docs: 更新协议文档，删除废弃计划文档
- 更新 http-error-codes, user-points-chat-data-protocol - 更新 divination-run-protocol, profile-protocol - 删除废弃的后端和前端设计计划文档
This commit is contained in:
qzl
@@ -0,0 +1,89 @@
|
||||
# Bug: 追问1次后无法进入查看历史记录
|
||||
|
||||
日期：2026-04-08
|
||||
状态：已确认（未修复）
|
||||
|
||||
## 问题描述
|
||||
|
||||
追问1次之后追问入口关闭，但用户也没法点击进去查看追问的历史记录。
|
||||
|
||||
## 根因分析
|
||||
|
||||
`_loadFollowUpEligibility()` 方法（`divination_result_screen.dart:65-83`）将两个概念混为一谈：
|
||||
|
||||
```dart
|
||||
Future<void> _loadFollowUpEligibility() async {
|
||||
...
|
||||
final messages = await widget.divinationApi!.getSessionMessages(
|
||||
threadId: widget.data.threadId!,
|
||||
);
|
||||
final userCount = messages.where((msg) => msg.role == 'user').length;
|
||||
...
|
||||
setState(() {
|
||||
_canFollowUp = userCount < 2; // 同时控制"能追问"和"能进入"
|
||||
...
|
||||
});
|
||||
}
|
||||
```
|
||||
|
||||
按钮逻辑（`divination_result_screen.dart:367`）：
|
||||
```dart
|
||||
onPressed: (!_canFollowUp || _followUpEligibilityLoading)
|
||||
? null // _canFollowUp = false 时按钮完全禁用
|
||||
: () {
|
||||
Navigator.of(context).push(
|
||||
MaterialPageRoute<void>(
|
||||
builder: (_) => FollowUpChatScreen(...),
|
||||
),
|
||||
);
|
||||
},
|
||||
```
|
||||
|
||||
**问题**：
|
||||
- `_canFollowUp = userCount < 2` 控制的是"是否还有追问配额"
|
||||
- 但 `onPressed` 把它和"是否能进入查看历史"混用了
|
||||
- 导致用户追问1次后（userCount=2），`_canFollowUp=false`，按钮被禁用
|
||||
- 用户**无法进入追问页面查看历史记录**
|
||||
|
||||
**业务逻辑分析**：
|
||||
- 每 session 最多2次追问（首问1次 + 追问1次）
|
||||
- 追问1次后，用户不能再发送新追问
|
||||
- 但**应该仍能进入查看历史记录**
|
||||
|
||||
### 正确的逻辑应该是
|
||||
|
||||
```dart
|
||||
// 追问配额判断
|
||||
_canFollowUp = userCount < 2;
|
||||
|
||||
// 能否进入查看历史（与追问次数无关，只要有历史消息就能进入）
|
||||
_canEnterFollowUpChat = messages.isNotEmpty; // 或者只要有 threadId 就能进
|
||||
|
||||
// 按钮文案
|
||||
buttonText = _canFollowUp ? l10n.followUpEntryHint : l10n.followUpViewHistory;
|
||||
|
||||
// 按钮是否禁用
|
||||
onPressed = (_followUpEligibilityLoading)
|
||||
? null // 只在加载中时禁用
|
||||
: () { ... } // 始终可点击进入查看
|
||||
```
|
||||
|
||||
## 证据
|
||||
|
||||
- 结果页代码：`apps/lib/features/divination/presentation/screens/divination_result_screen.dart`
|
||||
- `_buildFollowUpBar`: 行 338-394
|
||||
- `_loadFollowUpEligibility`: 行 65-83
|
||||
- 按钮 onPress: 行 367
|
||||
- 追问聊天页：`apps/lib/features/divination/presentation/screens/follow_up_chat_screen.dart`
|
||||
|
||||
## 修复方向
|
||||
|
||||
1. 分离"能否追问"和"能否进入查看历史"的逻辑
|
||||
2. 按钮文案根据状态显示不同内容（能追问显示追问提示，已用完显示"查看历史"）
|
||||
3. 只要有 `threadId`，用户就应该能进入查看历史
|
||||
|
||||
---
|
||||
|
||||
## 相关文档
|
||||
|
||||
- 随访工程计划：`docs/plans/2026-04-08-followup-session-history-eng-plan.md`
|
||||
@@ -0,0 +1,261 @@
|
||||
# Bug: 追问页面 UI 问题
|
||||
|
||||
日期：2026-04-08
|
||||
状态：已确认（未修复）
|
||||
|
||||
## Bug 1: 追问页发送时顶部和消息下方重复加载UI
|
||||
|
||||
### 问题描述
|
||||
|
||||
在追问页面，用户输入信息发送后，顶部和消息下方同时出现加载UI，但顶部的加载UI是多余的。
|
||||
|
||||
### 根因分析
|
||||
|
||||
**文件**：`apps/lib/features/divination/presentation/screens/follow_up_chat_screen.dart`
|
||||
|
||||
**问题代码**：
|
||||
|
||||
1. 顶部 step 指示器（第 85-124 行）：
|
||||
```dart
|
||||
if (_sending && _currentStepName != null)
|
||||
Container(
|
||||
// 显示 step 进度，如 "解读中..."、"推理中..."
