Releases: ChatLunaLab/chatluna
Releases · ChatLunaLab/chatluna
v1.3.9
大家好!此次更新修复了之前版本含有的 Bug。
新特性
- core: 无限上下文支持设置总结的上下文上限 (#682)
修复 & 改进
无
What's Changed
- [Feature] support configurable infinite context threshold and manual compression by @dingyi222666 in #682
- [chore] bump version to 1.3.9 by @dingyi222666 in #683
Full Changelog: v1.3.8...v1.3.9
v1.3.8
大家好!此次更新修复了之前版本含有的 Bug。
新特性
- core: 支持结合 chatluna storage service,提供文件的读取支持 (#679)
修复 & 改进
What's Changed
- [Fix] prevent image handling for thought parts in Gemini adapter by @dingyi222666 in #671
- [Fix] Deadlock in message_delay for multi-user concurrent conversations by @springtiming in #673
- [Feature] support file message transformation for OneBot platform by @dingyi222666 in #679
- chore(build): bump versions and update dependencies by @dingyi222666 in #680
Full Changelog: v1.3.7...v1.3.8
v1.3.7
大家好!此次更新修复了之前版本含有的 Bug。
新特性
无
修复 & 改进
- core: 修复房间系统的一些问题,主要是消息队列的触发 (#668)
What's Changed
- [Fix] refine message queue delay logic and state management by @dingyi222666 in #668
- [Chore] bump version to 1.3.7 by @dingyi222666 in #669
Full Changelog: v1.3.6...v1.3.7
v1.3.6
大家好!此次更新修复了之前版本含有的 Bug。
新特性
无
修复 & 改进
- gemini-adapter, openai-adapter: 支持 gemini 3 flash, openai 系列模型的思考变体选择 (#665)
- core: 修复房间系统的一些问题 (#662 #663)
What's Changed
- fix(core): refine room resolution logic in clear_room middleware by @springtiming in #662
- [Fix] Catch errors from after-chat hooks to avoid process crash by @springtiming in #663
- [Fix] Filter tool-related messages before compression in infinite context by @dingyi222666 in #664
- [Feature] support model thinking levels and dynamic model fetching by @dingyi222666 in #665
- chore(release): bump version to 1.3.6 by @dingyi222666 in #666
Full Changelog: v1.3.5...v1.3.6
v1.3.5
大家好!此次更新修复了之前版本含有的 Bug。
新特性
- dify-adapter: 支持图片多模态输入,优化很多东西东西 (#653)
修复 & 改进
- gemini-adapter: 图像模型支持选择 2K,4K 变体 (#645)
- extension-tools: 命令工具使用聚合选择,不会在注册很多工具浪费 token (#649
- core: 修复房间系统的一些问题 (#656)
What's Changed
- [Feature] Add Gemini 3 thinking levels and improve model detection by @dingyi222666 in #647
- [Refactor] Simplify Gemini thinking level logic by @dingyi222666 in #648
- feat(extension-tools): refactor command tools to unified interface by @dingyi222666 in #649
- [Fix] Disable thinking parameters for non-Gemini 3 models by @dingyi222666 in #654
- [Feature] Support Dify image uploads with data URL normalizatio by @springtiming in #653
- [Fix] Improve room system and infinite context handling by @dingyi222666 in #656
- chore(release): bump version to 1.3.5 by @dingyi222666 in #657
New Contributors
- @springtiming made their first contribution in #653
Full Changelog: v1.3.4...v1.3.5
v1.3.4
v1.3.3
v1.3.1
大家好!此次更新修复了之前版本含有的 Bug。
新特性
修复 & 改进
What's Changed
- [Feature] Add Gemini 3.0 Pro and thinking mode support by @dingyi222666 in #628
- [Feature] Add Gemini 3.0 thinking level and systemInstruction casing support by @dingyi222666 in #629
- fix(adapter-gemini): clear thinkingLevel when thinking mode is disabled by @dingyi222666 in #630
- [Feature] Add Gemini 3.0 image generation support by @dingyi222666 in #631
- [Feature] MCP client reconnection and Gemini API improvements by @dingyi222666 in #635
- [Fix] Optimize GIF frame extraction with incremental decoding by @dingyi222666 in #637
- [Feature] Add Gemini 3.0 image generation with configurable sizes by @dingyi222666 in #638
- [Release] Version 1.3.1 by @dingyi222666 in #639
Full Changelog: v1.3.0...v1.3.1
v1.3.0
