Files
HomeAgent/internal/plugins/agentcli/plugin_test.go
JianFeeeee 687e5655bc feat(sdk): 多模态贯通插件边界——公开接口、内核桥接与统一输入主干
记忆系统在 1.1.0 支持了二进制多媒体节点,但那条链路只对**内核自己**开放:
用户在 qq 发图能落进 CAS、能被记忆引用,而插件调 Commit / DocMemory().Insert
交进来的媒体一律无处安放。原因是三层都断着,且**每一层都不报错**。

## 一、公开 SDK:补上媒体的表达能力(全部新增,无签名变更)

- `Triple` += `SentenceText`、`MediaDigests`
- `Doc` += `MediaDigests`、`Attachments`;新增 `MediaAttachment`
- `TextEvent` += `Attachments`
- `DocMemoryAPI` += `InsertWithMedia`
- `IOInjector` += `InjectInputMedia` / `InjectInputMediaSync` / `InjectInterruptMedia`
- `PluginSDK` 补上一直缺失的 `SetToolBlocks` 包装(接口里有、便捷方法里没有)

`MediaAttachment` 一个类型服务两个方向:给 `Data`+`MIME` 是新内容(CAS 按字节
去重),只给 `Digest` 是引用已有内容。读路径**只回元数据不回字节**——一次检索
可能命中几十份媒体,全塞回去会把跨进程消息撑爆。

媒体注入不能搭 `SetToolBlocks` 的车:那个方法只在工具处理函数内部可用,且媒体
要等下一条 tool message 才到模型手上。插件主动发起一轮带媒体的对话、以及中断
注入,需要自己的签名,且媒体在**本轮**就送到模型。

## 二、内核桥接层:原先在静默裁字段

`internal/sdk/memory_impl.go` 此前只搬自己认识的几个字段,其余丢弃且返回 nil:

- 图记忆丢 `Confidence`/`SubjectType`/`ObjectType`/`SentenceText`,又走 `Commit`
  而非 `CommitWithMedia`(不回 sentenceIDs)→ 媒体绑定链 `SentenceText → sentences
  → sentence_id → media_refs` 一步都走不通,插件即便按格式写好标记也永远挂不上;
- 知识库 `Query` 只回 ID/Title/Content,`Insert` 只写这三个;`Remove` 不解引用,
  于是那些媒体永久处于「被引用」状态,GC 收不掉、磁盘只增不减
  (内核的归档路径 `releaseDocMedia` 做了这一步,插件路径漏了同一步)。

规则改为:**内部结构有的字段一律透传**。标记格式处理作为包级私有辅助留在桥接
层自己手里,但必须与内核 `mediaSummaryForEvent` 字节兼容——两边要能互读对方
写下的标记。

标记插入必须在 `ds.Insert` **之前**(向量索引取 `Summary + " " + Content`,
之后补的标记检索不到),引用绑定必须在**之后**(owner_id 是 Insert 生成的 ID)。

## 三、跨进程链路:不接线就是全体外部插件编译失败

`go test` 直接把这一层拍出来了——`procIO does not implement sdk.IOInjector`。
公开接口加方法后,生成模板不跟上,**每个外部插件都编不过**,是硬失败不是软降级。
六处接线:`protocol.go` 四个 method 常量、`capability.go` 能力归属、
`corehandler.go` 四个分派分支、`proc_core.go` 委托、`proc_main.go.tmpl` 模板侧
实现、以及三个测试替身。

## 四、统一输入主干:把模态从「函数选择」降级为「字段」

`processTextInput` / `processMediaInput` 合并为 `processInput`。这个分叉是历史
产物而非设计:`processTextInput` 本来就处理媒体(`bindEventMedia` +
`mediaSummaryForEvent`,与媒体路径尾部完全相同),`process()` 只看
`stageCtx.Extra["media_blocks"]`、根本不认识 `evt.Type`。模态是输入的**属性**,
不是输入的**种类**。

媒体路径由此获得它一直缺的六项:去重、`no_memory`、通道 `Cleaner`、中断语义、
`_consolidation_` 路由、正确的 `EventRawInput`。

最后一项是个真 bug:媒体路径发布 `"content": evt.Payload`(一个 map),而
`webui/handler.go` 断言 `.(string)` → 断言失败、`content == ""`、提前返回。
**用户发的图从来没出现在 WebUI 聊天记录里。**

