v1.3.0

[dingyi222666]

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 ([Feature] Improve agent system with infinite context and better prompts #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, [Feature] 新增指令用于手动调用模型 #386)
• mcp-client: 为 ChatLuna 提供 MCP 客户端支持，支持使用代理 ([Feature] 支持 mcp 使用代理 #618, [Refactor] Refactor MCP commands to use core command pattern #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 ([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 #542)
• long-memory: 支持 Guild 层级记忆，实现群组共享记忆功能 ([Feature] Long-term memory enhancements and prompt rendering improvements #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 (添加 Neo4j 向量数据库支持 #439)
• vector-store-service: 实现 LightRAG (API) 检索器，提供新的 RAG 策略 ([Feature] Add LightRAG (API) retriever implementation #582)
• vector-store-service: 重构向量数据库架构，支持数据库持久化 ([Refactor] 新的向量数据库实现 #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 ([Feature] 希望能有类似于search-service一样的调用别的模型以实现聊天模型没有的视觉功能 #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） ([Feature] Add GIF animation support and improve image handling #572)

搜索服务

• search-service: 支持搜索服务的上下文压缩，调用小模型总结文本 (7d77cf6, [Feature] 搜索服务支持 Contextual Compression 上下文压缩 #444)

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 ([Feature] Enhance command tool with improved normalization and blacklist support #599)
• extension-tools: 新增文件系统工具，支持文件忽略模式和批量操作 (feat: switch to sydney preset and remove code sandbox #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 (feat: switch to sydney preset and remove code sandbox #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 (fix(extension-mcp,extension-tools): improve MCP client initialization and tool selector logic #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 (Feature: Add support for Qwen3 models and improve model handling #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 (feat(adapters,core): enhance model testing and token usage tracking #584)
• zhipu-adapter: 新增 GLM-4.6 模型支持（200k 上下文窗口） (feat(adapters,core): enhance model testing and token usage tracking #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, [Question] 问一下onebot平台接收到markdown格式的消息要怎样过滤没有用的字符呢？ #460)
• core: 修复流式文本吞字问题 (7565371, [Bug] 搜索模式会吞回答第一个字 #417)
• core: 修复发图片容易触发对话的问题 ([Bug] 做图片数据检测，防止提示词误触 #438, cf07d23)
• core: 修复用房间名直接触发聊天的 Bug (cf07d23, e8d0bb3, [Bug] 群里用房间名前缀触发对话失效 #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 ([Fix] Migrate token usage to response_metadata and add stream retry #621)
• core: 实现流式缓冲架构，使用 Web Streams API 提升性能 ([Fix] Fix room resolution and proxy configuration issues #568)
• core: 优化超时处理，使用标准 Promise API ([Refactor] Optimize timeout handling in model requests #596)
• adapters: 修复流式响应处理中的 continue 语句问题 ([Fix] Fix streaming response handling and bump version to 1.3.0-alpha.60 #587)

错误处理

• core,extension-mcp: 改进工具错误处理和异步操作 ([Fix] Improve tool error handling and async operations #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 ([Fix] Add error handling for image reading failures #603)
• core: 改进代理配置验证和错误处理 ([Fix] Fix room resolution and proxy configuration issues #568)

配置与架构

• adapters: 标准化 modelMaxContextSize 处理 ([Fix] Standardize modelMaxContextSize handling and bump versions #620)
• adapters: 用 maxContextRatio 替换 maxTokens，支持百分比上下文管理 ([Refactor] Replace maxTokens with maxContextRatio and version updates #565)
• adapters: API 密钥配置改为可选，支持启用/禁用切换 ([Feature] Add notification support for model configuration guidance #585)
• core: 新增通知系统，指导用户配置模型 ([Feature] Add notification support for model configuration guidance #585)
• core: 新增模型可用性验证 (feat(adapters,core): enhance model testing and token usage tracking #584)
• core: 新增模型测试命令 chatluna.model.test (feat(adapters,core): enhance model testing and token usage tracking #584)

