myagent-ai 1.9.2 → 1.9.4

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "myagent-ai",
3
- "version": "1.9.2",
3
+ "version": "1.9.4",
4
4
  "description": "本地桌面端执行型AI助手 - Open Interpreter 风格 | Local Desktop Execution-Oriented AI Assistant",
5
5
  "main": "main.py",
6
6
  "bin": {
@@ -3,7 +3,7 @@
3
3
 
4
4
  (function() {
5
5
  var base = new URL('.', window.location.href).href;
6
- var v = '?v=5';
6
+ var v = '?v=6';
7
7
  var files = [
8
8
  { url: base + 'left_sessions.html' + v, container: 'leftSessionsContainer' },
9
9
  { url: base + 'middle_chat.html' + v, container: 'middleChatContainer' },
@@ -620,17 +620,17 @@ function toggleExecEventsPanel(header) {
620
620
  function renderInlineExecEvent(data, msgIdx) {
621
621
  // V2 Tool Event handling (called with full part: {type:'v2_tool', data:{...}})
622
622
  if (data.type === 'v2_tool') {
623
- var inner = data.data;
624
- var isStart = inner.status === 'running' || inner.type === 'tool_start';
625
- var iconEmoji = isStart ? '⚡' : (inner.success ? '✅' : '❌');
626
- var title = inner.title || (inner.tool_name || '工具调用');
627
- var metaParts = [];
623
+ let inner = data.data;
624
+ let isStart = inner.status === 'running' || inner.type === 'tool_start';
625
+ let iconEmoji = isStart ? '⚡' : (inner.success ? '✅' : '❌');
626
+ let title = inner.title || (inner.tool_name || '工具调用');
627
+ let metaParts = [];
628
628
  if (inner.tool_name) metaParts.push(escapeHtml(inner.tool_name));
629
629
  if (inner.timeout) metaParts.push('超时 ' + inner.timeout + 's');
630
630
  if (inner.callback !== undefined) metaParts.push(inner.callback ? '回调LLM' : '无回调');
631
- var metaText = metaParts.join(' · ');
631
+ let metaText = metaParts.join(' · ');
632
632
 
633
- var bodyHtml = '';
633
+ let bodyHtml = '';
634
634
  if (inner.type === 'tool_result' && inner.summary) {
635
635
  bodyHtml += '<div class="inline-exec-summary">' + escapeHtml(inner.summary) + '</div>';
636
636
  }
@@ -639,7 +639,7 @@ function renderInlineExecEvent(data, msgIdx) {
639
639
  }
640
640
  // Show params for tool_start
641
641
  if (inner.type === 'tool_start' && inner.params) {
642
- var paramPreview = typeof inner.params === 'string' ? inner.params : JSON.stringify(inner.params);
642
+ let paramPreview = typeof inner.params === 'string' ? inner.params : JSON.stringify(inner.params);
643
643
  if (paramPreview.length > 200) paramPreview = paramPreview.substring(0, 200) + '...';
644
644
  bodyHtml += '<div class="inline-exec-code">' + escapeHtml(paramPreview) + '</div>';
645
645
  }
@@ -814,6 +814,31 @@ function showToolResultModal(msgIndex, eventId) {
814
814
  document.body.appendChild(overlay);
815
815
  }
816
816
 