`media_blocks` 同时接受 `[]agentAPI.ContentBlock` 与 `[]pubsdk.ContentBlock`:
字段一致但 Go 不自动转换,只认一种的后果是另一种被静默丢弃。

## 五、模型可调用的三个工具

`memory_commit` 的 `sentence_text` **从未暴露给模型**,而它是绑定链上的必经环节;
连同 `media_digests` 一起补进 JSON schema 与工具文档。`doc_commit` 加
`media_digests`。`doc_query` 把关联媒体单独一行附在结果末尾(正文按 2000 字截断,
标记通常就在尾部)。

标记由**内核**生成而非插件/模型拼装:要求调用方知道格式,等于让一个拼写错误
静默切断引用绑定,而全链路无人报错。

## 六、WebUI 上传走真实媒体链路

图片/音频读回字节拼 data URL 注入 `media_blocks`(8MB 上限,超限退回按路径处理)。
此前只注入一句「文件已保存到 <路径>」,指望模型自己调 `files_read`——但那返回
文本,图片字节对模型永远不可见。附件类型识别扩展到 audio 并在缺 Content-Type
时按扩展名兜底(判错不只是卡片样式问题,图片被当普通文件就进不了视觉链路)。

## 测试

- `internal/sdk/memory_impl_test.go`(12 例,此前该包**没有任何测试文件**)
- `internal/agent/core/inputunify_test.go`(统一主干 + 双静态类型 + 三工具媒体)
- `third_party/homeagent-sdk/sdk/stress_test.go`(13 例并发压测)

压测抓到两处**真**竞态(不是理论风险):`PluginSDK` 的 API 字段与 `autoRestart`
无锁,而写方(内核注入 API、插件 `SetAutoRestart`)与读方(插件后台 goroutine
注入、内核 registry 读 `AutoRestart`)天然跨 goroutine。加 `apiMu` 修掉;约定
只在持锁期间取字段值,取完即释放再调用——持锁调用会把 `InjectInputSync` 这类
阻塞到 agent 回复(可达数分钟)的方法与 `SetIOInjector` 串起来,让插件重载卡死。

测试还抓出两个自身缺陷:`bindDocMedia` 把同一份媒体数两次(`AddRef` 幂等所以表
是对的,但日志说「绑定 2 个」而实际 1 条——误导后续排查),以及用单字符实体名
时 `validEntityName` 静默跳过、`Commit` 返回 nil 却什么都没写。

存量插件不需要改一行也不需要重编:新增方法由插件调用、内核实现,不调就不受影响。
17 个 example 插件源码零改动通过类型检查。
2026-09-06 09:51:31 +08:00

668 lines
15 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.

