myagent-ai 1.17.3 → 1.18.1

package/agents/main_agent.py

@@ -706,46 +706,35 @@ class MainAgent(BaseAgent):
                         )
                     break
 
-                # [v1.16.13→17.0] 特殊处理模型不支持图片输入 — 去掉图片用纯文本重试
-                # 支持中英文错误关键词匹配（ModelScope 等国产 API 可能返回中文错误）
+                # [v1.18.0] 特殊处理模型不支持图片输入
+                # 策略: 保留图片数据，break 后让 _stream_process_message 返回 ⚠️ 标记
+                # model chain 的 _try_model_chain_stream_inner 检测到 ⚠️ 后会继续尝试下一个模型
                 _vision_keywords = [
                     "doesn't support image", "does not support image", "model_incompatible",
                     "image input", "not support vision", "unsupported multimodal", "image capability",
                     "不支持图片", "不支持图像", "图片输入", "图像输入", "不支持多模态",
                     "视觉", "image_url", "multimodal", "vision",
                 ]
-                if any(kw.lower() in _llm_error.lower() for kw in _vision_keywords) and context.metadata.get("user_images"):
-                    logger.warning(f"[{task_id}] 模型不支持图片输入，去掉图片用纯文本重试")
-                    context.metadata["user_images"] = []
-                    # 用纯文本消息替换最后一条多模态消息
-                    _text_only_msg = context.user_message or "请处理上述上下文。"
-                    if len(messages) > 0 and isinstance(messages[-1].content, list):
-                        messages[-1] = Message(role="user", content=_text_only_msg)
-                    # 重试 LLM 调用
-                    if stream_response and self.llm:
-                        response = await self._call_llm_stream(
-                            messages, text_delta_callback=text_delta_callback,
-                            stream_response=stream_response,
+                _is_vision_error = (
+                    any(kw.lower() in _llm_error.lower() for kw in _vision_keywords)
+                    and context.metadata.get("user_images")
+                )
+
+                if _is_vision_error:
+                    logger.warning(f"[{task_id}] 模型 {self.llm.model} 不支持图片输入，"
+                                   f"将切换到支持图片的模型重试")
+                    # 返回 ⚠️ 标记的错误，让 model chain 继续尝试下一个模型
+                    # 注意: 不清除 context.metadata["user_images"]，下一个模型仍可使用
+                    _vision_skip_msg = f"⚠️ 模型 {self.llm.model} 不支持图片，正在切换..."
+                    context.working_memory["final_response"] = _vision_skip_msg
+                    await self._emit_v2_event("v2_reasoning", {"content": _vision_skip_msg}, stream_callback)
+                    if self.memory:
+                        self.memory.add_session(
+                            session_id=context.session_id,
+                            role="assistant",
+                            content=_vision_skip_msg,
                         )
-                    else:
-                        response = await self._call_llm(messages)
-                    if response.success:
-                        # 纯文本重试成功，给回复加上提示前缀
-                        _vision_prefix = "⚠️ 当前模型不支持图片识别，已自动使用纯文本模式处理（图片未发送给模型）。\n\n"
-                        llm_raw = _vision_prefix + response.content
-                        context.working_memory["final_response"] = llm_raw
-                        await self._emit_v2_event("v2_reasoning", {"content": llm_raw}, stream_callback)
-                        if self.memory:
-                            self.memory.add_session(
-                                session_id=context.session_id,
-                                role="assistant",
-                                content=llm_raw,
-                            )
-                        break
-                    else:
-                        # 纯文本也失败了，走下面的通用错误处理
-                        _llm_error = response.error or ""
-                        logger.error(f"[{task_id}] 纯文本重试也失败: {_llm_error}")
+                    break  # 退出 agent 循环，让 model chain 尝试下一个模型
 
