feat(upload): implement WeCom media upload via WebSocket API

pkg/agent/loop.go

@@ -193,6 +193,7 @@ func registerSharedTools(
 					Channel: channel,
 					ChatID:  chatID,
 					Content: content,
+					Final:   true,
 				})
 			})
 			agent.Tools.Register(messageTool)
@@ -315,6 +316,7 @@ func (al *AgentLoop) Run(ctx context.Context) error {
 						Channel: msg.Channel,
 						ChatID:  msg.ChatID,
 						Content: response,
+						Final:   true,
 					})
 					logger.InfoCF("agent", "Published outbound response",
 						map[string]any{

pkg/agent/loop_mcp.go

@@ -64,7 +64,7 @@ func (al *AgentLoop) ensureMCPInitialized(ctx context.Context) error {
 		return nil
 	}
 
-	if al.cfg.Tools.MCP.Servers == nil || len(al.cfg.Tools.MCP.Servers) == 0 {
+	if len(al.cfg.Tools.MCP.Servers) == 0 {
 		logger.WarnCF("agent", "MCP is enabled but no servers are configured, skipping MCP initialization", nil)
 		return nil
 	}

pkg/bus/types.go

@@ -34,6 +34,7 @@ type OutboundMessage struct {
 	ChatID           string `json:"chat_id"`
 	Content          string `json:"content"`
 	ReplyToMessageID string `json:"reply_to_message_id,omitempty"`
+	Final            bool   `json:"final,omitempty"`
 }
 
 // MediaPart describes a single media attachment to send.

pkg/channels/wecom/aibot_ws.go

@@ -38,10 +38,8 @@ const (
 	wsInitialReconnect  = time.Second
 
 	// WeCom requires finish=true within 6 minutes of the first stream frame.
-	// wsStreamTickInterval controls how often we send an in-progress hint.
 	// wsStreamMaxDuration is a safety margin below the 6-minute hard limit.
-	wsStreamTickInterval = 30 * time.Second
-	wsStreamMaxDuration  = 5*time.Minute + 30*time.Second
+	wsStreamMaxDuration = 5*time.Minute + 30*time.Second
 
 	// wsImageDownloadTimeout caps the time we spend downloading an inbound image.
 	wsImageDownloadTimeout = 30 * time.Second
@@ -286,6 +284,11 @@ func (c *WeComAIBotWSChannel) Send(ctx context.Context, msg bus.OutboundMessage)
 	// and there will be no matching entry in reqStates; fall through to proactive push.
 	task, route, ok := c.getReqState(msg.ChatID)
 	if !ok {
+		if !msg.Final {
+			// Intermediate message (tool feedback, etc.) with no active WS session.
+			// Nothing to stream to; silently drop it.
+			return nil
+		}
 		// No req_id record found — this is a cron/scheduler-originated message.
 		// Send it as a proactive markdown push using the chat ID directly.
 		logger.InfoCF("wecom_aibot", "Send: no req_id state, delivering via proactive push (cron/scheduler)",
@@ -299,6 +302,12 @@ func (c *WeComAIBotWSChannel) Send(ctx context.Context, msg bus.OutboundMessage)
 	}
 
 	if task == nil {
+		if !msg.Final {
+			// Intermediate message: stream for this req_id has already closed (deadline
+			// or new-message cancel).  Nothing to do.
+			return nil
+		}
+
 		if time.Now().Before(route.ReadyAt) {
 			// Keep using aibot_respond_msg within stream window; do not proactively
 			// push unless wsStreamMaxDuration has elapsed.
@@ -318,6 +327,19 @@ func (c *WeComAIBotWSChannel) Send(ctx context.Context, msg bus.OutboundMessage)
 		return nil
 	}
 
