[pull] master from AstrBotDevs:master

astrbot/builtin_stars/builtin_commands/commands/conversation.py

@@ -1,3 +1,6 @@
+from sqlalchemy import case, func, select
+from sqlmodel import col
+
 from astrbot.api import sp, star
 from astrbot.api.event import AstrMessageEvent, MessageEventResult
 from astrbot.core import logger
@@ -7,6 +10,7 @@
     DEERFLOW_THREAD_ID_KEY,
 )
 from astrbot.core.agent.runners.deerflow.deerflow_api_client import DeerFlowAPIClient
+from astrbot.core.db.po import ProviderStat
 from astrbot.core.utils.active_event_registry import active_event_registry
 
 from .utils.rst_scene import RstScene
@@ -246,3 +250,62 @@ async def new_conv(self, message: AstrMessageEvent) -> None:
                 f"✅ Switched to new conversation: {cid[:4]}。"
             ),
         )
+
+    async def stats(self, message: AstrMessageEvent) -> None:
+        """Show token usage statistics for the current conversation."""
+        umo = message.unified_msg_origin
+        cid = await self.context.conversation_manager.get_curr_conversation_id(umo)
+
+        if not cid:
+            message.set_result(
+                MessageEventResult().message(
+                    "❌ You are not in a conversation. Use /new to create one."
+                ),
+            )
+            return
+
+        db = self.context.get_db()
+        async with db.get_db() as session:
+            result = await session.execute(
+                select(
+                    func.count(case((col(ProviderStat.id).is_not(None), 1))).label(
+                        "record_count",
+                    ),
+                    func.coalesce(func.sum(ProviderStat.token_input_other), 0).label(
+                        "total_input_other",
+                    ),
+                    func.coalesce(func.sum(ProviderStat.token_input_cached), 0).label(
+                        "total_input_cached",
+                    ),
+                    func.coalesce(func.sum(ProviderStat.token_output), 0).label(
+                        "total_output",
+                    ),
+                ).where(
+                    col(ProviderStat.agent_type) == "internal",
+                    col(ProviderStat.conversation_id) == cid,
+                )
+            )
+            stats = result.one()
+
+        if stats.record_count == 0:
+            message.set_result(
+                MessageEventResult().message(
+                    "📊 No stats available for this conversation yet."
+                ),
+            )
+            return
+
+        total_input_other = stats.total_input_other
+        total_input_cached = stats.total_input_cached
+        total_output = stats.total_output
+        total_tokens = total_input_other + total_input_cached + total_output
+
+        ret = (
+            f"📊 Conversation Token usage (ID: {cid[:8]}...)\n"
+            f"Total:          {total_tokens:,}\n"
+            f"Input (cached): {total_input_cached:,}\n"
+            f"Input (other):  {total_input_other:,}\n"
+            f"Output:         {total_output:,}\n"
+        )
+
+        message.set_result(MessageEventResult().message(ret))

astrbot/builtin_stars/builtin_commands/main.py

@@ -47,6 +47,11 @@ async def new_conv(self, message: AstrMessageEvent) -> None:
         """Create new conversation"""
         await self.conversation_c.new_conv(message)
 
+    @filter.command("stats")
+    async def stats(self, message: AstrMessageEvent) -> None:
+        """Show token usage statistics for the current conversation"""
+        await self.conversation_c.stats(message)
+
     @filter.permission_type(filter.PermissionType.ADMIN)
     @filter.command("provider")
     async def provider(

astrbot/cli/__init__.py

@@ -1 +1 @@
-__version__ = "4.23.5"
+__version__ = "4.23.6"

astrbot/core/agent/runners/tool_loop_agent_runner.py

@@ -183,10 +183,10 @@ async def _complete_with_assistant_response(self, llm_resp: LLMResponse) -> None
         self.stats.end_time = time.time()
 
         parts = []
-        if llm_resp.reasoning_content or llm_resp.reasoning_signature:
+        if llm_resp.reasoning_content is not None or llm_resp.reasoning_signature:
             parts.append(
                 ThinkPart(
-                    think=llm_resp.reasoning_content,
+                    think=llm_resp.reasoning_content or "",
                     encrypted=llm_resp.reasoning_signature,
                 )
             )
@@ -876,10 +876,10 @@ async def step(self):
 
