WengChat v2.19.4

What's Changed

• Updated the MinerU adapter for the official v1 API, verified against MinerU 4.0.11. Fixed PDF page-range syntax and stopped sending page ranges for images and other non-PDF files.
• Fixed the MinerU v0 page-limit off-by-one error and removed unavailable task counters from the v1 health display.
• Prevented batch ZIP downloads from overwriting files with identical names by placing each document in its own directory and preserving relative image paths.
• Pinned the default Docker Compose images to MinerU 3.4.5 (v0) and 4.0.11 (v1). The selected profile's tag can still be overridden with MINERU_TAG.
• Merged upstream NextChat icon, screenshot, and AI-assistant wording updates while preserving WengChat branding and local features.
• Fixed a Jest mock type mismatch so the full TypeScript check passes.

更新内容

• 适配正式 MinerU v1 API，并使用 MinerU 4.0.11 完成验证。修正 PDF 页码范围语法，图片及其他非 PDF 文件不再发送页码范围。
• 修复 MinerU v0 最大页数多解析一页的问题，移除 v1 健康信息中服务不提供的任务计数。
• 批量 ZIP 下载为每份文档建立独立目录，避免同名文件覆盖，并保留图片相对路径。
• Docker Compose 默认使用 MinerU 3.4.5（v0）和 4.0.11（v1），仍可通过 MINERU_TAG 覆盖所选 profile 的镜像标签。
• 合并上游 NextChat 的图标、截图和通用 AI 助手文案更新，保留 WengChat 品牌与本地功能。
• 修复 Jest mock 类型不匹配问题，使完整 TypeScript 检查通过。

Full Changelog: v2.19.3...v2.19.4

WengChat v2.19.3

What's Changed

• Merged upstream NextChat changes.
• Added support for MinerU v0 and v1 APIs with automatic server-version detection.
• Added image support for GPT-6 models and updated model-specific request parameters.
• Fixed bugs.

更新内容

• 合并上游 NextChat 的更新。
• 支持 MinerU v0 和 v1 API，并自动识别服务端版本。
• 支持 GPT-6 系列模型的图片输入，并调整对应的请求参数。
• 修复 Bug。

Full Changelog: v2.19.2...v2.19.3

WengChat v2.19.2

What's Changed

• Rebranded NextChat as WengChat.
• Fixed CVE-2026-7644 (Issue #2) by disabling MCP by default, enforcing authentication, and blocking remote process configuration.

更新内容

• 将 NextChat 更名为 WengChat。
• 修复 CVE-2026-7644 (Issue #2)：MCP 默认关闭、强制鉴权且禁止远程配置进程。

Full Changelog: v2.19.1...v2.19.2

v2.19.1

What's Changed

Bug Fixes

• Fixed Clear All button not resetting conversion log to initial placeholder
• Fixed infinite retry loop when multiple files fail conversion: each file now attempted only once per run, previously failed files are retried on next click

Improvements

• Download All changed to package-and-download with zip extraction (MinerU zips and MarkItDown .md files bundled together)
• Updated MinerU engine to latest API: renamed backends (vlm-auto-engine → vlm-engine, hybrid-auto-engine → hybrid-engine), removed deprecated OCR languages (ch_lite, en, japan, chinese_cht, latin), added parse effort setting (medium / high) for hybrid backend

更新内容

问题修复

• 修复 全部清除 按钮未将转换信息重置为初始占位文本
• 修复多个文件转换失败时无限重试循环：每轮每个文件仅尝试一次，再次点击按钮时失败文件可重新尝试

改进优化

• 下载全部 改为 zip 打包提取（MinerU 的 zip 包与 MarkItDown 的 .md 文件统一打包）
• 更新 MinerU 引擎至最新 API：重命名解析后端（vlm-auto-engine → vlm-engine，hybrid-auto-engine → hybrid-engine），移除已废弃的 OCR 语言（ch_lite、en、japan、chinese_cht、latin），新增 hybrid 后端的解析精度设置（medium / high）

Full Changelog: v2.19.0...v2.19.1

v2.19.0

What's Changed

Features

• Added File Conversion page supporting MarkItDown and MinerU engines with configurable parse settings, health check, and batch conversion with download
• Added Docker Compose for optional service deployment

Improvements

• Added docker-build.sh helper script
• Restructured README and clean up docs/ directory

功能更新

• 新增文件转换页面，支持 MarkItDown 和 MinerU 双引擎、可配置解析参数、健康检查、批量转换与下载
• 新增 Docker Compose，支持按需部署可选服务

改进优化

• 新增 docker-build.sh 构建辅助脚本
• 重构 README 并清理 docs/ 文件夹

Full Changelog: v2.18.0...v2.19.0

v2.18.0

What's Changed

Features

• Added support for the OpenAI Responses format /v1/responses and the gpt-5.4-pro model
• Added support for GPT-5 series model parameters, including reasoning_effort, response_format, and verbosity
• Added support for image-to-image generation via /v1/images/edits and the gpt-image-2 model

Bug Fixes

• Increased the maximum number of uploaded images from 3 to 15
• Updated translations and localization
• Fixed an issue where api-version could be empty for Azure OpenAI
• Fixed an issue where the default model and default summary model could not be loaded
• Fixed model invocation issues when generating conversation titles

功能更新

• 支持 OpenAI Responses 格式 /v1/responses 和 gpt-5.4-pro 模型
• 支持 GPT-5 系列模型的推理强度 reasoning_effort、响应格式 response_format、详细程度 verbosity 参数
• 支持图生图模式 /v1/images/edits 和 gpt-image-2 模型

Bug 修复

• 上传图片数量由 3 扩大到 15
• 更新翻译和本地化
• 修复 Azure OpenAI 的 api-version 为空的问题
• 修复默认模型、默认摘要模型无法读取的问题
• 修复起对话标题时的模型调用相关问题

Full Changelog: v2.17.0...v2.18.0

v2.17.0

What's Changed

• 修复同步时切换页面触发多次同步的问题
• 支持调用 gpt-image-1 家族的模型生成图像
• 支持自定义对话摘要模型

Full Changelog: v2.16.2...v2.17.0

v2.16.2

What's Changed

• 20251223 客制化更新 by @Na2CuCl4 in #1
• 更新 eslint
• 更新多语言支持
• 修改服务器响应时间为 10 分钟
• 修改默认插件
• 移除广告
• 实现上传文件功能
• 实现自动更新
• 其他细节调整

New Contributors

• @Na2CuCl4 made their first contribution in #1

Full Changelog: https://github.com/Na2CuCl4/NextChat/commits/v2.16.2