|
||||
child: Row(
|
||||
children: [
|
||||
CircularProgressIndicator(...),
|
||||
Text(_stepLabel(_currentStepName!)),
|
||||
],
|
||||
),
|
||||
)
|
||||
```
|
||||
|
||||
2. 消息下方 streaming placeholder（第 565-584 行）：
|
||||
```dart
|
||||
isStreamingPlaceholder
|
||||
? Row(
|
||||
mainAxisSize: MainAxisSize.min,
|
||||
children: [
|
||||
CircularProgressIndicator(...),
|
||||
Text(l10n.followUpGenerating), // "生成中..."
|
||||
],
|
||||
)
|
||||
```
|
||||
|
||||
**事件触发流程**：
|
||||
|
||||
| 顺序 | 事件 | 顶部指示器 | 消息下方 |
|
||||
|------|------|----------|---------|
|
||||
| 1 | `_submitText()` 调用 | 无 | streaming placeholder 显示 "生成中..." |
|
||||
| 2 | `STEP_STARTED(stepName='divination')` | 显示 "解读中..." | 继续显示 |
|
||||
| 3 | `STEP_FINISHED` | 消失 | 继续显示 |
|
||||
| 4 | `STEP_STARTED(stepName='worker')` | 显示 "推理中..." | **同时显示** "生成中..." |
|
||||
| 5 | `STEP_FINISHED` | 消失 | 继续显示 |
|
||||
| 6 | `TEXT_MESSAGE_END` | 无 | placeholder 消失，显示真实内容 |
|
||||
|
||||
**问题**：在 worker 阶段（步骤 4），顶部显示 "推理中..." 同时消息下方显示 "生成中..."，造成重复反馈。用户只需要一个加载反馈即可。
|
||||
|
||||
**修复方向**：
|
||||
|
||||
1. **方案A**：只在 divination 阶段显示顶部 step 指示器，worker 阶段隐藏（因为 streaming placeholder 已经提供反馈）
|
||||
|
||||
```dart
|
||||
// 第 85 行修改
|
||||
if (_sending && _currentStepName != null && _currentStepName != 'worker')
|
||||
```
|
||||
|
||||
2. **方案B**：移除顶部 step 指示器，完全依赖 streaming placeholder
|
||||
|
||||
3. **方案C**：在 worker 阶段隐藏 streaming placeholder，只显示顶部指示器
|
||||
|
||||
推荐 **方案A**，因为：
|
||||
- divination 阶段没有 streaming placeholder，需要顶部指示器提供反馈
|
||||
- worker 阶段有 streaming placeholder，不需要顶部指示器
|
||||
- 改动最小，只加一个条件
|
||||
|
||||
---
|
||||
|
||||
## Bug 2: 语音模式录制时缺少动画UI
|
||||
|
||||
### 问题描述
|
||||
|
||||
根据 social-app 的实现，按住说话模式下录音时应该有动画效果，但 eryao 的实现只有文字，没有动画。
|
||||
|
||||
### 根因分析
|
||||
|
||||
**对比**：
|
||||
|
||||
| 功能 | social-app | eryao |
|
||||
|------|------------|-------|
|
||||
| 录音中动画 | `recordingAnimation` widget（外部传入） | 只有文字 "录音中..." |
|
||||
| 录音提示文字 | 带动画的 widget | 只有文字 |
|
||||
| 录音中隐藏输入区域 | `IgnorePointer` + `Opacity` | 未实现 |
|
||||
|
||||
**eryao 当前实现**（`message_composer.dart:159-172`）：
|
||||
```dart
|
||||
if (_isRecording) {
|
||||
if (!showRecordingInlineFeedback) {
|
||||
return Text(recordingText, ...);
|
||||
}
|
||||
return Column(
|
||||
mainAxisAlignment: MainAxisAlignment.center,
|
||||
children: [
|
||||
const SizedBox(width: 16, height: 16), // 空白占位
|
||||
const SizedBox(height: AppSpacing.xs),
|
||||
Text(recordingText, ...), // "录音中..."