             # 将结果添加到上下文中
             parts = []
-            if llm_resp.reasoning_content or llm_resp.reasoning_signature:
+            if llm_resp.reasoning_content is not None or llm_resp.reasoning_signature:
                 parts.append(
                     ThinkPart(
-                        think=llm_resp.reasoning_content,
+                        think=llm_resp.reasoning_content or "",
                         encrypted=llm_resp.reasoning_signature,
                     )
                 )
@@ -1361,10 +1361,10 @@ async def _finalize_aborted_step(
         self.stats.end_time = time.time()
 
         parts = []
-        if llm_resp.reasoning_content or llm_resp.reasoning_signature:
+        if llm_resp.reasoning_content is not None or llm_resp.reasoning_signature:
             parts.append(
                 ThinkPart(
-                    think=llm_resp.reasoning_content,
+                    think=llm_resp.reasoning_content or "",
                     encrypted=llm_resp.reasoning_signature,
                 )
             )

astrbot/core/config/default.py

@@ -5,7 +5,7 @@
 
 from astrbot.core.utils.astrbot_path import get_astrbot_data_path
 
-VERSION = "4.23.5"
+VERSION = "4.23.6"
 DB_PATH = os.path.join(get_astrbot_data_path(), "data_v4.db")
 PERSONAL_WECHAT_CONFIG_METADATA = {
     "weixin_oc_base_url": {

astrbot/core/provider/entities.py

@@ -353,7 +353,7 @@ class LLMResponse:
     """Tool call IDs."""
     tools_call_extra_content: dict[str, dict[str, Any]] = field(default_factory=dict)
     """Tool call extra content. tool_call_id -> extra_content dict"""
-    reasoning_content: str = ""
+    reasoning_content: str | None = None
     """The reasoning content extracted from the LLM, if any."""
     reasoning_signature: str | None = None
     """The signature of the reasoning content, if any."""
@@ -404,8 +404,6 @@ def __init__(
             raw_completion (ChatCompletion, optional): 原始响应, OpenAI 格式. Defaults to None.
 
         """
-        if reasoning_content is None:
-            reasoning_content = ""
         if tools_call_args is None:
             tools_call_args = []
         if tools_call_name is None:

astrbot/core/provider/sources/anthropic_source.py

@@ -39,7 +39,7 @@ def _ensure_usable_response(
         stop_reason: str | None = None,
     ) -> None:
         has_text_output = bool((llm_response.completion_text or "").strip())
-        has_reasoning_output = bool(llm_response.reasoning_content.strip())
+        has_reasoning_output = bool((llm_response.reasoning_content or "").strip())
         has_tool_output = bool(llm_response.tools_call_args)
         if has_text_output or has_reasoning_output or has_tool_output:
             return

astrbot/core/provider/sources/gemini_source.py

@@ -462,7 +462,7 @@ def _ensure_usable_response(
         finish_reason: str | None = None,
     ) -> None:
         has_text_output = bool((llm_response.completion_text or "").strip())
-        has_reasoning_output = bool(llm_response.reasoning_content.strip())
+        has_reasoning_output = bool((llm_response.reasoning_content or "").strip())
         has_tool_output = bool(llm_response.tools_call_args)
         if has_text_output or has_reasoning_output or has_tool_output:
             return