817
+ // ══════════════════════════════════════════════════════
818
+ // ── V2 Content Assembler (V2 内容组装) ──
819
+ // ══════════════════════════════════════════════════════
820
+
821
+ function _assembleV2Content(msg, msgParts) {
822
+ // Priority 1: V2 reasoning text (user-facing response from v2_reasoning events)
823
+ if (msg._v2Reasoning && msg._v2Reasoning.trim()) {
824
+ return msg._v2Reasoning.trim();
825
+ }
826
+ // Priority 2: V2 ask user text
827
+ if (msg._askUser && msg._askUser.trim()) {
828
+ return msg._askUser.trim();
829
+ }
830
+ // Priority 3: V1 text parts (backward compat — non-V2 mode)
831
+ var textParts = msgParts.filter(function(p) { return p.type === 'text'; });
832
+ if (textParts.length > 0) {
833
+ return textParts.map(function(p) { return p.content; }).join('\n\n');
834
+ }
835
+ // Priority 4: raw content from message (server-stored response)
836
+ if (msg.content && msg.content.trim() && msg.content !== '(无回复)') {
837
+ return msg.content.trim();
838
+ }
839
+ return '(无回复)';
840
+ }
841
+
817
842
  // ══════════════════════════════════════════════════════
818
843
  // ── Send Message (核心 SSE 流式消息发送) ──
819
844
  // ══════════════════════════════════════════════════════
@@ -827,7 +852,8 @@ async function sendMessage() {
827
852
  // ── 如果正在生成,弹出处理选择框 ──
828
853
  if (state.isGenerating) {
829
854
  state.tempInputText = text;
830
- document.getElementById('injectModal').style.display = 'flex';
855
+ var modal = document.getElementById('injectModal');
856
+ if (modal) { modal.style.display = 'flex'; }
831
857
  return;
832
858
  }
833
859
 
@@ -903,10 +929,17 @@ async function sendMessage() {
903
929
  let msgIdx = state.messages.length;
904
930
  let sessionIdReceived = sessionId;
905
931
  let fullThought = '';
932
+ let _isV2Mode = false; // Track whether V2 structured output events are received
933
+ let _v2RawXml = ''; // In V2 mode, accumulate raw XML from text_delta separately
934
+ let _v2ReasoningText = ''; // Accumulated V2 reasoning text for content
906
935
 
