.name, len(p.feeds), pollMin)
return nil
}
func (p *Plugin) Stop() error {
p.stopOnce.Do(func() { close(p.stopCh) })
p.pollTicker.Stop()
p.wg.Wait()
p.saveData()
fmt.Printf("[%s] stopped\n", p.name)
return nil
}
func (p *Plugin) pollLoop() {
defer p.wg.Done()
p.checkAllFeeds()
for {
select {
case <-p.pollTicker.C:
p.checkAllFeeds()
case <-p.stopCh:
return
}
}
}
func (p *Plugin) checkAllFeeds() {
p.mu.RLock()
feeds := make([]FeedSub, len(p.feeds))
copy(feeds, p.feeds)
p.mu.RUnlock()
for _, feed := range feeds {
select {
case <-p.stopCh:
return
default:
}
p.checkFeed(feed)
}
}
func (p *Plugin) checkFeed(sub FeedSub) {
parsed, err := p.fp.ParseURL(sub.URL)
if err != nil {
return
}
title := parsed.Title
if title == "" {
title = sub.URL
}
var newArticles []*gofeed.Item
for _, item := range parsed.Items {
guid := item.GUID
if guid == "" {
guid = item.Link
}
if guid == "" {
continue
}
guid = sub.URL + "|" + guid
p.mu.RLock()
seen := p.seenGUIDs[guid]
p.mu.RUnlock()
if !seen {
newArticles = append(newArticles, item)
}
}
if len(newArticles) == 0 {
return
}
now := time.Now()
toInject := make([]*gofeed.Item, 0, len(newArticles))
p.mu.Lock()
for _, item := range newArticles {
guid := item.GUID
if guid == "" {
guid = item.Link
}
if guid == "" {
continue
}
key := sub.URL + "|" + guid
if t, ok := p.injected[key]; ok && now.Sub(t) < injectDedupWindow {
continue
}
p.injected[key] = now
p.seenGUIDs[key] = true
toInject = append(toInject, item)
}
p.mu.Unlock()
if len(toInject) == 0 {
return
}
var lines []string
lines = append(lines, fmt.Sprintf("📡 %s (%s) — %d 篇新文章:", title, sub.URL, len(toInject)))
for _, item := range toInject {
pubDate := ""
if item.PublishedParsed != nil {
pubDate = item.PublishedParsed.Format("01-02 15:04")
}
line := fmt.Sprintf(" • %s", item.Title)
if pubDate != "" {
line += fmt.Sprintf(" [%s]", pubDate)
}
if item.Link != "" {
line += "\n " + item.Link
}
lines = append(lines, line)
}
// 中断注入是「系统通知」NoMemory 写明意图:这类提醒不参与记忆计算,
// 原文仍进上下文(模型当轮看得到)。
p.sdk.InjectInterruptTextOpts("rss", "rss", strings.Join(lines, "\n"),
sdk.InjectOptions{NoMemory: true})
p.saveData()
}
func (p *Plugin) handleSubscribe(args map[string]interface{}) (interface{}, error) {
url := readArg(args, "url")
if url == "" {
return map[string]interface{}{"isError": true, "content": "URL is required"}, nil
}
p.mu.RLock()
for _, f := range p.feeds {
if f.URL == url {
p.mu.RUnlock()
return map[string]interface{}{"isError": true, "content": "Already subscribed to: " + url}, nil
}
}
p.mu.RUnlock()
interval := readArgInt(args, "interval", 30)
if interval < 5 {
interval = 5
}
parsed, err := p.fp.ParseURL(url)
if err != nil {
return map[string]interface{}{"isError": true, "content": "Failed to parse feed: " + err.Error()}, nil
}
feedTitle := parsed.Title
if feedTitle == "" {
feedTitle = url
}
sub := FeedSub{
URL: url,
Title: feedTitle,
AddedAt: time.Now().Format("2006-01-02 15:04"),
Interval: interval,
}
guidCount := 0
p.mu.Lock()
for _, item := range parsed.Items {
guid := item.GUID
if guid == "" {
guid = item.Link
}
if guid == "" {
continue
}
p.seenGUIDs[url+"|"+guid] = true
guidCount++
}
p.mu.Unlock()
p.mu.Lock()
p.feeds = append(p.feeds, sub)
p.mu.Unlock()
p.saveData()
return map[string]interface{}{
"content": fmt.Sprintf("Subscribed to: %s\nTitle: %s\nArticles found: %d\nPoll interval: %d min", url, feedTitle, guidCount, interval),
}, nil
}
func (p *Plugin) handleUnsubscribe(args map[string]interface{}) (interface{}, error) {
url := readArg(args, "url")
if url == "" {
return map[string]interface{}{"isError": true, "content": "URL is required"}, nil
}
p.mu.Lock()
found := false
for i, f := range p.feeds {
if f.URL == url {
p.feeds = append(p.feeds[:i], p.feeds[i+1:]...)