|
||||
const SizedBox(height: AppSpacing.xs),
|
||||
Text(recordingHintText, ...), // "上滑取消"
|
||||
],
|
||||
);
|
||||
}
|
||||
```
|
||||
|
||||
**social-app 实现**（`message_composer.dart:222-239`）：
|
||||
```dart
|
||||
if (_isRecording) {
|
||||
if (!showRecordingInlineFeedback) {
|
||||
return Text(resolvedRecordingText, ...);
|
||||
}
|
||||
return Column(
|
||||
mainAxisAlignment: MainAxisAlignment.center,
|
||||
children: [
|
||||
recordingAnimation, // 动画widget（外部传入）
|
||||
const SizedBox(height: AppSpacing.xs),
|
||||
Text(resolvedRecordingText, ...),
|
||||
const SizedBox(height: AppSpacing.xs),
|
||||
Text(resolvedRecordingHintText, ...),
|
||||
],
|
||||
);
|
||||
}
|
||||
```
|
||||
|
||||
**差异**：
|
||||
1. `recordingAnimation` widget 缺失 - 需要外部传入，eryao 直接用空白占位
|
||||
2. 录音中图标/按钮没有动画效果
|
||||
|
||||
**修复方向**：
|
||||
|
||||
1. 在 `FollowUpChatScreen` 或其父组件中创建 `recordingAnimation` widget
|
||||
2. 将 `recordingAnimation` 通过 `MessageComposer` 传入
|
||||
3. 参考 social-app 的实现，使用脉冲动画或波形动画
|
||||
|
||||
**参考实现**（需要添加 `recording_animation.dart`）：
|
||||
|
||||
```dart
|
||||
class RecordingAnimation extends StatefulWidget {
|
||||
final double size;
|
||||
final Color color;
|
||||
|
||||
@override
|
||||
State<RecordingAnimation> createState() => _RecordingAnimationState();
|
||||
}
|
||||
|
||||
class _RecordingAnimationState extends State<RecordingAnimation>
|
||||
with SingleTickerProviderStateMixin {
|
||||
late AnimationController _controller;
|
||||
late Animation<double> _scaleAnimation;
|
||||
|
||||
@override
|
||||
void initState() {
|
||||
super.initState();
|
||||
_controller = AnimationController(
|
||||
duration: const Duration(milliseconds: 1000),
|
||||
vsync: this,
|
||||
)..repeat(reverse: true);
|
||||
_scaleAnimation = Tween<double>(begin: 1.0, end: 1.2).animate(
|
||||
CurvedAnimation(parent: _controller, curve: Curves.easeInOut),
|
||||
);
|
||||
}
|
||||
|
||||
@override
|
||||
void dispose() {
|
||||
_controller.dispose();
|
||||
super.dispose();
|
||||
}
|
||||
|
||||
@override
|
||||
Widget build(BuildContext context) {
|
||||
return AnimatedBuilder(
|
||||
animation: _scaleAnimation,
|
||||
builder: (context, child) {
|
||||
return Transform.scale(
|
||||
scale: _scaleAnimation.value,
|
||||
child: Container(
|
||||
width: widget.size,
|
||||
height: widget.size,
|
||||
decoration: BoxDecoration(
|
||||
shape: BoxShape.circle,
|
||||
color: widget.color.withValues(alpha: 0.3),
|
||||
),
|
||||
child: Icon(Icons.mic, color: widget.color, size: widget.size * 0.6),
|
||||
),
|
||||
);
|
||||
},
|
||||
);
|
||||
}
|
||||
}
|
||||
```
|
||||
|
||||
---
|
||||
|
||||
## Bug 2 & 3: 直接复用 social-app 输入框（不需要 plus 按钮）
|