大家好!此次更新是 ChatLuna v1.3.0 的正式版本,包含大量新特性、性能优化和 Bug 修复。这是一个重要的里程碑版本,建议所有用户更新。
新特性
核心功能
- core: 无限上下文对话支持,自动压缩历史消息,支持超长对话 (#602)
- actions: 支持注册指令调用 ChatLuna,支持多模态识图输入,一次性指令生成 (https://github.com/ChatLunaLab/actions, #386)
- mcp-client: 为 ChatLuna 提供 MCP 客户端支持,支持使用代理 (#618, #605)
- variable-extension: 扩展 ChatLuna 的变量输入,支持群内消息变量和农历变量 (8016c34)
长期记忆
- long-memory: 长期记忆系统重构,新增三种长期记忆引擎(Basic, HippoRAG, EMGAS),优化提取 Prompt 和性能 (#542)
- long-memory: 支持 Guild 层级记忆,实现群组共享记忆功能 (#577)
- long-memory: 增强记忆提取,使用文本指纹提升召回质量
向量数据库与 RAG
- vector-store-service: 新增 Neo4j 向量存储支持 (#439)
- vector-store-service: 实现 LightRAG (API) 检索器,提供新的 RAG 策略 (#582)
- vector-store-service: 重构向量数据库架构,支持数据库持久化 (#543)
图片服务
- image-service: 图片识别服务,使用支持多模态输入的模型为不支持多模态的模型提供图片输入 (#392, 47aafef)
- image-service: 新增 GIF 动画帧提取功能,支持可配置的提取策略(首帧/平均帧/头部帧) (#572)
搜索服务
工具系统
- extension-tools: 增强命令工具,支持命令黑名单和改进的命名规范化 (#599)
- extension-tools: 新增文件系统工具,支持文件忽略模式和批量操作 (#575)
- extension-tools: 新增任务列表(todos)工具,支持任务分解和进度跟踪
- extension-tools: 新增定时任务(cron)工具,支持数据库持久化和灵活的时间解析 (#575)
- extension-tools: 新增域名请求头配置,支持不同域名使用不同的请求头 (#581)
模型适配器
- adapter-doubao: 新增字节豆包适配器
- qwen-adapter: 支持 Qwen3 系列模型 (#558)
- gemini-adapter: 支持 Gemini 2.5 Flash Image Preview 图像生成模型
- claude-adapter: 新增 Claude Opus 4.5 和 Claude Sonnet 4.5 模型支持 (#584)
- zhipu-adapter: 新增 GLM-4.6 模型支持(200k 上下文窗口) (#584)
修复 & 改进
核心系统
- core: 新增纯文本渲染格式,自动去除 markdown 附加格式文本 (f60f338, #460)
- core: 修复流式文本吞字问题 (7565371, #417)
- core: 修复发图片容易触发对话的问题 (#438, cf07d23)
- core: 修复用房间名直接触发聊天的 Bug (cf07d23, e8d0bb3, #432)
- core: 优化 ReAct 模式下的 Prompt (c8c6bde)
- core: 优化锁超时的错误日志 (de2adfb)
- core: 调用命令聊天时支持图片多模态(需要模型本身支持多模态)(6ed564c)
流式响应与性能
- core,adapters: 迁移 token 使用统计到 response_metadata,新增流式重试机制 (#621)
- core: 实现流式缓冲架构,使用 Web Streams API 提升性能 (#568)
- core: 优化超时处理,使用标准 Promise API (#596)
- adapters: 修复流式响应处理中的 continue 语句问题 (#587)
错误处理
- core,extension-mcp: 改进工具错误处理和异步操作 (#622)
- core: 新增 chatluna/after-chat-error 事件,改进错误清理流程
- core: 改进图片读取失败的错误处理 (#603)
- core: 改进代理配置验证和错误处理 (#568)
配置与架构
- adapters: 标准化 modelMaxContextSize 处理 (#620)
- adapters: 用 maxContextRatio 替换 maxTokens,支持百分比上下文管理 (#565)
- adapters: API 密钥配置改为可选,支持启用/禁用切换 (#585)
- core: 新增通知系统,指导用户配置模型 (#585)
- core: 新增模型可用性验证 (#584)
- core: 新增模型测试命令 chatluna.model.test (#584)
工具系统改进
- core: 新增 chatluna.tool.list 命令,支持分页显示工具 (#588)
- core: 重构工具选择和向量存储操作 (#592)
- core: 新增工具缓存到 PlatformService
- core: 改进工具管理,新增唯一 ID 和名称分配
- extension-mcp: 改进 MCP 工具处理和超时支持 (#613)
- extension-mcp: 简化 MCP 命令,改进配置解析和工具注册 (#579)
依赖与兼容性
- core: 迁移到 Vue 响应式系统,改进代理和预设架构 (#570)
- core: 替换 uuid 和 md5 为 Node.js 内置 crypto API (#590)
- core: 升级 @langchain/core 到 0.3.62
- core,adapters: 移除未使用的依赖,标准化 crypto 导入 (#590)
其他改进
- core: 新增本地文件日志记录用于调试 (#571)
- core: 改进缓存密钥迁移和错误处理 (#567)
- core: 新增优先级支持到消息转换拦截器 (#604)
- core: 改进中间件配置和作用域处理 (#564)
- core: 新增最小长度阈值用于转发消息 (#581)
架构重构
- refactor: 实现 Vue 响应式计算系统,替换静态方法调用 (#548)
- refactor: 重构 ChatChain 支持并行中间件执行 (#548)
- refactor: 迁移预设系统到 Vue 响应式 (#570)
- refactor: 提取并中心化 chunkArray 工具函数
- refactor: 重构向量存储架构,支持数据库持久化 (#543)
- refactor: 实现 shared-prompt-renderer,支持高级模板功能 (#569)
- refactor: 移除平台客户端注册中未使用的 ctx 参数 (#586)
破坏性变更
- 移除 ChatGLM 和 GPT Free 适配器
- 移除 Neo4j 向量存储支持
- 默认聊天模式从 'chat' 更改为 'agent'
- 默认预设从 'chatgpt' 更改为 'sydney'
- PresetService 方法现在返回 ComputedRef 而非 Promise
- 长期记忆系统完全重构,不兼容旧版本数据
What's Changed
- chore(deps): bump marked from 15.0.12 to 16.1.1 by @dependabot[bot] in #516
- update glm model list by @LoserFox in #527
