feat: remotedevice 二进制分块协议(录像回传+音频下发) + 工具描述增强

回应 GUI c29abe9 需服务端配套项([阻塞]两项 + [中]一项):

1. readFrame 支持 0x2 二进制帧(之前遇 0x2 直接断连):
   - 返回 opcode, 二进制帧上限放宽至 8MB
2. handleWS 二进制聚合协议(设备→网关, 录像回传):
   - cmd_data_start 开启按 req_id 聚合 → 0x2 帧追加 → cmd_data_end 聚合完成
   - 结果存 cmdresult(data_base64 字段), device_ctl_cmdresult 可取回
   - 超限防护(声明 total×2 或硬上限 64MB)
3. PushData 下发协议(网关→设备, 音频/TTS):
   - cmd_speech_start → 0x2 分块(8KB) → cmd_speech_end
   - GUI 端已实现接收侧(speakeruse 播放链路打通)
4. device_ctl_cmdrun 工具描述补齐:
   - screensue 带参示例(<内容>/<秒> <内容>)
   - camerasue 录像说明(cmdresult 含 data_base64)
   - speakeruse 文字朗读

测试: 最小 WS 客户端端到端验证二进制分块上传/PushData 下发/离线报错
This commit is contained in:
JianFeeeee
2026-08-21 10:53:38 +08:00
parent 8d368913a9
commit f2e3215c77
3 changed files with 457 additions and 20 deletions