||||
|
||||
### 需求
|
||||
|
||||
**直接复用** social-app 的 `MessageComposer`，但**不需要 plus 按钮**。
|
||||
|
||||
### 复用清单
|
||||
|
||||
| 功能 | social-app | eryao | 处理 |
|
||||
|------|------------|-------|------|
|
||||
| `recordingAnimation` | required Widget | 空白占位 | **需要添加** |
|
||||
| 双层阴影 | 有 | 无 | **需要添加** |
|
||||
| `AppRadius.xxl` | 用于圆角 | 无此值 | **需要添加** |
|
||||
| Plus 按钮 | 有 | 无 | **不需要**，保持无 |
|
||||
| 图标 | `LucideIcons` | Material Icons | 保持 Material Icons |
|
||||
|
||||
### 实现步骤
|
||||
|
||||
#### 1. 添加 `AppRadius.xxl` 到 `design_tokens.dart`
|
||||
|
||||
```dart
|
||||
class AppRadius {
|
||||
static const double sm = 8;
|
||||
static const double md = 12;
|
||||
static const double lg = 16;
|
||||
static const double xl = 20;
|
||||
static const double xxl = 32; // 新增
|
||||
static const double full = 999;
|
||||
}
|
||||
```
|
||||
|
||||
#### 2. 创建 `RecordingAnimation` widget
|
||||
|
||||
放在 `apps/lib/shared/widgets/` 下，参考 social-app 的脉冲动画效果。
|
||||
|
||||
#### 3. 修改 `MessageComposer`
|
||||
|
||||
- 添加 `recordingAnimation` 参数（required）
|
||||
- 添加双层阴影
|
||||
- 使用 `AppRadius.xxl` 替代 `AppRadius.full`
|
||||
- 移除 plus 按钮（eryao 原有的无 plus 逻辑保持不变）
|
||||
- 保持 Material Icons
|
||||
|
||||
#### 4. 修改 `FollowUpChatScreen`
|
||||
|
||||
- 创建 `RecordingAnimation` 实例
|
||||
- 传入 `MessageComposer`
|
||||
|
||||
---
|
||||
|
||||
## 相关文件
|
||||
|
||||
- `apps/lib/features/divination/presentation/screens/follow_up_chat_screen.dart`
|
||||
- `apps/lib/shared/widgets/message_composer.dart`
|
||||
- `apps/lib/shared/widgets/recording_animation.dart`（新建）
|
||||
- `apps/lib/shared/theme/design_tokens.dart`
|
||||
- `/home/qzl/Code/social-app/apps/lib/shared/widgets/message_composer.dart`
|
||||
|
||||
## 修复优先级
|
||||
|
||||
1. **高优先级**：Bug 1（重复加载UI）
|
||||
2. **高优先级**：Bug 2 & 3（直接复用 social-app 输入框）
|
||||
@@ -0,0 +1,200 @@
|
||||
# Bug: 追问时 agent_output 被重新生成，导致 sign_level 被覆盖
|
||||
|
||||
日期：2026-04-08
|
||||
状态：已确认根因（未修复）
|
||||
|
||||
## 问题描述
|
||||
|
||||
首次解卦完成后，用户继续追问时，agent 会重新生成 `agent_output`，重新计算卦的结论和标签。
|
||||
|
||||
**具体表现**：
|
||||
- 首次解卦结论：`中下签`
|
||||
- 追问后结论：`下下签`（同一个卦，结果被重新生成）
|
||||
|
||||
## 根因分析
|
||||
|
||||
**问题定位**：`backend/src/core/agentscope/runtime/runner.py`
|
||||
|
||||
### 根因链条
|
||||
|
||||
1. **`runner.py:execute()` 方法（第 70-133 行）**
|
||||
|
||||
无论 `runtime_mode` 是 `chat` 还是 `follow_up`，都会执行以下逻辑：
|
||||
|
||||
```python
|
||||
# 第 105-119 行：始终推导卦象
|
||||
derived_divination = self._resolve_derived_divination(run_input=run_input)
|
||||
await self._emit_step_event(
|
||||
pipeline=pipeline,
|
||||
run_input=run_input,
|
||||
step_name="divination",
|
||||
event_type="DIVINATION_DERIVED", # <-- 追问时不应发射此事件
|
||||
...