907
936
  function flushCurrentText() {
908
937
  if (currentText.trim()) {
909
- msgParts.push({type: 'text', content: currentText});
938
+ // In V2 mode, text_delta contains raw XML — do NOT add to msgParts as user-facing text
939
+ if (!_isV2Mode) {
940
+ msgParts.push({type: 'text', content: currentText});
941
+ }
942
+ // Always accumulate for backward compat content field
910
943
  }
911
944
  currentText = '';
912
945
  }
@@ -949,12 +982,16 @@ async function sendMessage() {
949
982
  } else if (evt.type === 'text_delta') {
950
983
  // Incremental streaming token
951
984
  currentText += evt.content;
952
- // Build backward-compat content from all parts + streaming text
953
- const allText = msgParts.filter(p => p.type === 'text').map(p => p.content).join('\n\n')
954
- + (currentText.trim() ? '\n\n' + currentText : '');
985
+ if (_isV2Mode) {
986
+ // In V2 mode, text_delta contains raw XML store separately, not as user-facing content
987
+ _v2RawXml = currentText;
988
+ // Show V2 reasoning text as streaming content (if available)
989
+ state.messages[msgIdx]._streamingText = _v2ReasoningText || '';
990
+ } else {
991
+ // V1 mode: text_delta IS the user-facing content
992
+ state.messages[msgIdx]._streamingText = currentText;
993
+ }
955
994
  state.messages[msgIdx].parts = [...msgParts];
956
- state.messages[msgIdx]._streamingText = currentText;
957
- state.messages[msgIdx].content = allText;
958
995
  throttledStreamUpdate(msgIdx);
959
996
  // ── 分段流式 TTS:推送增量文本 ──
960
997
  if (ttsManager.enabled && !ttsManager._streamActive) {
@@ -981,7 +1018,7 @@ async function sendMessage() {
981
1018
  if (state.messages[msgIdx]) {
982
1019
  state.messages[msgIdx].streaming = false;
983
1020
  state.messages[msgIdx].parts = [...msgParts];
984
- state.messages[msgIdx].content = msgParts.filter(p => p.type === 'text').map(p => p.content).join('\n\n') || '(无回复)';
1021
+ state.messages[msgIdx].content = _assembleV2Content(state.messages[msgIdx], msgParts);
985
1022
  state.messages[msgIdx]._streamingText = '';
986
1023
  if (allExecEvents.length > 0) state.messages[msgIdx].exec_events = [...allExecEvents];
987
1024
  }
@@ -999,7 +1036,7 @@ async function sendMessage() {
999
1036
  flushCurrentText();
1000
1037
  state.messages[msgIdx].parts = [...msgParts];
1001
1038
  state.messages[msgIdx]._streamingText = '';
1002
- state.messages[msgIdx].content = msgParts.filter(p => p.type === 'text').map(p => p.content).join('\n\n') || '';
1039
+ state.messages[msgIdx].content = _assembleV2Content(state.messages[msgIdx], msgParts);
1003
1040
  throttledStreamUpdate(msgIdx);
1004
1041
  // 停止当前轮次的 TTS 播放,防止旧迭代语音与新迭代语音互相打断
1005
1042
  if (ttsManager.enabled && ttsManager._streamActive) {
@@ -1040,7 +1077,10 @@ async function sendMessage() {
1040
1077
  }
1041
1078
  }
1042
1079
  }
1043
- // V2 Structured Output Events
1080
+ // V2 Structured Output Events — mark V2 mode
1081
+ } else if (evt.type === 'v2_context') {
1082
+ // V2 context event — confirms we're in V2 mode
1083
+ _isV2Mode = true;
1044
1084
  } else if (evt.type === 'v2_output_parsed') {
1045
1085
  // LLM output was parsed into structured format
1046
1086
  // evt.data contains: {usersays_correct, task_plan, tools_to_call, remember, recall, ask_user, finish}
@@ -1119,11 +1159,22 @@ async function sendMessage() {
1119
1159
  state.messages[msgIdx]._memorySaved = (state.messages[msgIdx]._memorySaved || '') +
1120
1160
  (evt.content ? evt.content.substring(0, 100) : '');
1121
1161
  } else if (evt.type === 'v2_reasoning') {
1122
- // V2 reasoning text from model (similar to thought_delta)
1162
+ // V2 reasoning text from model this IS the user-facing content in V2 mode
1123
1163
  if (!state.messages[msgIdx]._v2Reasoning) {
1124
1164
  state.messages[msgIdx]._v2Reasoning = '';
1125
1165
  }
1166
+ // Add separator between multiple reasoning events (e.g., beforecalltext + final text)
1167
+ if (state.messages[msgIdx]._v2Reasoning.length > 0 && evt.content) {
1168
+ state.messages[msgIdx]._v2Reasoning += '\n\n';
1169
+ }
1126
1170
  state.messages[msgIdx]._v2Reasoning += evt.content;
1171
+ _v2ReasoningText = state.messages[msgIdx]._v2Reasoning;
1172
+ // In V2 mode, update streaming text to show reasoning content
1173
+ if (_isV2Mode) {
1174
+ state.messages[msgIdx]._streamingText = _v2ReasoningText;
1175
+ // Also update content for real-time display
1176
+ state.messages[msgIdx].content = _v2ReasoningText;
1177
+ }
1127
1178
  throttledStreamUpdate(msgIdx);
1128
1179
  } else if (evt.type === 'done') {
1129
1180
  flushCurrentText();
@@ -1134,7 +1185,8 @@ async function sendMessage() {
1134
1185
  state.messages[msgIdx].parts = [...msgParts];
1135
1186
  state.messages[msgIdx]._streamingText = '';
1136
1187
  state.messages[msgIdx].exec_events = [...allExecEvents];
1137
- state.messages[msgIdx].content = msgParts.filter(p => p.type === 'text').map(p => p.content).join('\n\n') || '(无回复)';
1188
+ // Assemble final content: prefer V2 reasoning/ask text over raw XML
1189
+ state.messages[msgIdx].content = _assembleV2Content(state.messages[msgIdx], msgParts);
1138
1190
  } else if (evt.type === 'reasoning_delta') {
1139
1191
  // 模型推理过程增量文本(OpenAI o1/o3/DeepSeek-R1 等推理模型)
1140
1192
  if (!state.messages[msgIdx].reasoning) state.messages[msgIdx].reasoning = '';
@@ -1163,7 +1215,8 @@ async function sendMessage() {
1163
1215
  state.messages[msgIdx].parts = [...msgParts];
1164
1216
  state.messages[msgIdx]._streamingText = '';
1165
1217
  state.messages[msgIdx].exec_events = allExecEvents;
1166
- state.messages[msgIdx].content = msgParts.filter(p => p.type === 'text').map(p => p.content).join('\n\n') || '(无回复)';
1218
+ // Assemble final content: prefer V2 reasoning/ask text over raw XML
1219
+ state.messages[msgIdx].content = _assembleV2Content(state.messages[msgIdx], msgParts);
1167
1220
  }
1168
1221
 