- feat(docs): update star history chart to support dark and light themes by @weiruchenai1 in #536
- [Refactor] 新的向量数据库实现 by @dingyi222666 in #543
- fix: improve multimodal image processing and message handling by @dingyi222666 in #545
- fix: core chat message handling improvements and version updates by @dingyi222666 in #546
- [Fix] 修复 Zod 工具调用兼容性和消息延迟中间件问题 by @dingyi222666 in #547
- [Refactor] 响应式系统实现,更好的重载实现,减少 async 传播性 by @dingyi222666 in #548
- chore: bump package versions to alpha.42/alpha.4 by @dingyi222666 in #549
- [Feature] 实现 RAG 检索增强生成系统 by @dingyi222666 in #551
- [Fix] 修复核心服务错误处理和生命周期管理问题 by @dingyi222666 in #552
- chore: 发布 alpha.43/alpha.5 版本并修复工具调用问题 by @dingyi222666 in #554
- [Refactor] 长期记忆实现重构,新增三种长期记忆引擎,优化提取 Prompt 和性能 by @Hoshino-Yumetsuki in #542
- Feature: Add support for Qwen3 models and improve model handling by @dingyi222666 in #558
- Fix: Bump adapter versions and fix shared utils initialization by @dingyi222666 in #560
- Fix: search-service typo and model reference issues by @dingyi222666 in #562
- Fix: Core middleware configuration and scope handling issues by @dingyi222666 in #564
- [Refactor] Replace maxTokens with maxContextRatio and version updates by @dingyi222666 in #565
- chore(deps): update dependencies across multiple packages by @dingyi222666 in #566
- [Fix] Improve error handling and migrate cache keys by @dingyi222666 in #567
- [Fix] Fix room resolution and proxy configuration issues by @dingyi222666 in #568
- [Feature] Implement shared-prompt-renderer with advanced template features by @dingyi222666 in #569
- [Refactor] Migrate to Vue reactivity system and enhance agent/preset architecture by @dingyi222666 in #570
- [Feature] Add local file logging for debugging by @dingyi222666 in #571
- [Feature] Add GIF animation support and improve image handling by @dingyi222666 in #572
- feat: switch to sydney preset and remove code sandbox by @dingyi222666 in #575
- [Fix] Fix double brace escape sequence behavior by @dingyi222666 in #576
- [Feature] Long-term memory enhancements and prompt rendering improvements by @dingyi222666 in #577
- feat: MCP tools management and adapter configuration enhancements by @dingyi222666 in #578
- [Fix] MCP client configuration parsing and tool registration by @dingyi222666 in #579
- fix(extension-mcp,extension-tools): improve MCP client initialization and tool selector logic by @dingyi222666 in #581
- [Feature] Add LightRAG (API) retriever implementation by @dingyi222666 in #582
- Fix: MCP client improvements and core stability enhancements by @dingyi222666 in #583
- feat(adapters,core): enhance model testing and token usage tracking by @dingyi222666 in #584
- [Feature] Add notification support for model configuration guidance by @dingyi222666 in #585
- [Refactor] Remove unused ctx parameter from platform client registration by @dingyi222666 in #586
- [Fix] Fix streaming response handling and bump version to 1.3.0-alpha.60 by @dingyi222666 in #587
- [Feature] Add tool listing command and refactor reactive chain management by @dingyi222666 in #588
- [Refactor] Migrate to Node.js crypto APIs and enhance prompt context by @dingyi222666 in #590
- [Fix] Refactor tool selection and fix vector store operations by @dingyi222666 in #592
- [Chore] Bump version to 1.3.0-alpha.64 by @dingyi222666 in https://github...
v1.2.1
大家好!此次更新修复了之前版本含有的 Bug。各位可以选择性更新。
新特性
无。
修复 & 改进
- core: 优化 ReAct 模式下的 Prompt (c8c6bde)
- core: 优化锁超时的错误日志 (de2adfb)
- core: 配合 chatluna character 伪装添加一个小东西 (7c981ef)
Full Changelog: v1.2.0...v1.2.1