|
||||
)
|
||||
```
|
||||
|
||||
2. **`runner.py:_execute_worker_step()` 方法（第 200-245 行）**
|
||||
|
||||
始终将 `derived_divination` 传递给 worker：
|
||||
|
||||
```python
|
||||
# 第 294-302 行：始终将 divination_derived 放入 worker_output
|
||||
await emitter.emit_final_text_end(
|
||||
worker_output={
|
||||
**worker_payload.model_dump(mode="json", exclude_none=True),
|
||||
"divination_derived": derived_divination.model_dump(...), # <-- 追问时不应包含
|
||||
},
|
||||
...
|
||||
)
|
||||
```
|
||||
|
||||
3. **`stage_emitter.py:emit_final_text_end()` 方法（第 46-73 行）**
|
||||
|
||||
始终将所有字段放入 `TEXT_MESSAGE_END` 事件：
|
||||
|
||||
```python
|
||||
payload = {
|
||||
"messageId": message_id,
|
||||
"role": "assistant",
|
||||
"stage": self._stage,
|
||||
"status": worker_output.get("status"),
|
||||
"sign_level": worker_output.get("sign_level"), # <-- 追问时不应有
|
||||
"conclusion": worker_output.get("conclusion", []), # <-- 追问时不应有
|
||||
"focus_points": worker_output.get("focus_points", []),
|
||||
"advice": worker_output.get("advice", []),
|
||||
"keywords": worker_output.get("keywords", []),
|
||||
"answer": worker_output.get("answer", ""),
|
||||
"error": worker_output.get("error"),
|
||||
"divination_derived": worker_output.get("divination_derived"), # <-- 追问时不应有
|
||||
...
|
||||
}
|
||||
```
|
||||
|
||||
4. **`store.py:_persist_text_message()` 方法（第 125-160 行）**
|
||||
|
||||
从事件中提取所有字段并完整存储：
|
||||
|
||||
```python
|
||||
worker_output_fields = (
|
||||
"status",
|
||||
"sign_level", # <-- 追问时被重新生成并覆盖
|
||||
"conclusion", # <-- 追问时被重新生成并覆盖
|
||||
"focus_points",
|
||||
"advice",
|
||||
"keywords",
|
||||
"answer",
|
||||
"error",
|
||||
"divination_derived", # <-- 追问时被重新生成
|
||||
"ui_hints",
|
||||
)
|
||||
```
|
||||
|
||||
### 问题本质
|
||||
|
||||
`runtime_mode=follow_up` 时，系统仍在：
|
||||
1. 重新推导卦象（调用 `_resolve_derived_divination`）
|
||||
2. 发射完整的 `DIVINATION_DERIVED` 事件
|
||||
3. 生成包含所有结构化字段的 `worker_output`
|
||||
4. 将所有字段存储到数据库
|
||||
|
||||
但根据工程计划（`docs/plans/2026-04-08-followup-session-history-eng-plan.md:37-40`），追问时的预期行为是：
|
||||
|
||||
```
|
||||
[一次追问]
|
||||
user -> /agent/runs(runtime_mode=follow_up)
|
||||
-> assistant(content [+ optional metadata.agent_output.answer])
|
||||
```
|
||||
|
||||
即：追问时只输出 `answer`（内容），不重新生成卦象结构。
|
||||
|
||||
## 证据
|
||||
|
||||
### 数据库证据
|
||||
|
||||
Session `015fe0f9-0500-43ab-911a-4ce8e3160032`：
|
||||
|
||||
| 时间 | 角色 | sign_level | divination_derived | 问题 |
|
||||
|------|------|------------|-------------------|------|
|
||||
| 05:21:19 | user | - | - | 首问问题 |
|
||||
| 05:21:28 | assistant | 中下签 | 完整 | 首答（正确） |
|
||||
| 05:22:24 | user | - | - | 追问 |
|
||||
| 05:22:33 | assistant | **下下签** | 完整 | 追问答（sign_level 被重新生成） |
|
||||
|
||||