+	if !msg.Final {
+		// Intermediate message (tool feedback, context notices, etc.) with an active
+		// stream.  Send directly as a non-finishing stream chunk so the user sees
+		// live progress without touching answerCh (which the goroutine reserves for
+		// the single final answer).
+		for _, chunk := range splitWSContent(msg.Content, wsStreamMaxContentBytes) {
+			c.wsSendStreamChunk(task.ReqID, task.StreamID, false, chunk)
+		}
+		return nil
+	}
+
+	// Final answer: deliver to the waiting goroutine via answerCh, which will
+	// send it with finish=true and clean up the stream state.
 	// Non-blocking fast path: when answerCh has space, deliver without racing
 	// against task.ctx.Done() (which fires when the task is canceled by a new
 	// incoming message, but the response must still be sent).
@@ -852,37 +874,34 @@ func (c *WeComAIBotWSChannel) dispatchWSAgentTask(
 		c.HandleMessage(taskCtx, peer, reqID, userID, reqID,
 			content, mediaRefs, metadata, sender)
 
-		// Wait for the agent reply. While waiting, send periodic finish=false
-		// hints so the user knows processing is still in progress.
+		// Wait for the agent reply.
 		// WeCom requires finish=true within 6 minutes of the first stream frame;
 		// wsStreamMaxDuration enforces that limit with a safety margin.
-		waitHints := []string{
-			"⏳ Processing, please wait...",
-			"⏳ Still processing, please wait...",
-			"⏳ Almost there, please wait...",
-		}
-		ticker := time.NewTicker(wsStreamTickInterval)
-		defer ticker.Stop()
 		deadlineTimer := time.NewTimer(wsStreamMaxDuration)
 		defer deadlineTimer.Stop()
-		tickCount := 0
 		for {
 			select {
 			case answer := <-task.answerCh:
-				// Split the answer into byte-bounded chunks and send as stream frames.
-				// All but the last carry finish=false; the final frame closes the stream.
+
+				// Split the answer into byte-bounded chunks.
+				// Each chunk must fit within wsStreamMaxContentBytes (20480 bytes).
+				// WeCom's stream update semantics REPLACE content (not append), so all
+				// chunks beyond the first need a fresh stream.id — otherwise only the
+				// last chunk would be visible to the user.
+				//
+				// Delivery plan:
+				//   chunk[0] — close the already-opened stream (finish=true, same streamID)
+				//   chunk[1+] — create + immediately close a new stream per chunk
 				chunks := splitWSContent(answer, wsStreamMaxContentBytes)
 				for i, chunk := range chunks {
-					c.wsSendStreamChunk(reqID, streamID, i == len(chunks)-1, chunk)
+					sid := streamID
+					if i > 0 {
+						sid = wsGenerateID()
+					}
+					c.wsSendStreamChunk(reqID, sid, true, chunk)
 				}
 				c.deleteReqState(reqID)
 				return
-			case <-ticker.C:
-				hint := waitHints[tickCount%len(waitHints)]
-				tickCount++
-				logger.DebugCF("wecom_aibot", "Sending stream progress hint",
-					map[string]any{"chat_id": actualChatID, "tick": tickCount})
-				c.wsSendStreamChunk(reqID, streamID, false, hint)
 			case <-deadlineTimer.C:
 				logger.WarnCF("wecom_aibot",
 					"Stream response deadline reached, closing stream; late reply will be pushed",
@@ -1028,26 +1047,10 @@ func (c *WeComAIBotWSChannel) wsSendActivePush(chatID string, chatType uint32, c
 }
 
 // writeWSAndWait writes cmd to the active connection and validates the command response.
+// It is a thin wrapper around callWSCommand that discards the response body.
 func (c *WeComAIBotWSChannel) writeWSAndWait(cmd wsCommand, timeout time.Duration) error {
-	if cmd.Headers.ReqID == "" {
-		return fmt.Errorf("req_id is empty")
-	}
-
-	c.connMu.Lock()
-	conn := c.conn
-	c.connMu.Unlock()
-	if conn == nil {
-		return fmt.Errorf("websocket not connected")
-	}
-
-	resp, err := c.sendAndWait(conn, cmd.Headers.ReqID, cmd, timeout)
-	if err != nil {
-		return err
-	}
-	if resp.ErrCode != 0 {
-		return fmt.Errorf("%s rejected (errcode=%d): %s", cmd.Cmd, resp.ErrCode, resp.ErrMsg)
-	}
-	return nil
+	_, err := c.callWSCommand(cmd, timeout)
+	return err
 }
 