工具系统改进

• core: 新增 chatluna.tool.list 命令，支持分页显示工具 ([Feature] Add tool listing command and refactor reactive chain management #588)
• core: 重构工具选择和向量存储操作 ([Fix] Refactor tool selection and fix vector store operations #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 ([Fix] Improve MCP tool handling and timeout support #613)
• extension-mcp: 简化 MCP 命令，改进配置解析和工具注册 ([Fix] MCP client configuration parsing and tool registration #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 ([Refactor] Migrate to Vue reactivity system and enhance agent/preset architecture #570)
• core: 替换 uuid 和 md5 为 Node.js 内置 crypto API ([Refactor] Migrate to Node.js crypto APIs and enhance prompt context #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 ([Refactor] Migrate to Node.js crypto APIs and enhance prompt context #590)

其他改进

• core: 新增本地文件日志记录用于调试 ([Feature] Add local file logging for debugging #571)
• core: 改进缓存密钥迁移和错误处理 ([Fix] Improve error handling and migrate cache keys #567)
• core: 新增优先级支持到消息转换拦截器 ([Feature] Add priority support to message transform interceptors #604)
• core: 改进中间件配置和作用域处理 (Fix: Core middleware configuration and scope handling issues #564)
• core: 新增最小长度阈值用于转发消息 (fix(extension-mcp,extension-tools): improve MCP client initialization and tool selector logic #581)