View File

@ -0,0 +1,297 @@
package remotedevice
import (
"bufio"
"encoding/base64"
"encoding/binary"
"encoding/json"
"math/rand"
"net"
"net/http"
"net/http/httptest"
"strings"
"testing"
"time"
)
// ===== 测试用最小 WS 客户端(模拟 GUI 设备桥)=====
type testWSClient struct {
conn net.Conn
rw *bufio.ReadWriter
}
func dialTestWS(t *testing.T, url, token string) *testWSClient {
t.Helper()
req := "GET /api/v1/device/ws?token=" + token + " HTTP/1.1\r\n" +
"Host: test\r\nUpgrade: websocket\r\nConnection: Upgrade\r\n" +
"Sec-WebSocket-Key: dGhlIHNhbXBsZSBub25jZQ==\r\nSec-WebSocket-Version: 13\r\n\r\n"
conn, err := net.Dial("tcp", strings.TrimPrefix(url, "http://"))
if err != nil {
t.Fatalf("dial: %v", err)
}
if _, err := conn.Write([]byte(req)); err != nil {
t.Fatalf("write upgrade: %v", err)
}
br := bufio.NewReader(conn)
resp, err := readLine(br)
if err != nil {
t.Fatalf("read upgrade resp: %v", err)
}
if !strings.Contains(resp, "101") {
t.Fatalf("expected 101, got %s", resp)
}
for {
line, err := readLine(br)
if err != nil {
t.Fatalf("read headers: %v", err)
}
if line == "\r\n" || line == "" {
break
}
}
return &testWSClient{conn: conn, rw: &bufio.ReadWriter{Reader: br, Writer: bufio.NewWriter(conn)}}
}
func readLine(br *bufio.Reader) (string, error) {
var sb strings.Builder
for {
b, err := br.ReadByte()
if err != nil {
return sb.String(), err
}
sb.WriteByte(b)
if b == '\n' {
return sb.String(), nil
}
}
}
// sendText 发送客户端文本帧带掩码RFC6455 要求客户端帧必须掩码)
func (c *testWSClient) sendText(payload []byte) {
c.sendFrame(0x1, payload)
}
func (c *testWSClient) sendBinary(payload []byte) {
c.sendFrame(0x2, payload)
}
func (c *testWSClient) sendFrame(opcode byte, payload []byte) {
maskKey := make([]byte, 4)
rand.Read(maskKey)
masked := make([]byte, len(payload))
for i := range payload {
masked[i] = payload[i] ^ maskKey[i%4]
}
var hdr []byte
hdr = append(hdr, 0x80|opcode)
n := len(payload)
switch {
case n < 126:
hdr = append(hdr, 0x80|byte(n))
case n <= 0xffff:
hdr = append(hdr, 0x80|126)
ext := make([]byte, 2)
binary.BigEndian.PutUint16(ext, uint16(n))
hdr = append(hdr, ext...)
default:
hdr = append(hdr, 0x80|127)
ext := make([]byte, 8)
binary.BigEndian.PutUint64(ext, uint64(n))
hdr = append(hdr, ext...)
}
c.rw.Write(hdr)
c.rw.Write(maskKey)
c.rw.Write(masked)
c.rw.Flush()
}
// readMsg 读一帧(跳过 pong返回 opcode 与 payload
func (c *testWSClient) readMsg() (byte, []byte, error) {
for {
payload, isClose, opcode, err := readFrame(c.rw.Reader)
if err != nil || isClose {
return 0, nil, err
}
if opcode == 0xa {
continue
}
return opcode, payload, nil
}
}
func (c *testWSClient) close() { c.conn.Close() }
// ===== 端到端hello/bind/cmd + 二进制分块回传(录像协议)=====
func TestWSBinaryChunkUpload(t *testing.T) {
reg := NewRegistry()
token := "test-token-123"
reg.SetAcceptToken(func(provided string) bool { return provided == token })
srv := httptest.NewServer(http.HandlerFunc(reg.ServeWS))
defer srv.Close()
url := srv.URL
cli := dialTestWS(t, url, token)
defer cli.close()
// hello 登记
cli.sendText([]byte(`{"op":"hello","device":{"device_id":"gui-test","name":"测试机","kind":"computer","caps":["cmd"]}}`))
op, payload, err := cli.readMsg()
if err != nil {
t.Fatalf("read hello_ack: %v", err)
}
if op != 0x1 {
t.Fatalf("expected text frame, got %x", op)
}
var ack map[string]interface{}
json.Unmarshal(payload, &ack)
if ack["op"] != "hello_ack" {
t.Fatalf("expected hello_ack, got %v", ack)
}
// 模拟设备收到 cmd 后以二进制分块回传cmd_data_start → 0x2×N → cmd_data_end
videoData := make([]byte, 20000) // 跨多个 8KB 块
for i := range videoData {
videoData[i] = byte(i % 251)
}
go func() {