 // cancelAllTasks cancels every pending agent task; called when the connection drops.

pkg/channels/wecom/aibot_ws_upload.go

@@ -0,0 +1,234 @@
+package wecom
+
+import (
+	"context"
+	"crypto/md5"
+	"encoding/base64"
+	"encoding/json"
+	"fmt"
+	"os"
+	"path/filepath"
+	"time"
+
+	"github.com/sipeed/picoclaw/pkg/logger"
+)
+
+// Upload temporary media constants.
+// Ref: https://developer.work.weixin.qq.com/document/path/101463#上传临时素材
+const (
+	// wsUploadChunkSize is the maximum raw (pre-base64) bytes per chunk.
+	// The API limit is 512 KB before base64 encoding.
+	wsUploadChunkSize = 512 << 10 // 512 KB
+
+	// wsUploadMaxChunks is the maximum number of chunks per upload session.
+	wsUploadMaxChunks = 100
+
+	wsUploadInitTimeout   = 30 * time.Second
+	wsUploadChunkTimeout  = 60 * time.Second // larger: base64 payload may be big
+	wsUploadFinishTimeout = 30 * time.Second
+)
+
+// ---- Request / response body types ----
+
+// wsUploadInitBody is the body for aibot_upload_media_init.
+type wsUploadInitBody struct {
+	Type        string `json:"type"`
+	Filename    string `json:"filename"`
+	TotalSize   int64  `json:"total_size"`
+	TotalChunks int    `json:"total_chunks"`
+	MD5         string `json:"md5,omitempty"`
+}
+
+// wsUploadInitResponse is the body received in the aibot_upload_media_init response.
+type wsUploadInitResponse struct {
+	UploadID string `json:"upload_id"`
+}
+
+// wsUploadChunkBody is the body for aibot_upload_media_chunk.
+type wsUploadChunkBody struct {
+	UploadID   string `json:"upload_id"`
+	ChunkIndex int    `json:"chunk_index"`
+	Base64Data string `json:"base64_data"`
+}
+
+// wsUploadFinishBody is the body for aibot_upload_media_finish.
+type wsUploadFinishBody struct {
+	UploadID string `json:"upload_id"`
+}
+
+// wsUploadFinishResponse is the body received in the aibot_upload_media_finish response.
+type wsUploadFinishResponse struct {
+	Type      string `json:"type"`
+	MediaID   string `json:"media_id"`
+	CreatedAt string `json:"created_at"`
+}
+
+// ---- Public API ----
+
+// UploadWSMedia uploads a local file as a temporary WeCom media asset via the
+// WebSocket long-connection API and returns the resulting media_id (valid 3 days).
+//
+// mediaType must be one of:
+//   - "image"  — PNG / JPG(JPEG) / GIF, ≤ 2 MB
+//   - "voice"  — AMR, ≤ 2 MB
+//   - "video"  — MP4, ≤ 10 MB
+//   - "file"   — any format, ≤ 20 MB
+//
+// filename is the logical file name sent to WeCom. When empty it is derived
+// from the base name of filePath.
+//
+// The upload is split into at most wsUploadMaxChunks chunks of wsUploadChunkSize
+// bytes each. The full operation must complete within wsUploadSessionTTL (30 min).
+// Each individual command call is also independently time-bounded.
+func (c *WeComAIBotWSChannel) UploadWSMedia(
+	ctx context.Context,
+	filePath, filename, mediaType string,
+) (string, error) {
+	// ---- Validate media type ----
+	switch mediaType {
+	case "image", "voice", "video", "file":
+	default:
+		return "", fmt.Errorf("unsupported media type %q: must be image, voice, video, or file", mediaType)
+	}
+
+	// ---- Read file ----
+	data, err := os.ReadFile(filePath)
+	if err != nil {
+		return "", fmt.Errorf("read file: %w", err)
+	}
+	totalSize := int64(len(data))
+	if totalSize == 0 {
+		return "", fmt.Errorf("file is empty: %s", filePath)
+	}
+
+	if filename == "" {
+		filename = filepath.Base(filePath)
+	}
+
+	// ---- Compute chunk count ----
+	totalChunks := int((totalSize + wsUploadChunkSize - 1) / wsUploadChunkSize)
+	if totalChunks > wsUploadMaxChunks {
+		return "", fmt.Errorf(
+			"file too large: requires %d chunks but limit is %d (max ~%d MB)",
+			totalChunks, wsUploadMaxChunks,
+			int64(wsUploadMaxChunks)*wsUploadChunkSize>>20,