//go:build linux || windows
package agentcli
import (
"encoding/json"
"fmt"
"strings"
"sync"
"testing"
"time"
sdk "gitcode.com/JianFeeeee/HomeAgent/internal/sdk"
sdkpub "gitcode.com/JianFeeeee/homeagent-sdk/sdk"
)
type toolCapture struct {
handlers map[string]sdk.ToolHandler
defs map[string]sdk.ToolDef
}
func newToolCapture() *toolCapture {
return &toolCapture{
handlers: make(map[string]sdk.ToolHandler),
defs: make(map[string]sdk.ToolDef),
}
}
func (tc *toolCapture) RegisterTool(name string, def sdk.ToolDef, handler sdk.ToolHandler) error {
tc.handlers[name] = handler
tc.defs[name] = def
return nil
}
func (tc *toolCapture) RegisterStage(stage sdk.Stage, handler sdk.StageHandler) {}
func (tc *toolCapture) RegisterAPI(name string) error { return nil }
func setupPlugin() (*Plugin, *toolCapture, error) {
p := New("agentcli")
tc := newToolCapture()
sdk := sdk.New("agentcli", sdk.SDKConfig{
RegTool: tc.RegisterTool,
RegStage: tc.RegisterStage,
RegAPI: tc.RegisterAPI,
Settings: sdk.NewSettings("agentcli", nil),
})
if err := p.Start(sdk); err != nil {
return nil, nil, err
}
return p, tc, nil
}
func TestKeyMapping(t *testing.T) {
tests := []struct {
key string
expected []byte
}{
{"enter", []byte{0x0D}},
{"tab", []byte{0x09}},
{"escape", []byte{0x1B}},
{"esc", []byte{0x1B}},
{"backspace", []byte{0x7F}},
{"delete", []byte{0x1B, 0x5B, 0x33, 0x7E}},
{"home", []byte{0x1B, 0x5B, 0x48}},
{"end", []byte{0x1B, 0x5B, 0x46}},
{"up", []byte{0x1B, 0x5B, 0x41}},
{"down", []byte{0x1B, 0x5B, 0x42}},
{"left", []byte{0x1B, 0x5B, 0x44}},
{"right", []byte{0x1B, 0x5B, 0x43}},
{"page_up", []byte{0x1B, 0x5B, 0x35, 0x7E}},
{"page_down", []byte{0x1B, 0x5B, 0x36, 0x7E}},
{"ctrl_a", []byte{0x01}},
{"ctrl_z", []byte{0x1A}},
{"alt_a", []byte{0x1B, 'a'}},
{"alt_z", []byte{0x1B, 'z'}},
{"f1", []byte{0x1B, 0x5B, 0x50}},
{"f4", []byte{0x1B, 0x5B, 0x53}},
{"f12", []byte{0x1B, 0x5B, 0x32, 0x34, 0x7E}},
}
for _, tt := range tests {
t.Run(tt.key, func(t *testing.T) {
result, err := mapKey(tt.key)
if err != nil {
t.Fatalf("unexpected error: %v", err)
}
if !bytesEqual(result, tt.expected) {
t.Fatalf("expected %v, got %v", tt.expected, result)
}
})
}
}
func TestKeyMappingInvalid(t *testing.T) {
_, err := mapKey("unknown_key")
if err == nil {
t.Fatal("expected error for unknown key")
}
}
func bytesEqual(a, b []byte) bool {
if len(a) != len(b) {
return false
}
for i := range a {
if a[i] != b[i] {
return false
}
}
return true
}
func TestListEmpty(t *testing.T) {
p, tc, err := setupPlugin()
if err != nil {
t.Fatal(err)
}
defer p.Stop()
handler := tc.handlers["terminal_list"]
result, err := handler(map[string]interface{}{})
if err != nil {
t.Fatal(err)
}
data, _ := json.Marshal(result)
var resp map[string]interface{}
json.Unmarshal(data, &resp)
if resp["status"] != "ok" {
t.Fatalf("expected status ok, got %v", resp["status"])
}
if resp["count"].(float64) != 0 {
t.Fatalf("expected count 0, got %v", resp["count"])
}
}
func TestCreateAndCloseTerminal(t *testing.T) {
p, tc, err := setupPlugin()
if err != nil {
t.Fatal(err)
}
defer p.Stop()
createHandler := tc.handlers["terminal_create"]
result, err := createHandler(map[string]interface{}{
"command": "echo hello",
"timeout": "10s",
})
if err != nil {
t.Fatal(err)
}
data, _ := json.Marshal(result)
var createResp map[string]interface{}
json.Unmarshal(data, &createResp)
if v, ok := createResp["error"]; ok {
t.Skipf("PTY not available in this environment: %v", v)
}
if createResp["status"] != "created" {
t.Fatalf("expected status created, got %v", createResp["status"])
}