time.Sleep(100 * time.Millisecond)
cli.sendText(mustJSON(map[string]interface{}{
"op": "cmd_data_start", "req_id": "req-video-1",
"kind": "camera_video", "mime": "video/mp4",
"total": len(videoData), "chunk_size": 8192,
}))
const chunk = 8192
for off := 0; off < len(videoData); off += chunk {
end := off + chunk
if end > len(videoData) {
end = len(videoData)
}
cli.sendBinary(videoData[off:end])
}
cli.sendText(mustJSON(map[string]interface{}{
"op": "cmd_data_end", "req_id": "req-video-1", "status": "ok", "total": len(videoData),
}))
}()
// 服务端等待聚合结果AwaitResult 由 deliverResult 唤醒)
res, err := reg.AwaitResult("req-video-1", 5*time.Second)
if err != nil {
t.Fatalf("await aggregated result: %v", err)
}
if res["status"] != "ok" {
t.Fatalf("expected status ok, got %v", res["status"])
}
if got, _ := res["size"].(int); got != len(videoData) {
t.Fatalf("size mismatch: got %v want %d", res["size"], len(videoData))
}
// 校验 base64 数据完整性
stored, ok := reg.GetResult("req-video-1")
if !ok {
t.Fatal("result not persisted")
}
b64, _ := stored["data_base64"].(string)
if len(b64) == 0 {
t.Fatal("data_base64 empty")
}
decoded, err := base64.StdEncoding.DecodeString(b64)
if err != nil {
t.Fatalf("decode base64: %v", err)
}
if string(decoded) != string(videoData) {
t.Fatal("decoded data mismatch")
}
}
// ===== 端到端PushData 下发音频(网关→设备 cmd_speech 协议)=====
func TestWSPushDataAudio(t *testing.T) {
reg := NewRegistry()
token := "test-token-456"
reg.SetAcceptToken(func(provided string) bool { return provided == token })
srv := httptest.NewServer(http.HandlerFunc(reg.ServeWS))
defer srv.Close()
cli := dialTestWS(t, srv.URL, token)
defer cli.close()
cli.sendText([]byte(`{"op":"hello","device":{"device_id":"audio-dev","name":"音频机","kind":"speaker"}}`))
if _, _, err := cli.readMsg(); err != nil {
t.Fatalf("read hello_ack: %v", err)
}
audioData := []byte("RIFF....fake-wav-audio-data-for-testing....")
// 异步下发音频
errCh := make(chan error, 1)
go func() {
errCh <- reg.PushData("audio-dev", "req-speech-1", "speech", "audio/wav", audioData)
}()
// 设备侧按协议读取start 文本帧 → N 个二进制帧 → end 文本帧
var start map[string]interface{}
var chunks [][]byte
var end map[string]interface{}
deadline := time.After(5 * time.Second)
for end == nil {
select {
case <-deadline:
t.Fatal("timeout reading speech protocol frames")
default:
}
op, payload, err := cli.readMsg()
if err != nil {
t.Fatalf("read frame: %v", err)
}
switch op {
case 0x1:
var msg map[string]interface{}
json.Unmarshal(payload, &msg)
switch msg["op"] {
case "cmd_speech_start":
start = msg
case "cmd_speech_end":
end = msg
}
case 0x2:
chunks = append(chunks, payload)
}
}
if err := <-errCh; err != nil {
t.Fatalf("PushData error: %v", err)
}
if start == nil || start["op"] != "cmd_speech_start" {
t.Fatal("missing cmd_speech_start")
}
if start["mime"] != "audio/wav" || start["kind"] != "speech" {
t.Fatalf("unexpected start fields: %v", start)
}
if int(start["total"].(float64)) != len(audioData) {
t.Fatalf("total mismatch: %v", start["total"])
}
if end["req_id"] != "req-speech-1" {
t.Fatalf("unexpected end: %v", end)
}
var got []byte
for _, c := range chunks {
got = append(got, c...)
}
if string(got) != string(audioData) {
t.Fatalf("audio data mismatch: got %d bytes want %d", len(got), len(audioData))
}
}
// ===== PushData 对离线设备报错 =====
func TestPushDataOfflineDevice(t *testing.T) {
reg := NewRegistry()
err := reg.PushData("no-such-device", "req-x", "speech", "audio/wav", []byte{1})
if err == nil || !strings.Contains(err.Error(), "not online") {
t.Fatalf("expected not online error, got %v", err)
}
}