+		)
+	}
+
+	// ---- Compute MD5 ----
+	rawMD5 := md5.Sum(data) //nolint:gosec // MD5 is required by the WeCom API spec
+	fileMD5 := fmt.Sprintf("%x", rawMD5)
+
+	// ---- Step 1: Init ----
+	initEnv, err := c.callWSCommand(wsCommand{
+		Cmd:     "aibot_upload_media_init",
+		Headers: wsHeaders{ReqID: wsGenerateID()},
+		Body: wsUploadInitBody{
+			Type:        mediaType,
+			Filename:    filename,
+			TotalSize:   totalSize,
+			TotalChunks: totalChunks,
+			MD5:         fileMD5,
+		},
+	}, wsUploadInitTimeout)
+	if err != nil {
+		return "", fmt.Errorf("upload init: %w", err)
+	}
+	var initResp wsUploadInitResponse
+	if err = json.Unmarshal(initEnv.Body, &initResp); err != nil {
+		return "", fmt.Errorf("parse upload init response: %w", err)
+	}
+	if initResp.UploadID == "" {
+		return "", fmt.Errorf("upload init returned empty upload_id")
+	}
+	uploadID := initResp.UploadID
+
+	logger.InfoCF("wecom_aibot", "Media upload initialized", map[string]any{
+		"upload_id":    uploadID,
+		"type":         mediaType,
+		"filename":     filename,
+		"total_size":   totalSize,
+		"total_chunks": totalChunks,
+	})
+
+	// ---- Step 2: Upload chunks ----
+	for i := 0; i < totalChunks; i++ {
+		if ctxErr := ctx.Err(); ctxErr != nil {
+			return "", fmt.Errorf("upload aborted before chunk %d: %w", i, ctxErr)
+		}
+
+		start := int64(i) * wsUploadChunkSize
+		end := start + wsUploadChunkSize
+		if end > totalSize {
+			end = totalSize
+		}
+
+		_, err = c.callWSCommand(wsCommand{
+			Cmd:     "aibot_upload_media_chunk",
+			Headers: wsHeaders{ReqID: wsGenerateID()},
+			Body: wsUploadChunkBody{
+				UploadID:   uploadID,
+				ChunkIndex: i,
+				Base64Data: base64.StdEncoding.EncodeToString(data[start:end]),
+			},
+		}, wsUploadChunkTimeout)
+		if err != nil {
+			return "", fmt.Errorf("upload chunk %d/%d: %w", i, totalChunks-1, err)
+		}
+
+		logger.DebugCF("wecom_aibot", "Media chunk uploaded", map[string]any{
+			"upload_id": uploadID,
+			"chunk":     i,
+			"total":     totalChunks,
+		})
+	}
+
+	// ---- Step 3: Finish ----
+	finishEnv, err := c.callWSCommand(wsCommand{
+		Cmd:     "aibot_upload_media_finish",
+		Headers: wsHeaders{ReqID: wsGenerateID()},
+		Body:    wsUploadFinishBody{UploadID: uploadID},
+	}, wsUploadFinishTimeout)
+	if err != nil {
+		return "", fmt.Errorf("upload finish: %w", err)
+	}
+	var finishResp wsUploadFinishResponse
+	if err := json.Unmarshal(finishEnv.Body, &finishResp); err != nil {
+		return "", fmt.Errorf("parse upload finish response: %w", err)
+	}
+	if finishResp.MediaID == "" {
+		return "", fmt.Errorf("upload finish returned empty media_id")
+	}
+
+	logger.InfoCF("wecom_aibot", "Media upload complete", map[string]any{
+		"upload_id": uploadID,
+		"media_id":  finishResp.MediaID,
+		"type":      finishResp.Type,
+	})
+	return finishResp.MediaID, nil
+}
+
+// ---- Internal helper ----
+
+// callWSCommand sends a WebSocket command and returns the raw server envelope.
+// It validates the errcode field and returns an error when non-zero.
+// Use callWSCommand (over writeWSAndWait) when the response body must be inspected.
+func (c *WeComAIBotWSChannel) callWSCommand(cmd wsCommand, timeout time.Duration) (wsEnvelope, error) {
+	if cmd.Headers.ReqID == "" {
+		return wsEnvelope{}, fmt.Errorf("req_id is empty")
+	}
+	c.connMu.Lock()
+	conn := c.conn
+	c.connMu.Unlock()
+	if conn == nil {
+		return wsEnvelope{}, fmt.Errorf("websocket not connected")
+	}
+	env, err := c.sendAndWait(conn, cmd.Headers.ReqID, cmd, timeout)
+	if err != nil {
+		return wsEnvelope{}, err
+	}
+	if env.ErrCode != 0 {
+		return wsEnvelope{}, fmt.Errorf("%s rejected (errcode=%d): %s", cmd.Cmd, env.ErrCode, env.ErrMsg)
+	}
+	return env, nil
+}