id := createResp["id"].(string)
if id == "" {
t.Fatal("expected non-empty terminal id")
}
// Give the terminal a moment to output
time.Sleep(200 * time.Millisecond)
// Read output
readHandler := tc.handlers["terminal_read"]
result, err = readHandler(map[string]interface{}{
"id": id,
"clear": true,
})
if err != nil {
t.Fatal(err)
}
data, _ = json.Marshal(result)
var readResp map[string]interface{}
json.Unmarshal(data, &readResp)
if readResp["status"] != "ok" {
t.Fatalf("expected status ok, got %v", readResp["status"])
}
// Close
closeHandler := tc.handlers["terminal_close"]
result, err = closeHandler(map[string]interface{}{
"id": id,
})
if err != nil {
t.Fatal(err)
}
data, _ = json.Marshal(result)
var closeResp map[string]interface{}
json.Unmarshal(data, &closeResp)
if closeResp["status"] != "closed" {
t.Fatalf("expected status closed, got %v", closeResp["status"])
}
}
func TestCreateTerminalMissingArgs(t *testing.T) {
p, tc, err := setupPlugin()
if err != nil {
t.Fatal(err)
}
defer p.Stop()
handler := tc.handlers["terminal_create"]
result, err := handler(map[string]interface{}{})
if err != nil {
t.Fatal(err)
}
data, _ := json.Marshal(result)
var resp map[string]interface{}
json.Unmarshal(data, &resp)
if v, ok := resp["error"]; ok {
t.Skipf("PTY not available in this environment: %v", v)
}
if resp["status"] != "created" {
t.Fatalf("expected status created, got %v", resp["status"])
}
id := resp["id"].(string)
p.handleClose(map[string]interface{}{"id": id})
}
func TestWriteToNonexistentTerminal(t *testing.T) {
_, tc, err := setupPlugin()
if err != nil {
t.Fatal(err)
}
handler := tc.handlers["terminal_write"]
result, err := handler(map[string]interface{}{
"id": "nonexistent",
"input": "test",
})
if err != nil {
t.Fatal(err)
}
data, _ := json.Marshal(result)
var resp map[string]interface{}
json.Unmarshal(data, &resp)
if _, ok := resp["error"]; !ok {
t.Fatal("expected error for nonexistent terminal")
}
}
func TestReadNonexistentTerminal(t *testing.T) {
_, tc, err := setupPlugin()
if err != nil {
t.Fatal(err)
}
handler := tc.handlers["terminal_read"]
result, err := handler(map[string]interface{}{
"id": "nonexistent",
})
if err != nil {
t.Fatal(err)
}
data, _ := json.Marshal(result)
var resp map[string]interface{}
json.Unmarshal(data, &resp)
if _, ok := resp["error"]; !ok {
t.Fatal("expected error for nonexistent terminal")
}
}
func TestResizeNonexistentTerminal(t *testing.T) {
_, tc, err := setupPlugin()
if err != nil {
t.Fatal(err)
}
handler := tc.handlers["terminal_resize"]
result, err := handler(map[string]interface{}{
"id": "nonexistent",
"rows": float64(40),
"cols": float64(120),
})
if err != nil {
t.Fatal(err)
}
data, _ := json.Marshal(result)
var resp map[string]interface{}
json.Unmarshal(data, &resp)
if _, ok := resp["error"]; !ok {
t.Fatal("expected error for nonexistent terminal")
}
}
func TestCloseNonexistentTerminal(t *testing.T) {
_, tc, err := setupPlugin()
if err != nil {
t.Fatal(err)
}
handler := tc.handlers["terminal_close"]
result, err := handler(map[string]interface{}{
"id": "nonexistent",
})
if err != nil {
t.Fatal(err)
}
data, _ := json.Marshal(result)
var resp map[string]interface{}
json.Unmarshal(data, &resp)
if _, ok := resp["error"]; !ok {
t.Fatal("expected error for nonexistent terminal")
}
}
func TestTerminalWriteRequiresId(t *testing.T) {
_, tc, err := setupPlugin()
if err != nil {
t.Fatal(err)
}
handler := tc.handlers["terminal_write"]
result, err := handler(map[string]interface{}{
"input": "test",
})
if err != nil {
t.Fatal(err)
}
data, _ := json.Marshal(result)