                 # 其他 LLM 错误
                 error_msg = f"LLM 调用失败: {response.error}"

package/core/deps_checker.py

@@ -68,14 +68,20 @@ DEPENDENCIES: List[DepInfo] = [
     # ── PDF 处理 ──
     DepInfo("PyPDF2", "PyPDF2", "3.0.0", "pdf", "all",
             note="纯 Python PDF 文本提取（pdftotext 失败时的备用方案）"),
+    DepInfo("pypdf", "pypdf", "4.0.0", "pdf", "all",
+            note="PDF 文本提取 (PyPDF2 新版, 优先使用)"),
+    DepInfo("reportlab", "reportlab", "4.0.0", "pdf", "all",
+            note="PDF 文档生成引擎"),
+    DepInfo("PIL", "Pillow", "10.0.0", "pdf", "all",
+            note="图片处理 (PDF 插图, 报告图表)"),
 
     # ── 文档处理 (Excel/Word/PPT) ──
     DepInfo("openpyxl", "openpyxl", "3.1.0", "doc", "all",
-            note="Excel (.xlsx) 文件读取"),
+            note="Excel (.xlsx) 文件读取和生成"),
     DepInfo("docx", "python-docx", "1.1.0", "doc", "all",
-            note="Word (.docx) 文件读取"),
+            note="Word (.docx) 文件读取和生成"),
     DepInfo("pptx", "python-pptx", "0.6.21", "doc", "all",
-            note="PowerPoint (.pptx) 文件读取"),
+            note="PowerPoint (.pptx) 文件读取和生成"),
     DepInfo("xlrd", "xlrd", "2.0.0", "doc", "all",
             note="旧版 Excel (.xls) 文件读取"),
 

package/core/vnc_manager.py

@@ -597,6 +597,12 @@ class VNCManager:
 
             env = {**os.environ, "DISPLAY": self.display}
 
+            # [v1.18.0] proot/Termux 兼容: 可能需要额外的安全参数
+            cmd.append("-nobell")
+            cmd.append("-noxdamage")
+            # 跳过 Xinerama 检查（proot 环境下可能失败）
+            env["X11VNC_NO_UNIXPW"] = "1"
+
             logger.info(f"启动 x11vnc: {' '.join(cmd)}")
             self._x11vnc_process = subprocess.Popen(
                 cmd,
@@ -612,7 +618,7 @@ class VNCManager:
             if self._x11vnc_process.poll() is not None:
                 stderr = ""
                 try:
-                    stderr = self._x11vnc_process.stderr.read().decode("utf-8", errors="replace")[:500]
+                    stderr = self._x11vnc_process.stderr.read().decode("utf-8", errors="replace")[:2000]
                 except Exception:
                     pass
                 logger.error(f"x11vnc 启动失败: {stderr}")

package/main.py

@@ -390,6 +390,19 @@ class MyAgentApp:
         ]:
             self.skill_registry.register(skill_cls())
 
+        # ── 文档生成技能 (v1.17.3, 融合 MiniMax Skills 设计理念) ──
+        from skills.pdf_skill import PDFCreateSkill, PDFReadSkill
+        from skills.docx_skill import DOCXCreateSkill, DOCXReadSkill
+        from skills.xlsx_skill import XLSXCreateSkill, XLSXReadSkill, XLSXEditSkill
+        from skills.ppt_skill import PPTCreateSkill, PPTReadSkill
+        for skill_cls in [
+            PDFCreateSkill, PDFReadSkill,
+            DOCXCreateSkill, DOCXReadSkill,
+            XLSXCreateSkill, XLSXReadSkill, XLSXEditSkill,
+            PPTCreateSkill, PPTReadSkill,
+        ]:
+            self.skill_registry.register(skill_cls())
+
     async def process_message(
         self,
         user_message: str,