pkg/channels/wecom/aibot_ws_upload_test.go

@@ -0,0 +1,268 @@
+package wecom
+
+import (
+	"context"
+	"encoding/base64"
+	"encoding/json"
+	"os"
+	"path/filepath"
+	"testing"
+	"time"
+)
+
+// ---------------------------------------------------------------------------
+// UploadWSMedia — pre-flight validation tests (no real WS required)
+// ---------------------------------------------------------------------------
+
+// TestUploadWSMedia_InvalidMediaType verifies that unsupported media types are
+// rejected before any network activity occurs.
+func TestUploadWSMedia_InvalidMediaType(t *testing.T) {
+	ch := newTestWSChannel(t)
+
+	for _, bad := range []string{"", "pdf", "audio", "IMAGE", "gif"} {
+		_, err := ch.UploadWSMedia(context.Background(), "/dev/null", "test.pdf", bad)
+		if err == nil {
+			t.Errorf("expected error for media type %q, got nil", bad)
+		}
+	}
+}
+
+// TestUploadWSMedia_FileNotFound verifies that a non-existent path returns an error.
+func TestUploadWSMedia_FileNotFound(t *testing.T) {
+	ch := newTestWSChannel(t)
+	_, err := ch.UploadWSMedia(context.Background(), "/nonexistent/path/to/file.bin", "", "file")
+	if err == nil {
+		t.Fatal("expected error for missing file, got nil")
+	}
+}
+
+// TestUploadWSMedia_EmptyFile verifies that an empty file is rejected before
+// the upload starts.
+func TestUploadWSMedia_EmptyFile(t *testing.T) {
+	ch := newTestWSChannel(t)
+
+	tmp, err := os.CreateTemp(t.TempDir(), "empty-*.bin")
+	if err != nil {
+		t.Fatal(err)
+	}
+	tmp.Close()
+
+	_, err = ch.UploadWSMedia(context.Background(), tmp.Name(), "", "file")
+	if err == nil {
+		t.Fatal("expected error for empty file, got nil")
+	}
+}
+
+// TestUploadWSMedia_FileTooLarge verifies that a file requiring more than
+// wsUploadMaxChunks chunks is rejected before the upload starts.
+func TestUploadWSMedia_FileTooLarge(t *testing.T) {
+	ch := newTestWSChannel(t)
+
+	// Synthesize a file that would need 101 chunks (one byte over the limit).
+	size := int64(wsUploadMaxChunks)*int64(wsUploadChunkSize) + 1
+
+	tmp, err := os.CreateTemp(t.TempDir(), "large-*.bin")
+	if err != nil {
+		t.Fatal(err)
+	}
+	truncErr := tmp.Truncate(size)
+	tmp.Close()
+	if truncErr != nil {
+		t.Fatal(truncErr)
+	}
+
+	_, err = ch.UploadWSMedia(context.Background(), tmp.Name(), "", "file")
+	if err == nil {
+		t.Fatal("expected error for oversized file, got nil")
+	}
+}
+
+// TestUploadWSMedia_NotConnected verifies that calling UploadWSMedia when no
+// WebSocket connection is active returns a clear error.
+func TestUploadWSMedia_NotConnected(t *testing.T) {
+	ch := newTestWSChannel(t)
+
+	tmp := filepath.Join(t.TempDir(), "test.txt")
+	if err := os.WriteFile(tmp, []byte("hello world"), 0o600); err != nil {
+		t.Fatal(err)
+	}
+
+	// conn is nil by default in a test channel — callWSCommand must return an error.
+	_, err := ch.UploadWSMedia(context.Background(), tmp, "", "file")
+	if err == nil {