var resp map[string]interface{}
json.Unmarshal(data, &resp)
if _, ok := resp["error"]; !ok {
t.Fatal("expected error when id is missing")
}
}
func TestTerminalWriteRequiresContent(t *testing.T) {
_, tc, err := setupPlugin()
if err != nil {
t.Fatal(err)
}
handler := tc.handlers["terminal_write"]
result, err := handler(map[string]interface{}{
"id": "test",
})
if err != nil {
t.Fatal(err)
}
data, _ := json.Marshal(result)
var resp map[string]interface{}
json.Unmarshal(data, &resp)
if _, ok := resp["error"]; !ok {
t.Fatal("expected error when both input and key are missing")
}
}
func TestToolsRegistered(t *testing.T) {
p, tc, err := setupPlugin()
if err != nil {
t.Fatal(err)
}
defer p.Stop()
expectedTools := []string{
"terminal_create",
"terminal_write",
"terminal_read",
"terminal_resize",
"terminal_close",
"terminal_list",
}
for _, name := range expectedTools {
if _, ok := tc.handlers[name]; !ok {
t.Errorf("tool %s not registered", name)
}
}
}
// ——— Phase 6: 通知节流测试mock 终端 + 捕获注入) ———
type injectCapture struct {
mu sync.Mutex
texts []string
}
func (c *injectCapture) InjectInterruptText(source, channel, text string) {
c.mu.Lock()
c.texts = append(c.texts, text)
c.mu.Unlock()
}
func (c *injectCapture) InjectText(source, channel, text string) {
c.mu.Lock()
c.texts = append(c.texts, text)
c.mu.Unlock()
}
func (c *injectCapture) InjectTextNoMemory(source, channel, text string) {
c.mu.Lock()
c.texts = append(c.texts, text)
c.mu.Unlock()
}
func (c *injectCapture) SetToolBlocks(blocks []sdkpub.ContentBlock) {
// 测试桩:忽略多模态块
}
func (c *injectCapture) InjectInputSync(source, channel, text string) string { return "" }
// 三个带媒体的注入方法同样记录文本:本测试只关心「注入了什么话」,
// 媒体块的转发在 core 的 injectedBlocks 测试里覆盖。
func (c *injectCapture) InjectInputMedia(source, channel, text string, blocks []sdkpub.ContentBlock) {
c.mu.Lock()
c.texts = append(c.texts, text)
c.mu.Unlock()
}
func (c *injectCapture) InjectInputMediaSync(source, channel, text string, blocks []sdkpub.ContentBlock) string {
return ""
}
func (c *injectCapture) InjectInterruptMedia(source, channel, text string, blocks []sdkpub.ContentBlock) {
c.mu.Lock()
c.texts = append(c.texts, text)
c.mu.Unlock()
}
func (c *injectCapture) snapshot() []string {
c.mu.Lock()
defer c.mu.Unlock()
out := make([]string, len(c.texts))
copy(out, c.texts)
return out
}
// mockTerm 可控输出流的假终端Read 从 data chan 取数据,可模拟进程退出/读取错误
type mockTerm struct {
mu sync.Mutex
data chan []byte
running bool
err error
}
func newMockTerm() *mockTerm {
return &mockTerm{data: make(chan []byte, 16), running: true}
}
func (m *mockTerm) Read(buf []byte) (int, error) {
for {
m.mu.Lock()
err := m.err
running := m.running
m.mu.Unlock()
if err != nil {
return 0, err
}
if !running {
return 0, fmt.Errorf("process exited")
}
select {
case data, ok := <-m.data:
if !ok {
return 0, fmt.Errorf("closed")
}
n := copy(buf, data)
return n, nil
case <-time.After(20 * time.Millisecond):
}
}
}
func (m *mockTerm) WriteString(s string) (int, error) { return len(s), nil }
func (m *mockTerm) Resize(rows, cols uint16) error { return nil }
func (m *mockTerm) Running() bool {
m.mu.Lock()
defer m.mu.Unlock()
return m.running
}
func (m *mockTerm) Kill() error { return nil }
func (m *mockTerm) Close() error { return nil }
func (m *mockTerm) push(data []byte) {
m.data <- data
}
func (m *mockTerm) setRunning(v bool) {
m.mu.Lock()
m.running = v
m.mu.Unlock()
}
func (m *mockTerm) setErr(err error) {
m.mu.Lock()
m.err = err
m.mu.Unlock()
}
func newTestSession(term ptyTerm) *TerminalSession {
return &TerminalSession{
id: "t1",
session: term,