1169
1222
  // Task list 已通过 SSE task_list_update 事件实时推送,无需再轮询
@@ -1230,6 +1283,19 @@ function stopGenerating() {
1230
1283
  if (state.abortController) {
1231
1284
  state.abortController.abort();
1232
1285
  }
1286
+ // 立即重置 UI 状态(防止 abort 后 finally 未执行导致 UI 卡死)
1287
+ state.isGenerating = false;
1288
+ state.abortController = null;
1289
+ hideTypingIndicator();
1290
+ stopExecTimerPolling();
1291
+ var sendBtn = document.getElementById('sendBtn');
1292
+ var stopBtn = document.getElementById('stopBtn');
1293
+ var input = document.getElementById('userInput');
1294
+ if (sendBtn) { sendBtn.style.display = ''; sendBtn.disabled = !(input && input.value.trim()); }
1295
+ if (stopBtn) stopBtn.style.display = 'none';
1296
+ // 关闭可能打开的 injectModal
1297
+ var injectModal = document.getElementById('injectModal');
1298
+ if (injectModal) injectModal.style.display = 'none';
1233
1299
  }
1234
1300
 
1235
1301
  // ══════════════════════════════════════════════════════
@@ -1237,7 +1303,8 @@ function stopGenerating() {
1237
1303
  // ══════════════════════════════════════════════════════
1238
1304
 
1239
1305
  function closeInjectModal() {
1240
- document.getElementById('injectModal').style.display = 'none';
1306
+ var modal = document.getElementById('injectModal');
1307
+ if (modal) modal.style.display = 'none';
1241
1308
  state.tempInputText = '';
1242
1309
  }
1243
1310
 
@@ -1438,7 +1505,8 @@ function handleKeyDown(e) {
1438
1505
  if (state.isGenerating) {
1439
1506
  // 任务执行中,使用统一的 injectModal 弹出选择对话框
1440
1507
  state.tempInputText = text;
1441
- document.getElementById('injectModal').style.display = 'flex';
1508
+ var modal = document.getElementById('injectModal');
1509
+ if (modal) { modal.style.display = 'flex'; }
1442
1510
  } else {
1443
1511
  sendMessage();
1444
1512
  }
@@ -142,4 +142,17 @@
142
142
  <div class="input-hint">MyAgent 可能会出错,请核实重要信息</div>
143
143
  </div>
144
144
  </div>
145
+
146
+ <!-- Inject Modal (生成中消息注入选择) -->
147
+ <div id="injectModal" class="modal-overlay" style="display:none">
148
+ <div class="modal">
149
+ <h3>任务执行中</h3>
150
+ <p>Agent 正在执行任务,请选择如何处理你的消息:</p>
151
+ <div style="display:flex;flex-direction:column;gap:8px;margin-bottom:0">
152
+ <button class="modal-btn" style="width:100%;justify-content:center;background:var(--accent);color:#fff" onclick="handleInjectChoice('queue')">排队等待</button>
153
+ <button class="modal-btn" style="width:100%;justify-content:center" onclick="handleInjectChoice('inject')">注入当前任务</button>
154
+ <button class="modal-btn" style="width:100%;justify-content:center;color:var(--text2)" onclick="closeInjectModal()">取消</button>
155
+ </div>
156
+ </div>
157
+ </div>
145
158
  </div>