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "myagent-ai",
-  "version": "1.17.3",
+  "version": "1.18.1",
   "description": "本地桌面端执行型AI助手 - Open Interpreter 风格 | Local Desktop Execution-Oriented AI Assistant",
   "main": "main.py",
   "bin": {

package/skills/docx_skill.py

@@ -0,0 +1,199 @@
+"""
+skills/docx_skill.py - DOCX Word 文档生成/编辑技能
+===============================================
+基于 python-docx 生成和编辑 Word 文档，融入 MiniMax DOCX Skill 设计理念。
+支持标题、段落、列表、表格、图片、页眉页脚、样式等。
+"""
+from __future__ import annotations
+
+import json
+import os
+from pathlib import Path
+from typing import Any, Dict, List, Optional
+
+from skills.base import Skill, SkillParameter, SkillResult
+
+
+class DOCXCreateSkill(Skill):
+    """生成 Word 文档"""
+    name = "docx_create"
+    description = (
+        "生成 Word (DOCX) 文档。支持多级标题、段落、列表、表格、图片等。"
+        "content 为 JSON 数组，每项: {type, text/items/headers/rows/path...}。"
+        "type 可选: h1/h2/h3/body/bullet/numbered/table/image/pagebreak/spacer。"
+    )
+    category = "doc"
+    dangerous = True
+    parameters = [
+        SkillParameter("content", "string",
+            "内容块 JSON 数组。示例: [{\"type\":\"h1\",\"text\":\"报告\"},{\"type\":\"body\",\"text\":\"正文\"}]",
+            required=True),
+        SkillParameter("output_path", "string", "输出 DOCX 文件路径", required=True),
+        SkillParameter("title", "string", "文档标题", required=False, default=""),
+        SkillParameter("author", "string", "作者", required=False, default=""),
+        SkillParameter("font", "string", "中文字体", required=False, default="SimHei"),
+    ]
+
+    async def execute(self, content: str = "", output_path: str = "",
+                      title: str = "", author: str = "",
+                      font: str = "SimHei", **kwargs) -> SkillResult:
+        try:
+            from docx import Document
+            from docx.shared import Inches, Cm, Pt, Emu, RGBColor
+            from docx.enum.text import WD_ALIGN_PARAGRAPH
+            from docx.enum.table import WD_TABLE_ALIGNMENT
+        except ImportError:
+            return SkillResult(success=False, error="python-docx 未安装: pip install python-docx")
+
+        try:
+            blocks = json.loads(content) if isinstance(content, str) else content
+            if not isinstance(blocks, list):
+                blocks = [{"type": "body", "text": str(blocks)}]
+        except json.JSONDecodeError as e:
+            return SkillResult(success=False, error=f"content JSON 解析失败: {e}")
+
+        try:
+            out = Path(output_path).expanduser().resolve()
+            out.parent.mkdir(parents=True, exist_ok=True)
+
+            doc = Document()
+            if title:
+                doc.core_properties.title = title
+            if author:
+                doc.core_properties.author = author
+
+            style = doc.styles["Normal"]
+            rpr = style.font
+            rpr.name = font
+            rpr.size = Pt(11)
+
+            for level, size in [("Heading 1", 22), ("Heading 2", 16), ("Heading 3", 13)]:
+                if level in doc.styles:
+                    hs = doc.styles[level]
+                    hs.font.name = font
+                    hs.font.size = Pt(size)
+                    hs.font.color.rgb = RGBColor(0x1a, 0x36, 0x5d)
+
+            for block in blocks:
+                if not isinstance(block, dict):
+                    continue
+                bt = block.get("type", "body")
+
+                if bt in ("h1", "h2", "h3"):
+                    heading_map = {"h1": 0, "h2": 1, "h3": 2}
+                    doc.add_heading(block.get("text", ""), level=heading_map[bt] + 1)
+                elif bt == "body":
+                    p = doc.add_paragraph(block.get("text", ""))
+                    p.paragraph_format.first_line_indent = Cm(0.74)
+                    p.paragraph_format.line_spacing = 1.5
+                elif bt == "bullet":
+                    for item in block.get("items", []):
+                        doc.add_paragraph(item, style="List Bullet")
+                elif bt == "numbered":
+                    for item in block.get("items", []):
+                        doc.add_paragraph(item, style="List Number")
+                elif bt == "table":
+                    headers = block.get("headers", [])
+                    rows = block.get("rows", [])
+                    if headers:
+                        table = doc.add_table(rows=1 + len(rows), cols=len(headers))
+                        table.style = "Light Grid Accent 1"
+                        table.alignment = WD_TABLE_ALIGNMENT.CENTER
+                        for j, h in enumerate(headers):
+                            cell = table.rows[0].cells[j]
+                            cell.text = str(h)
+                            for run in cell.paragraphs[0].runs:
+                                run.bold = True
+                        for i, row in enumerate(rows):
+                            for j, val in enumerate(row):
+                                if j < len(table.columns):
+                                    table.rows[i + 1].cells[j].text = str(val)
+                elif bt == "image":
+                    img_path = block.get("path", "")
+                    img_w = block.get("width", 0)
+                    if img_path and os.path.isfile(img_path):
+                        try:
+                            if img_w > 0:
+                                doc.add_picture(img_path, width=Emu(int(img_w)))
+                            else:
+                                doc.add_picture(img_path, width=Inches(5.5))
+                        except Exception as e:
+                            doc.add_paragraph(f"[图片加载失败: {e}]")
+                elif bt == "pagebreak":
+                    doc.add_page_break()
+                elif bt == "spacer":
+                    doc.add_paragraph("")
+
+            doc.save(str(out))
+            return SkillResult(
+                success=True,
+                message=f"Word 文档已生成: {out}",
+                files=[str(out)],
+                data={"path": str(out), "blocks": len(blocks)},
+            )
+        except Exception as e:
+            return SkillResult(success=False, error=f"DOCX 生成失败: {e}")
+
+
+class DOCXReadSkill(Skill):
+    """读取 Word 文档内容"""
+    name = "docx_read"
+    description = "读取 Word (DOCX) 文件，提取文本内容。支持提取表格数据。"
+    category = "doc"
+    parameters = [
+        SkillParameter("path", "string", "DOCX 文件路径", required=True),
+        SkillParameter("max_chars", "integer", "最大字符数", required=False, default=50000),
+    ]
+
+    async def execute(self, path: str = "", max_chars: int = 50000, **kwargs) -> SkillResult:
+        try:
+            from docx import Document
+        except ImportError:
+            return SkillResult(success=False, error="python-docx 未安装: pip install python-docx")
+
+        fp = Path(path).expanduser().resolve()
+        if not fp.exists():
+            return SkillResult(success=False, error=f"文件不存在: {path}")
+
+        try:
+            doc = Document(str(fp))
+            parts = []
+            for para in doc.paragraphs:
+                style_name = para.style.name if para.style else ""
+                prefix = ""
+                if "Heading" in style_name:
+                    level = style_name.replace("Heading", "").strip()
+                    prefix = f"[{'#' * int(level) if level.isdigit() else '#'}] "
+                elif "Bullet" in style_name:
+                    prefix = "  * "
+                elif "Number" in style_name or "List" in style_name:
+                    prefix = "  - "
+                text = para.text.strip()
+                if text:
+                    parts.append(f"{prefix}{text}")
+
+            for i, table in enumerate(doc.tables):
+                rows = []
+                for row in table.rows:
+                    cells = [cell.text.strip() for cell in row.cells]
+                    rows.append(" | ".join(cells))
+                if rows:
+                    parts.append(f"\n[表格 {i+1}]")
+                    parts.append(rows[0])
+                    if len(rows) > 1:
+                        parts.append("-" * len(rows[0]))
+                        parts.extend(rows[1:])
+
+            full = "\n".join(parts)
+            if len(full) > max_chars:
+                full = full[:max_chars] + f"\n\n... (截断，共 {len(full)} 字符)"
+
+            return SkillResult(
+                success=True,
+                message=f"已读取 Word 文档: {fp.name}",
+                data={"path": str(fp), "paragraphs": len(doc.paragraphs),
+                      "tables": len(doc.tables)},
+                output=full,
+            )
+        except Exception as e:
+            return SkillResult(success=False, error=f"DOCX 读取失败: {e}")