View File

@ -59,8 +59,12 @@ func (d *devicectlDevice) Tools() []agentIO.ToolDef {
Description: "向设备下发命令/操作异步accepted=true 后用 device_ctl_cmdresult 轮询结果)。" +
"command 支持两类(前缀区分):\n" +
"- shell-cmd: 在设备上执行原生 shell 命令,如 shell-cmd ls -la /tmp\n" +
"- homeagent-cmd: 调用设备端 HomeAgent 内置能力,如 homeagent-camerasue调用用户侧摄像头" +
"homeagent-screensue(用户侧屏幕显示内容\n" +
"- homeagent-cmd: 调用设备端 HomeAgent 内置能力\n" +
" · homeagent-screensue <显示内容/HTML> — 用户侧屏幕弹窗显示自定义内容(带参显示,如 homeagent-screensue 会议提醒:三点开会\n" +
" · homeagent-screensue <秒> <内容> — 显示指定时长后自动关闭;不带参数则常驻\n" +
" · homeagent-camerasue — 抓拍单张 jpeg结果为 base64 data URL\n" +
" · homeagent-camerasue <N秒> — 录像 N 秒 mp4二进制分块回传cmdresult 含 data_base64 字段)\n" +
" · homeagent-speakeruse <文字> — 设备端 TTS 语音朗读文字\n" +
"⚡ 高危:设备必须已授权,且该操作会改变设备行为。" +
"返回 accepted=true 表示已下发并等待设备执行,之后可用 device_ctl_cmdresult 查询结果。" +
"若设备未授权或离线,返回错误信息。",
@ -68,7 +72,7 @@ func (d *devicectlDevice) Tools() []agentIO.ToolDef {
"type": "object",
"properties": map[string]interface{}{
"device_id": map[string]interface{}{"type": "string", "description": "目标设备 ID"},
"command": map[string]interface{}{"type": "string", "description": "以 shell-cmd 或 homeagent-cmd 前缀开头。如 shell-cmd pwd、homeagent-camerasue"},
"command": map[string]interface{}{"type": "string", "description": "以 shell-cmd 或 homeagent-cmd 前缀开头。如 shell-cmd pwd、homeagent-screensue 三点开会、homeagent-camerasue 5录5秒、homeagent-speakeruse 你好"},
},
"required": []interface{}{"device_id", "command"},
},