astrbot/core/provider/sources/openai_source.py

@@ -671,9 +671,9 @@ async def _query_stream(
             reasoning = self._extract_reasoning_content(chunk)
             _y = False
             llm_response.id = chunk.id
-            llm_response.reasoning_content = ""
+            llm_response.reasoning_content = None
             llm_response.completion_text = ""
-            if reasoning:
+            if reasoning is not None:
                 llm_response.reasoning_content = reasoning
                 _y = True
             if delta and delta.content:
@@ -701,22 +701,28 @@ async def _query_stream(
     def _extract_reasoning_content(
         self,
         completion: ChatCompletion | ChatCompletionChunk,
-    ) -> str:
+    ) -> str | None:
         """Extract reasoning content from OpenAI ChatCompletion if available."""
-        reasoning_text = ""
+
+        def _get_reasoning_attr(obj: Any) -> str | None:
+            fields_set = getattr(obj, "model_fields_set", None)
+            if isinstance(fields_set, set) and self.reasoning_key in fields_set:
+                attr = getattr(obj, self.reasoning_key, "")
+                return "" if attr is None else str(attr)
+            attr = getattr(obj, self.reasoning_key, None)
+            return None if attr is None else str(attr)
+
         if not completion.choices:
-            return reasoning_text
+            return None
         if isinstance(completion, ChatCompletion):
             choice = completion.choices[0]
-            reasoning_attr = getattr(choice.message, self.reasoning_key, None)
-            if reasoning_attr:
-                reasoning_text = str(reasoning_attr)
+            reasoning_attr = _get_reasoning_attr(choice.message)
         elif isinstance(completion, ChatCompletionChunk):
             delta = completion.choices[0].delta
-            reasoning_attr = getattr(delta, self.reasoning_key, None)
-            if reasoning_attr:
-                reasoning_text = str(reasoning_attr)
-        return reasoning_text
+            reasoning_attr = _get_reasoning_attr(delta)
+        else:
+            return None
+        return reasoning_attr
 
     def _extract_usage(self, usage: CompletionUsage | dict) -> TokenUsage:
         ptd = getattr(usage, "prompt_tokens_details", None)
@@ -859,7 +865,9 @@ async def _parse_openai_completion(
 
         # parse the reasoning content if any
         # the priority is higher than the <think> tag extraction
-        llm_response.reasoning_content = self._extract_reasoning_content(completion)
+        reasoning_content = self._extract_reasoning_content(completion)
+        if reasoning_content is not None:
+            llm_response.reasoning_content = reasoning_content
 
         # parse tool calls if any
         if choice.message.tool_calls and tools is not None:
@@ -906,7 +914,7 @@ async def _parse_openai_completion(
                 "API 返回的 completion 由于内容安全过滤被拒绝(非 AstrBot)。",
             )
         has_text_output = bool((llm_response.completion_text or "").strip())
-        has_reasoning_output = bool(llm_response.reasoning_content.strip())
+        has_reasoning_output = bool((llm_response.reasoning_content or "").strip())
         if (
             not has_text_output
             and not has_reasoning_output
@@ -987,31 +995,34 @@ def _finally_convert_payload(self, payloads: dict) -> None:
             model in deepseek_reasoning_models
             or "api.deepseek.com" in self.client.base_url.host
         )
-
         for message in payloads.get("messages", []):
             if message.get("role") == "assistant" and isinstance(
                 message.get("content"), list
             ):
                 reasoning_content = ""
+                reasoning_content_present = False
                 new_content = []  # not including think part
                 for part in message["content"]:
                     if part.get("type") == "think":
+                        reasoning_content_present = True
                         reasoning_content += str(part.get("think"))
                     else:
                         new_content.append(part)
                 # Some providers (Grok, etc.) reject empty content lists.
                 # When all parts were think blocks, fall back to None.
                 message["content"] = new_content or None
-                if is_deepseek_v4_reasoning and not reasoning_content:
-                    logger.info(
-                        "Deepseek v4 model requires non-empty reasoning content, but got empty. Setting to 'none' to satisfy the requirement."
-                    )
-                    # Deepseek models require the field on assistant
-                    # history messages, even when the reasoning content is empty.
-                    message["reasoning_content"] = "none"
-                elif reasoning_content:
+                if reasoning_content_present:
                     message["reasoning_content"] = reasoning_content
 