两个 assistant 消息的 `divination_derived` 完全相同，但 `sign_level` 不同，证明是重新生成的。
|
||||
|
||||
## 修复方向
|
||||
|
||||
### 1. `runner.py`
|
||||
|
||||
在 `execute()` 方法中，根据 `runtime_mode` 决定是否推导卦象：
|
||||
|
||||
```python
|
||||
async def execute(self, ...):
|
||||
runtime_mode = self._resolve_runtime_mode(run_input=run_input)
|
||||
|
||||
if runtime_mode == RuntimeMode.CHAT:
|
||||
derived_divination = self._resolve_derived_divination(run_input=run_input)
|
||||
await self._emit_step_event(...) # DIVINATION_DERIVED
|
||||
else:
|
||||
derived_divination = None # follow_up 不推导
|
||||
```
|
||||
|
||||
### 2. `stage_emitter.py`
|
||||
|
||||
`emit_final_text_end()` 根据 `runtime_mode` 决定发送哪些字段：
|
||||
|
||||
```python
|
||||
async def emit_final_text_end(self, ..., runtime_mode: str):
|
||||
payload = {"messageId": ..., "role": "assistant", "stage": self._stage}
|
||||
|
||||
if runtime_mode == "chat":
|
||||
payload.update({
|
||||
"status": worker_output.get("status"),
|
||||
"sign_level": worker_output.get("sign_level"),
|
||||
"conclusion": worker_output.get("conclusion", []),
|
||||
"focus_points": worker_output.get("focus_points", []),
|
||||
"advice": worker_output.get("advice", []),
|
||||
"keywords": worker_output.get("keywords", []),
|
||||
"divination_derived": worker_output.get("divination_derived"),
|
||||
...
|
||||
})
|
||||
else: # follow_up
|
||||
payload["answer"] = worker_output.get("answer", "")
|
||||
```
|
||||
|
||||
### 3. `store.py`
|
||||
|
||||
`_persist_text_message()` 根据 `runtime_mode` 决定提取哪些字段：
|
||||
|
||||
```python
|
||||
async def _persist_text_message(self, ...):
|
||||
runtime_mode = self._resolve_runtime_mode(event=event)
|
||||
|
||||
if runtime_mode == "chat":
|
||||
worker_output_fields = ("status", "sign_level", "conclusion", ...)
|
||||
else: # follow_up
|
||||
worker_output_fields = ("answer",) # 只存储 answer
|
||||
```
|
||||
|
||||
### 4. `runtime_models.py`
|
||||
|
||||
`resolve_worker_output_model()` 应根据 `runtime_mode` 返回不同 schema：
|
||||
|
||||
```python
|
||||
def resolve_worker_output_model(runtime_mode: RuntimeMode = RuntimeMode.CHAT) -> type[WorkerAgentOutputLite]:
|
||||
if runtime_mode == RuntimeMode.FOLLOW_UP:
|
||||
return WorkerAgentOutputLite # 只有 answer
|
||||
return AgentOutput # 完整结构（继承自 WorkerAgentOutputRich）
|
||||
```
|
||||
|
||||
## 相关文件
|
||||
|
||||
- `backend/src/core/agentscope/runtime/runner.py` - 主编排逻辑
|
||||
- `backend/src/core/agentscope/runtime/stage_emitter.py` - 事件发射
|
||||
- `backend/src/core/agentscope/events/store.py` - 事件持久化
|
||||
- `backend/src/schemas/agent/runtime_models.py` - 输出 schema 定义
|
||||
|
||||
## 相关文档
|
||||
|
||||
- 工程计划：`docs/plans/2026-04-08-followup-session-history-eng-plan.md`
|
||||