View File

@ -260,6 +260,45 @@ func (r *Registry) PushCmd(deviceID, reqID, command, cmdType string) error {
})
}
// PushData 向设备分块下发二进制数据(网关→设备,如 TTS 音频)。
// 协议(与 GUI 设备桥协商):
// 文本帧 cmd_speech_start {op, req_id, kind, mime, total} → N 个二进制帧(0x2, ≤8KB) → 文本帧 cmd_speech_end {op, req_id}
// kind 为语义标记(如 speechmime 为数据 MIME 类型。设备聚合后按自身能力处理(播放等)。
func (r *Registry) PushData(deviceID, reqID, kind, mime string, data []byte) error {
r.mu.RLock()
c, ok := r.conns[deviceID]
r.mu.RUnlock()
if !ok {
return fmt.Errorf("device %s not online", deviceID)
}
if err := writeText(c.w, mustJSON(map[string]interface{}{
"op": "cmd_speech_start",
"req_id": reqID,
"kind": kind,
"mime": mime,
"total": len(data),
})); err != nil {
return fmt.Errorf("push data start: %w", err)
}
const chunkSize = 8192
for off := 0; off < len(data); off += chunkSize {
end := off + chunkSize
if end > len(data) {
end = len(data)
}
if err := writeBinary(c.w, data[off:end]); err != nil {
return fmt.Errorf("push data chunk: %w", err)
}
}
if err := writeText(c.w, mustJSON(map[string]interface{}{
"op": "cmd_speech_end",
"req_id": reqID,
})); err != nil {
return fmt.Errorf("push data end: %w", err)
}
return nil
}
// AwaitResult 等待某请求的结果(带超时)。
func (r *Registry) AwaitResult(reqID string, timeout time.Duration) (map[string]interface{}, error) {
ch := make(chan map[string]interface{}, 1)
@ -342,43 +381,50 @@ func httpUpgrade(w http.ResponseWriter, r *http.Request) (net.Conn, *bufio.ReadW
return conn, rw, nil
}
func readFrame(r *bufio.Reader) ([]byte, bool, error) {
// readFrame 读取一个 WS 帧。返回 (payload, isClose, err)。
// opcode: 0x1 文本 / 0x2 二进制(设备→网关大体积数据分块,如录像回传)。
func readFrame(r *bufio.Reader) ([]byte, bool, byte, error) {
b0, err := r.ReadByte()
if err != nil {
return nil, true, err
return nil, true, 0, err
}
opcode := b0 & 0x0f
b1, err := r.ReadByte()
if err != nil {
return nil, true, err
return nil, true, opcode, err
}
masked := b1&0x80 != 0
length := uint64(b1 & 0x7f)
if length == 126 {
var ext [2]byte
if _, err := io.ReadFull(r, ext[:]); err != nil {
return nil, true, err
return nil, true, opcode, err
}
length = uint64(binary.BigEndian.Uint16(ext[:]))
} else if length == 127 {
var ext [8]byte
if _, err := io.ReadFull(r, ext[:]); err != nil {
return nil, true, err
return nil, true, opcode, err
}
length = binary.BigEndian.Uint64(ext[:])
}
if length > 1<<20 {
return nil, true, fmt.Errorf("frame too large")
// 二进制帧允许更大(录像分块聚合,单帧仍限 8MB 防滥用)
maxFrame := uint64(1 << 20)
if opcode == 0x2 {
maxFrame = 8 << 20
}
if length > maxFrame {
return nil, true, opcode, fmt.Errorf("frame too large")
}
var maskKey [4]byte
if masked {
if _, err := io.ReadFull(r, maskKey[:]); err != nil {
return nil, true, err
return nil, true, opcode, err
}
}
payload := make([]byte, length)
if _, err := io.ReadFull(r, payload); err != nil {
return nil, true, err
return nil, true, opcode, err
}
if masked {
for i := range payload {
@ -386,23 +432,32 @@ func readFrame(r *bufio.Reader) ([]byte, bool, error) {
}
}
switch opcode {
case 0x1:
return payload, false, nil
case 0x1, 0x2:
return payload, false, opcode, nil
case 0x8:
return nil, true, nil
return nil, true, opcode, nil
case 0xa:
return nil, false, nil
return nil, false, opcode, nil
case 0x9:
return nil, false, errPing
return nil, false, opcode, errPing
default:
return nil, false, fmt.Errorf("unsupported opcode %x", opcode)
return nil, false, opcode, fmt.Errorf("unsupported opcode %x", opcode)
}
}
var errPing = fmt.Errorf("ping")
func writeText(w *bufio.Writer, payload []byte) error {
if err := writeFrameHeader(w, 0x1, len(payload)); err != nil {
return writeFrame(w, 0x1, payload)
}
// writeBinary 发送 WS 二进制帧0x2网关→设备大体积数据如 TTS 音频)分块下发。
func writeBinary(w *bufio.Writer, payload []byte) error {
return writeFrame(w, 0x2, payload)
}
func writeFrame(w *bufio.Writer, opcode byte, payload []byte) error {
if err := writeFrameHeader(w, opcode, len(payload)); err != nil {
return err
}
if _, err := w.Write(payload); err != nil {
@ -491,8 +546,12 @@ func (r *Registry) handleWS(conn net.Conn, rw *bufio.ReadWriter) {
}
}()
// 二进制分块聚合状态(设备→网关,如录像回传):
// cmd_data_start 开启 → 0x2 帧追加 → cmd_data_end 聚合存入 cmdresult
var dataAccum *dataAccumulator
for {
payload, isClose, err := readFrame(rw.Reader)
payload, isClose, opcode, err := readFrame(rw.Reader)
if err != nil {
if err == errPing {
if werr := writePong(rw.Writer); werr != nil {
@ -505,6 +564,23 @@ func (r *Registry) handleWS(conn net.Conn, rw *bufio.ReadWriter) {
if isClose {
return
}
if opcode == 0x2 {
// 二进制帧:处于聚合状态时追加数据块,否则忽略
if dataAccum != nil {
dataAccum.chunks = append(dataAccum.chunks, payload)
dataAccum.got += len(payload)
// 防滥用:超出声明 total 的 2 倍或硬上限 64MB 时放弃聚合
limit := int64(dataAccum.total)*2 + 1024
if limit < 64<<20 {
limit = 64 << 20
}
if int64(dataAccum.got) > limit {
log.Printf("[remotedevice] data accumulation exceeded limit for req %s, dropped", dataAccum.reqID)
dataAccum = nil
}
}
continue
}
var msg map[string]interface{}
if err := json.Unmarshal(payload, &msg); err != nil {
continue
@ -563,10 +639,70 @@ func (r *Registry) handleWS(conn net.Conn, rw *bufio.ReadWriter) {
if reqID != "" {
r.deliverResult(reqID, msg)
}
case "cmd_data_start":
reqID, _ := msg["req_id"].(string)
if reqID == "" {
continue
}
total, _ := msg["total"].(float64)
kind, _ := msg["kind"].(string)
mime, _ := msg["mime"].(string)
dataAccum = &dataAccumulator{
reqID: reqID,
kind: kind,
mime: mime,
total: int(total),
}
case "cmd_data_end":
reqID, _ := msg["req_id"].(string)
status, _ := msg["status"].(string)
if dataAccum == nil || dataAccum.reqID != reqID {
continue
}
acc := dataAccum
dataAccum = nil
if status != "ok" {
r.SaveResult(reqID, map[string]interface{}{
"op": "cmd_result", "req_id": reqID, "status": "error",
"error": "device reported transfer failure",
})
r.deliverResult(reqID, map[string]interface{}{
"op": "cmd_result", "req_id": reqID, "status": "error",
"error": "device reported transfer failure",
})
continue
}
data := make([]byte, 0, acc.got)
for _, c := range acc.chunks {
data = append(data, c...)
}
res := map[string]interface{}{
"op": "cmd_result",
"req_id": reqID,
"status": "ok",
"kind": acc.kind,
"mime": acc.mime,
"size": len(data),
"expected": acc.total,
// base64 编码完整二进制(录像 mp4 等),供 agent/上层取回后解码使用
"data_base64": base64.StdEncoding.EncodeToString(data),
}
r.SaveResult(reqID, res)
r.deliverResult(reqID, res)
}
}
}
// dataAccumulator 聚合设备→网关的二进制分块传输(如录像回传)。
type dataAccumulator struct {
reqID string
kind string
mime string
total int
chunks [][]byte
got int
}
func metaFromMsg(msg map[string]interface{}) DeviceMeta {
var meta DeviceMeta
if d, ok := msg["device"].(map[string]interface{}); ok {