+            if (
+                message.get("role") == "assistant"
+                and is_deepseek_v4_reasoning
+                and "reasoning_content" not in message
+            ):
+                # DeepSeek v4 reasoning models require the field on assistant
+                # history messages, even when the reasoning content is empty.
+                message["reasoning_content"] = ""
+
             # Gemini 的 function_response 要求 google.protobuf.Struct（即 JSON 对象），
             # 纯文本会触发 400 Invalid argument，需要包一层 JSON。
             if is_gemini and message.get("role") == "tool":

astrbot/dashboard/routes/stat.py

@@ -243,6 +243,7 @@ async def get_provider_token_stats(self):
             total_by_umo: dict[str, int] = defaultdict(int)
             total_by_bucket: dict[int, int] = defaultdict(int)
             range_total_tokens = 0
+            range_total_output_tokens = 0
             range_total_calls = 0
             range_success_calls = 0
             range_ttft_total_ms = 0.0
@@ -286,6 +287,7 @@ async def get_provider_token_stats(self):
                             record.end_time - record.start_time
                         ) * 1000
                         range_duration_samples += 1
+                        range_total_output_tokens += record.token_output
 
                 if created_at_local >= today_start_local:
                     today_total_calls += 1
@@ -371,7 +373,8 @@ async def get_provider_token_stats(self):
                             else 0
                         ),
                         "range_avg_tpm": (
-                            range_total_tokens / (range_duration_total_ms / 1000 / 60)
+                            range_total_output_tokens
+                            / (range_duration_total_ms / 1000 / 60)
                             if range_duration_total_ms > 0
                             else 0
                         ),