- 协议文档：`docs/protocols/divination/divination-run-protocol.md`
|
||||
@@ -0,0 +1,90 @@
|
||||
# Bug: 首页历史 Session 只显示4条，"更多"按钮缺失
|
||||
|
||||
日期：2026-04-08
|
||||
状态：功能缺失（未完成）
|
||||
|
||||
## 问题描述
|
||||
|
||||
前端主页接收后端传来的历史 session，永远只显示四个，"more"按钮不见了。
|
||||
|
||||
## 根因分析
|
||||
|
||||
### 1. 前端硬编码 `take(4)`
|
||||
|
||||
`apps/lib/features/home/presentation/screens/home_screen.dart:287`：
|
||||
```dart
|
||||
children: historyItems.take(4).map((item) {
|
||||
```
|
||||
|
||||
这是在 commit `6e82053`（重构首页为底部导航栏布局）时**有意添加**的设计选择，用于限制首页展示的历史记录数量。
|
||||
|
||||
### 2. 前端未实现"更多"按钮
|
||||
|
||||
- `l10n.more` 本地化字符串存在（`'更多'`/`'More'`），定义于 `apps/lib/l10n/app_localizations_zh.dart:115` 和 `app_localizations_en.dart:116`
|
||||
- 但搜索整个 Dart 代码库，`l10n.more` **没有任何地方使用它**
|
||||
- "更多"功能从未被实现
|
||||
|
||||
### 3. 后端 `hasMore` 硬编码为 `False`
|
||||
|
||||
`backend/src/v1/agent/service.py:655-659`：
|
||||
```python
|
||||
return HistorySnapshotResponse(
|
||||
scope="history_sessions_latest_assistant",
|
||||
threadId=None,
|
||||
day=None,
|
||||
hasMore=False, # 硬编码，未实际计算
|
||||
messages=messages,
|
||||
)
|
||||
```
|
||||
|
||||
后端 schema 定义了 `hasMore` 字段（`schemas.py:238`），但 service 层返回时**硬编码为 `False`**，从未实际计算是否还有更多数据。
|
||||
|
||||
### 4. 后端 API 行为正确
|
||||
|
||||
`backend/src/v1/agent/service.py:641-646`：
|
||||
```python
|
||||
raw_messages = await self._repository.get_latest_assistant_messages_by_user_sessions(
|
||||
user_id=str(current_user.id),
|
||||
visibility_mask=bit_mask(bit=int(SystemVisibilityBit.UI_HISTORY)),
|
||||
session_limit=50, # 后端返回最多50条
|
||||
)
|
||||
```
|
||||
|
||||
后端正确返回最多 50 条历史 session，前端只是没有利用这些数据。
|
||||
|
||||
## 问题性质
|
||||
|
||||
| 层级 | 现象 | 性质 |
|
||||
|------|------|------|
|
||||
| 后端 | `hasMore=False` 硬编码 | 缺陷：响应语义不正确 |
|
||||
| 前端 | `take(4)` 只显示4条 | 设计选择：有意的 UI 限制 |
|
||||
| 前端 | 无"更多"按钮 | 功能缺失：有 `l10n.more` 但未实现 |
|
||||
|
||||
## 证据
|
||||
|
||||
- 前端代码：`apps/lib/features/home/presentation/screens/home_screen.dart:287`
|
||||
- 后端代码：`backend/src/v1/agent/service.py:655`
|
||||
- 本地化字符串：`apps/lib/l10n/app_localizations_zh.dart:115`
|
||||
- 提交记录：`6e82053`（feat(home): 重构首页为底部导航栏布局）
|
||||
|
||||
## 修复方向
|
||||
|
||||
### 方案 A：实现完整分页（推荐）
|
||||
|
||||
1. **后端**：实现真正的 `hasMore` 计算逻辑
|
||||
- 添加 `offset` 参数支持分页
|
||||
- 实际计算 `hasMore = total_count > offset + limit`
|
||||
|
||||
2. **前端**：实现"更多"按钮或无限滚动
|
||||
- 监听滚动位置，滚动到底部时加载更多
|
||||
- 或添加"更多"按钮手动触发加载
|
||||
|
||||
### 方案 B：移除不存在的字符串
|
||||
|
||||
如果业务确定只需要显示4条，则：
|
||||
- 移除 `l10n.more` 本地化字符串，避免误导
|
||||
|
||||
## 相关文档
|
||||
|
||||
- 工程计划：`docs/plans/2026-04-05-divination-history-profile-eng-plan.md`
|
||||
- 随访计划：`docs/plans/2026-04-08-followup-session-history-eng-plan.md`
|
||||
Reference in New Issue
Block a user