found = true
break
}
}
if !found {
p.mu.Unlock()
return map[string]interface{}{"isError": true, "content": "Not subscribed to: " + url}, nil
}
for guid := range p.seenGUIDs {
if strings.HasPrefix(guid, url+"|") {
delete(p.seenGUIDs, guid)
}
}
p.mu.Unlock()
p.saveData()
return map[string]interface{}{"content": "Unsubscribed: " + url}, nil
}
func (p *Plugin) handleList(args map[string]interface{}) (interface{}, error) {
p.mu.RLock()
defer p.mu.RUnlock()
if len(p.feeds) == 0 {
return map[string]interface{}{"content": "No subscriptions. Use rss_subscribe to add one."}, nil
}
sort.Slice(p.feeds, func(i, j int) bool {
return p.feeds[i].Title < p.feeds[j].Title
})
var lines []string
lines = append(lines, fmt.Sprintf("📡 Subscriptions (%d):", len(p.feeds)))
for _, f := range p.feeds {
lines = append(lines, fmt.Sprintf(" • %s\n %s (every %dm, added %s)", f.Title, f.URL, f.Interval, f.AddedAt))
}
return map[string]interface{}{"content": strings.Join(lines, "\n")}, nil
}
func (p *Plugin) handleCheckNow(args map[string]interface{}) (interface{}, error) {
select {
case <-p.stopCh:
return map[string]interface{}{"isError": true, "content": "plugin is stopping"}, nil
default:
}
p.wg.Add(1)
go func() {
defer p.wg.Done()
p.checkAllFeeds()
}()
return map[string]interface{}{"content": "Checking all feeds for updates..."}, nil
}
func (p *Plugin) dataFile() string {
return filepath.Join(p.dataDir, "feeds.json")
}
func (p *Plugin) loadData() {
b, err := os.ReadFile(p.dataFile())
if err != nil {
return
}
var data struct {
Feeds []FeedSub `json:"feeds"`
SeenGUIDs map[string]bool `json:"seen"`
}
if json.Unmarshal(b, &data) != nil {
return
}
if data.Feeds != nil {
p.feeds = data.Feeds
}
if data.SeenGUIDs != nil {
p.seenGUIDs = data.SeenGUIDs
}
}
func (p *Plugin) saveData() {
p.mu.RLock()
defer p.mu.RUnlock()
data := struct {
Feeds []FeedSub `json:"feeds"`
SeenGUIDs map[string]bool `json:"seen"`
}{
Feeds: p.feeds,
SeenGUIDs: p.seenGUIDs,
}
b, _ := json.MarshalIndent(data, "", " ")
atomicWriteJSON(p.dataFile(), b)
}
// cleanupData 卸载时清理订阅数据目录feeds.json 等)
func (p *Plugin) cleanupData() {
p.mu.Lock()
defer p.mu.Unlock()
if p.dataDir == "" {
return
}
for _, f := range []string{"feeds.json"} {
path := filepath.Join(p.dataDir, f)
if err := os.Remove(path); err != nil && !os.IsNotExist(err) {
fmt.Printf("[%s] onRemove cleanup %s: %v\n", p.name, path, err)
}
}
}
// atomicWriteJSON 原子写 JSON先写临时文件再 rename避免进程崩溃截断数据文件。
func atomicWriteJSON(path string, data []byte) error {
tmp := path + ".tmp"
if err := os.WriteFile(tmp, data, 0644); err != nil {
return err
}
return os.Rename(tmp, path)
}