createdAt: time.Now(),
timeout: 10 * time.Minute,
stopCh: make(chan struct{}),
done: make(chan struct{}),
}
}
func startReadLoop(p *Plugin, s *sdk.PluginSDK, t *TerminalSession) {
p.wg.Add(1)
go p.readLoop(t, s)
}
func waitInjected(c *injectCapture, substr string, timeout time.Duration) bool {
deadline := time.Now().Add(timeout)
for time.Now().Before(deadline) {
for _, text := range c.snapshot() {
if strings.Contains(text, substr) {
return true
}
}
time.Sleep(20 * time.Millisecond)
}
return false
}
// Phase 6: 持续吐进度时,通知频率显著低于 500ms/条(节流生效)
func TestReadLoopNotifyThrottle(t *testing.T) {
p := New("agentcli")
p.notifyBytes = 2048
p.notifyInterval = 2 * time.Second
capture := &injectCapture{}
sdkInst := sdk.New("agentcli", sdk.SDKConfig{
RegTool: newToolCapture().RegisterTool,
RegStage: func(sdk.Stage, sdk.StageHandler) {},
RegAPI: func(string) error { return nil },
Settings: sdk.NewSettings("agentcli", nil),
})
sdkInst.SetIOInjector(capture)
term := newMockTerm()
ts := newTestSession(term)
startReadLoop(p, sdkInst, ts)
if !waitInjected(capture, "已启动", 2*time.Second) {
t.Fatal("expected startup notification")
}
// 持续以 100B/50ms(=2KB/s) 吐进度 3 秒
stop := make(chan struct{})
go func() {
ticker := time.NewTicker(50 * time.Millisecond)
defer ticker.Stop()
chunk := make([]byte, 100)
for i := range chunk {
chunk[i] = 'x'
}
for {
select {
case <-stop:
return
case <-ticker.C:
term.push(chunk)
}
}
}()
time.Sleep(3 * time.Second)
close(stop)
notifies := 0
for _, text := range capture.snapshot() {
if strings.Contains(text, "有新输出") {
notifies++
}
}
// 3 秒持续输出500ms/条 的旧行为应有 6 条;节流后 ≤3 条
if notifies > 3 {
t.Errorf("notify throttle ineffective: %d notifies in 3s (expected <=3)", notifies)
}
if notifies == 0 {
t.Error("expected at least one output notification")
}
close(ts.stopCh)
<-ts.done
}
// Phase 6: 进程退出 → 立即通知两条路径PTY Read 返回 EOF 走"读取结束"
// 或 reader 阻塞时顶部 terminalRunning 检测走"进程已退出"
func TestReadLoopNotifyOnExit(t *testing.T) {
p := New("agentcli")
p.notifyBytes = 2048
p.notifyInterval = 2 * time.Second
capture := &injectCapture{}
sdkInst := sdk.New("agentcli", sdk.SDKConfig{
RegTool: newToolCapture().RegisterTool,
RegStage: func(sdk.Stage, sdk.StageHandler) {},
RegAPI: func(string) error { return nil },
Settings: sdk.NewSettings("agentcli", nil),
})
sdkInst.SetIOInjector(capture)
term := newMockTerm()
ts := newTestSession(term)
startReadLoop(p, sdkInst, ts)
if !waitInjected(capture, "已启动", 2*time.Second) {
t.Fatal("expected startup notification")
}
term.setRunning(false)
gotExit := waitInjected(capture, "进程已退出", 2*time.Second)
gotReadEnd := waitInjected(capture, "读取结束", time.Second)
if !gotExit && !gotReadEnd {
t.Error("expected immediate notification on process exit (either 进程已退出 or 读取结束)")
}
close(ts.stopCh)
}
// Phase 6: 读取错误/EOF → 立即通知
func TestReadLoopNotifyOnReadError(t *testing.T) {
p := New("agentcli")
p.notifyBytes = 2048
p.notifyInterval = 2 * time.Second
capture := &injectCapture{}
sdkInst := sdk.New("agentcli", sdk.SDKConfig{
RegTool: newToolCapture().RegisterTool,
RegStage: func(sdk.Stage, sdk.StageHandler) {},
RegAPI: func(string) error { return nil },
Settings: sdk.NewSettings("agentcli", nil),
})
sdkInst.SetIOInjector(capture)
term := newMockTerm()
ts := newTestSession(term)
startReadLoop(p, sdkInst, ts)
if !waitInjected(capture, "已启动", 2*time.Second) {
t.Fatal("expected startup notification")
}
term.setErr(fmt.Errorf("read timeout"))
if !waitInjected(capture, "读取结束", 3*time.Second) {
t.Error("expected immediate notification on read error")
}
close(ts.stopCh)
<-ts.done
}