changelogs/v4.23.6.md

@@ -0,0 +1,56 @@
+- [更新日志(简体中文)](#chinese)
+- [Changelog(English)](#english)
+
+<a id="chinese"></a>
+
+## What's Changed
+
+### 新增
+
+- 新增 `/stats` 命令，可查看当前会话的 Token 使用统计，并按总量、输入（缓存）、输入（其他）与输出拆分展示。（[#7831](https://github.com/AstrBotDevs/AstrBot/pull/7831)）
+- 新增 Firecrawl Web 搜索与网页提取工具，支持搜索结果处理、网页内容提取、会话管理、请求校验与相关测试。（[#7764](https://github.com/AstrBotDevs/AstrBot/pull/7764)）
+- 微信客服文本消息新增 15 秒内去重，减少重复消息处理。（[#7788](https://github.com/AstrBotDevs/AstrBot/pull/7788)）
+
+### 优化
+
+- 优化 Provider 配置界面性能与响应式显示，改善相关组件的字体和布局体验。（[#7772](https://github.com/AstrBotDevs/AstrBot/pull/7772)）
+- 优化统计页 TPM 计算逻辑，TPM 现在仅统计输出 Token，并更新相关文案。（[#7827](https://github.com/AstrBotDevs/AstrBot/pull/7827)）
+- 优化 OpenAI 兼容 Provider 的空 assistant 消息过滤逻辑，流式与非流式路径统一处理空字符串和空列表内容，避免严格 Provider 拒绝历史消息。（[#7758](https://github.com/AstrBotDevs/AstrBot/pull/7758)）
+
+### 修复
+
+- 修复 DeepSeek v4 与 reasoning content 相关处理，支持空字符串 reasoning 内容，并在 assistant 消息中保留 reasoning 字段。（[#7823](https://github.com/AstrBotDevs/AstrBot/pull/7823), [#7830](https://github.com/AstrBotDevs/AstrBot/pull/7830)）
+- 修复 OpenRouter reasoning 字段属性名不正确的问题。（[#7821](https://github.com/AstrBotDevs/AstrBot/pull/7821)）
+- 修复超大图片未压缩可能导致后续处理异常的问题，并复用图片最大尺寸检查工具。（[#7807](https://github.com/AstrBotDevs/AstrBot/pull/7807)）
+- 修复 MiniMax TTS 默认输出 MP3 导致 QQ 官方平台语音转换出现 RIFF 错误的问题，默认输出格式改为 WAV。（[#7797](https://github.com/AstrBotDevs/AstrBot/pull/7797)）
+- 修复 Computer 沙盒图片下载未按图片发送的问题。（[#7785](https://github.com/AstrBotDevs/AstrBot/pull/7785)）
+- 修复 Windows 环境下部分 HTTPS 请求证书校验失败的问题，使用 certifi SSL context 提升兼容性。（[#7778](https://github.com/AstrBotDevs/AstrBot/pull/7778)）
+- 修复非安全上下文或部分对话框中复制功能不可用的问题，抽取共享剪贴板工具并增加 fallback。（[#7747](https://github.com/AstrBotDevs/AstrBot/pull/7747)）
+- 修复文件上传可能存在路径穿越的问题，并清理上传文件名中的 NUL 字节。（[#7751](https://github.com/AstrBotDevs/AstrBot/pull/7751)）
+
+<a id="english"></a>
+
+## What's Changed (EN)
+
+### New Features
+
+- Added a `/stats` command to show token usage for the current conversation, including total tokens, input cached tokens, input other tokens, and output tokens. ([#7831](https://github.com/AstrBotDevs/AstrBot/pull/7831))
+- Added Firecrawl web search and web extract tools with result handling, content extraction, session management, payload validation, and tests. ([#7764](https://github.com/AstrBotDevs/AstrBot/pull/7764))
+- Added 15-second deduplication for WeChat kefu text messages to reduce duplicate message handling. ([#7788](https://github.com/AstrBotDevs/AstrBot/pull/7788))
+
+### Improvements
+
+- Improved the Provider configuration UI performance and responsive layout, including font and component styling updates. ([#7772](https://github.com/AstrBotDevs/AstrBot/pull/7772))
+- Updated stats-page TPM calculation so TPM only counts output tokens, with matching label updates. ([#7827](https://github.com/AstrBotDevs/AstrBot/pull/7827))
+- Improved empty assistant message filtering for OpenAI-compatible providers by sharing the logic across streaming and non-streaming paths and handling empty string or empty list content. ([#7758](https://github.com/AstrBotDevs/AstrBot/pull/7758))
+
+### Bug Fixes
+
+- Fixed DeepSeek v4 and reasoning content handling by supporting empty-string reasoning content and preserving the reasoning field in assistant messages. ([#7823](https://github.com/AstrBotDevs/AstrBot/pull/7823), [#7830](https://github.com/AstrBotDevs/AstrBot/pull/7830))
+- Fixed the reasoning field attribute used for OpenRouter. ([#7821](https://github.com/AstrBotDevs/AstrBot/pull/7821))
+- Fixed oversized image handling by downscaling large images and sharing the image max-size check helper. ([#7807](https://github.com/AstrBotDevs/AstrBot/pull/7807))
+- Fixed MiniMax TTS output for QQ Official voice conversion by changing the default output format from MP3 to WAV. ([#7797](https://github.com/AstrBotDevs/AstrBot/pull/7797))
+- Fixed Computer sandbox image downloads so they are sent as images. ([#7785](https://github.com/AstrBotDevs/AstrBot/pull/7785))
+- Fixed HTTPS certificate verification issues on Windows by using a certifi SSL context. ([#7778](https://github.com/AstrBotDevs/AstrBot/pull/7778))
+- Fixed copy actions in insecure contexts and dialogs by extracting a shared clipboard utility with fallback behavior. ([#7747](https://github.com/AstrBotDevs/AstrBot/pull/7747))
+- Fixed path traversal risks in file uploads and removed embedded NUL bytes from upload filenames. ([#7751](https://github.com/AstrBotDevs/AstrBot/pull/7751))