+		t.Fatal("expected error when conn is nil, got nil")
+	}
+}
+
+// TestUploadWSMedia_ContextCanceledPreFlight verifies that a pre-canceled
+// context is detected during chunk iteration (even before any WS call, when
+// the connection is nil the WS error takes precedence for a single-chunk file,
+// so we test this path with a file that triggers ctx.Err() first).
+func TestUploadWSMedia_ContextCanceledPreFlight(t *testing.T) {
+	ch := newTestWSChannel(t)
+
+	tmp := filepath.Join(t.TempDir(), "test.txt")
+	if err := os.WriteFile(tmp, []byte("hello"), 0o600); err != nil {
+		t.Fatal(err)
+	}
+
+	ctx, cancel := context.WithCancel(context.Background())
+	cancel() // pre-cancel before the call
+
+	// With conn == nil, callWSCommand will return "websocket not connected".
+	// The pre-canceled context is also an error; either one satisfies the test.
+	_, err := ch.UploadWSMedia(ctx, tmp, "", "file")
+	if err == nil {
+		t.Fatal("expected error for canceled context or nil conn, got nil")
+	}
+}
+
+// ---------------------------------------------------------------------------
+// callWSCommand — unit tests without a live connection
+// ---------------------------------------------------------------------------
+
+// TestCallWSCommand_EmptyReqID verifies that callWSCommand rejects a command
+// whose req_id is empty before touching the connection.
+func TestCallWSCommand_EmptyReqID(t *testing.T) {
+	ch := newTestWSChannel(t)
+	cmd := wsCommand{
+		Cmd:  "aibot_upload_media_init",
+		Body: map[string]string{"type": "file"},
+		// Headers.ReqID intentionally left empty
+	}
+	_, err := ch.callWSCommand(cmd, 5*time.Second)
+	if err == nil {
+		t.Fatal("expected error for empty req_id, got nil")
+	}
+}
+
+// TestCallWSCommand_NotConnected verifies that callWSCommand returns a clear
+// error when the WebSocket connection is nil.
+func TestCallWSCommand_NotConnected(t *testing.T) {
+	ch := newTestWSChannel(t)
+	cmd := wsCommand{
+		Cmd:     "aibot_upload_media_init",
+		Headers: wsHeaders{ReqID: "test-req-id"},
+		Body:    map[string]string{"type": "file"},
+	}
+	_, err := ch.callWSCommand(cmd, 5*time.Second)
+	if err == nil {
+		t.Fatal("expected error when conn is nil, got nil")
+	}
+}
+
+// ---------------------------------------------------------------------------
+// Upload body / response type round-trip tests
+// ---------------------------------------------------------------------------
+
+// TestWsUploadChunkCount verifies the integer chunk-count formula used inside
+// UploadWSMedia without needing the math package.
+func TestWsUploadChunkCount(t *testing.T) {
+	tests := []struct {
+		name       string
+		size       int64
+		wantChunks int
+	}{
+		{"one byte", 1, 1},
+		{"exactly one chunk", int64(wsUploadChunkSize), 1},
+		{"one byte over", int64(wsUploadChunkSize) + 1, 2},
+		{"max allowed", int64(wsUploadMaxChunks) * int64(wsUploadChunkSize), wsUploadMaxChunks},
+	}
+	for _, tc := range tests {
+		t.Run(tc.name, func(t *testing.T) {
+			got := int((tc.size + int64(wsUploadChunkSize) - 1) / int64(wsUploadChunkSize))
+			if got != tc.wantChunks {
+				t.Errorf("size=%d: got %d chunks, want %d", tc.size, got, tc.wantChunks)
+			}
+		})
+	}
+}
+
+// TestWsUploadInitBodyJSON checks that wsUploadInitBody serializes and
+// deserializes correctly, including the omitempty MD5 field.
+func TestWsUploadInitBodyJSON(t *testing.T) {
+	b := wsUploadInitBody{
+		Type:        "image",
+		Filename:    "photo.jpg",
+		TotalSize:   1024,
+		TotalChunks: 1,
+		MD5:         "deadbeefdeadbeef",
+	}
+	data, err := json.Marshal(b)
+	if err != nil {
+		t.Fatalf("marshal: %v", err)
+	}
+	var got wsUploadInitBody
+	if err := json.Unmarshal(data, &got); err != nil {
+		t.Fatalf("unmarshal: %v", err)
+	}
+	if got != b {
+		t.Errorf("round-trip mismatch: got %+v, want %+v", got, b)
+	}
+
+	// MD5 must be omitted when empty.
+	b.MD5 = ""
+	data, _ = json.Marshal(b)
+	if contains(string(data), `"md5"`) {
+		t.Errorf("expected md5 key to be absent when empty, got: %s", data)
+	}
+}
+
+// TestWsUploadChunkBodyBase64 checks that wsUploadChunkBody correctly
+// round-trips arbitrary binary data through base64.
+func TestWsUploadChunkBodyBase64(t *testing.T) {
+	raw := []byte("binary\x00data\xff\xfe")
+	body := wsUploadChunkBody{
+		UploadID:   "uid-test-1",
+		ChunkIndex: 0,
+		Base64Data: base64.StdEncoding.EncodeToString(raw),
+	}
+	encoded, err := json.Marshal(body)
+	if err != nil {
+		t.Fatalf("marshal: %v", err)
+	}
+	var got wsUploadChunkBody
+	if err = json.Unmarshal(encoded, &got); err != nil {
+		t.Fatalf("unmarshal: %v", err)
+	}
+	decoded, err := base64.StdEncoding.DecodeString(got.Base64Data)
+	if err != nil {
+		t.Fatalf("decode base64: %v", err)
+	}
+	if string(decoded) != string(raw) {
+		t.Errorf("base64 round-trip mismatch: got %v, want %v", decoded, raw)
+	}
+}
+
+// TestWsUploadFinishResponseJSON checks that the finish response unmarshals
+// the fields WeCom returns.
+func TestWsUploadFinishResponseJSON(t *testing.T) {
+	raw := `{"type":"file","media_id":"MEDIAID_ABCDE","created_at":"1680000000"}`
+	var resp wsUploadFinishResponse
+	if err := json.Unmarshal([]byte(raw), &resp); err != nil {
+		t.Fatalf("unmarshal: %v", err)
+	}
+	if resp.MediaID != "MEDIAID_ABCDE" {
+		t.Errorf("media_id: got %q, want %q", resp.MediaID, "MEDIAID_ABCDE")
+	}
+	if resp.Type != "file" {
+		t.Errorf("type: got %q, want %q", resp.Type, "file")
+	}
+	if resp.CreatedAt != "1680000000" {
+		t.Errorf("created_at: got %q, want %q", resp.CreatedAt, "1680000000")
+	}
+}
+
+// contains is a helper to avoid importing "strings" for a single check.
+func contains(s, sub string) bool {
+	return len(s) >= len(sub) && (s == sub || len(s) > 0 && containsHelper(s, sub))
+}
+
+func containsHelper(s, sub string) bool {
+	for i := 0; i <= len(s)-len(sub); i++ {
+		if s[i:i+len(sub)] == sub {
+			return true
+		}
+	}
+	return false
+}