架构重构

• refactor: 实现 Vue 响应式计算系统，替换静态方法调用 ([Refactor] 响应式系统实现，更好的重载实现，减少 async 传播性 #548)
• refactor: 重构 ChatChain 支持并行中间件执行 ([Refactor] 响应式系统实现，更好的重载实现，减少 async 传播性 #548)
• refactor: 迁移预设系统到 Vue 响应式 ([Refactor] Migrate to Vue reactivity system and enhance agent/preset architecture #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 ([Refactor] 新的向量数据库实现 #543)
• refactor: 实现 shared-prompt-renderer，支持高级模板功能 ([Feature] Implement shared-prompt-renderer with advanced template features #569)
• refactor: 移除平台客户端注册中未使用的 ctx 参数 ([Refactor] Remove unused ctx parameter from platform client registration #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 chore(deps): bump marked from 15.0.12 to 16.1.1 #516
• update glm model list by @LoserFox in update glm model list #527
• feat(docs): update star history chart to support dark and light themes by @weiruchenai1 in feat(docs): update star history chart to support dark and light themes #536
• [Refactor] 新的向量数据库实现 by @dingyi222666 in [Refactor] 新的向量数据库实现 #543
• fix: improve multimodal image processing and message handling by @dingyi222666 in fix: improve multimodal image processing and message handling #545
• fix: core chat message handling improvements and version updates by @dingyi222666 in fix: core chat message handling improvements and version updates #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 [Fix] 修复 Zod 工具调用兼容性和消息延迟中间件问题 #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 [Refactor] 响应式系统实现，更好的重载实现，减少 async 传播性 #548
• chore: bump package versions to alpha.42/alpha.4 by @dingyi222666 in chore: bump package versions to alpha.42/alpha.4 #549
• [Feature] 实现 RAG 检索增强生成系统 by @dingyi222666 in [Feature] 实现 RAG 检索增强生成系统 #551
• [Fix] 修复核心服务错误处理和生命周期管理问题 by @dingyi222666 in [Fix] 修复核心服务错误处理和生命周期管理问题 #552
• chore: 发布 alpha.43/alpha.5 版本并修复工具调用问题 by @dingyi222666 in chore: 发布 alpha.43/alpha.5 版本并修复工具调用问题 #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 [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 #542
• Feature: Add support for Qwen3 models and improve model handling by @dingyi222666 in Feature: Add support for Qwen3 models and improve model handling #558
• Fix: Bump adapter versions and fix shared utils initialization by @dingyi222666 in Fix: Bump adapter versions and fix shared utils initialization #560
• Fix: search-service typo and model reference issues by @dingyi222666 in Fix: search-service typo and model reference issues #562
• Fix: Core middleware configuration and scope handling issues by @dingyi222666 in Fix: Core middleware configuration and scope handling issues #564
• [Refactor] Replace maxTokens with maxContextRatio and version updates by @dingyi222666 in [Refactor] Replace maxTokens with maxContextRatio and version updates #565
• chore(deps): update dependencies across multiple packages by @dingyi222666 in chore(deps): update dependencies across multiple packages #566
• [Fix] Improve error handling and migrate cache keys by @dingyi222666 in [Fix] Improve error handling and migrate cache keys #567
• [Fix] Fix room resolution and proxy configuration issues by @dingyi222666 in [Fix] Fix room resolution and proxy configuration issues #568
• [Feature] Implement shared-prompt-renderer with advanced template features by @dingyi222666 in [Feature] Implement shared-prompt-renderer with advanced template features #569
• [Refactor] Migrate to Vue reactivity system and enhance agent/preset architecture by @dingyi222666 in [Refactor] Migrate to Vue reactivity system and enhance agent/preset architecture #570
• [Feature] Add local file logging for debugging by @dingyi222666 in [Feature] Add local file logging for debugging #571
• [Feature] Add GIF animation support and improve image handling by @dingyi222666 in [Feature] Add GIF animation support and improve image handling #572
• feat: switch to sydney preset and remove code sandbox by @dingyi222666 in feat: switch to sydney preset and remove code sandbox #575
• [Fix] Fix double brace escape sequence behavior by @dingyi222666 in [Fix] Fix double brace escape sequence behavior #576
• [Feature] Long-term memory enhancements and prompt rendering improvements by @dingyi222666 in [Feature] Long-term memory enhancements and prompt rendering improvements #577
• feat: MCP tools management and adapter configuration enhancements by @dingyi222666 in feat: MCP tools management and adapter configuration enhancements #578
• [Fix] MCP client configuration parsing and tool registration by @dingyi222666 in [Fix] MCP client configuration parsing and tool registration #579
• fix(extension-mcp,extension-tools): improve MCP client initialization and tool selector logic by @dingyi222666 in fix(extension-mcp,extension-tools): improve MCP client initialization and tool selector logic #581
• [Feature] Add LightRAG (API) retriever implementation by @dingyi222666 in [Feature] Add LightRAG (API) retriever implementation #582
• Fix: MCP client improvements and core stability enhancements by @dingyi222666 in Fix: MCP client improvements and core stability enhancements #583
• feat(adapters,core): enhance model testing and token usage tracking by @dingyi222666 in feat(adapters,core): enhance model testing and token usage tracking #584
• [Feature] Add notification support for model configuration guidance by @dingyi222666 in [Feature] Add notification support for model configuration guidance #585
• [Refactor] Remove unused ctx parameter from platform client registration by @dingyi222666 in [Refactor] Remove unused ctx parameter from platform client registration #586
• [Fix] Fix streaming response handling and bump version to 1.3.0-alpha.60 by @dingyi222666 in [Fix] Fix streaming response handling and bump version to 1.3.0-alpha.60 #587
• [Feature] Add tool listing command and refactor reactive chain management by @dingyi222666 in [Feature] Add tool listing command and refactor reactive chain management #588
• [Refactor] Migrate to Node.js crypto APIs and enhance prompt context by @dingyi222666 in [Refactor] Migrate to Node.js crypto APIs and enhance prompt context #590
• [Fix] Refactor tool selection and fix vector store operations by @dingyi222666 in [Fix] Refactor tool selection and fix vector store operations #592
• [Chore] Bump version to 1.3.0-alpha.64 by @dingyi222666 in [Chore] Bump version to 1.3.0-alpha.64 #593
• [Fix] Fix adapter plugin apiKeys configuration save bug by @shangxueink in [Fix] Fix adapter plugin apiKeys configuration save bug #594
• [Fix] Improve tool system stability and simplify MCP commands by @dingyi222666 in [Fix] Improve tool system stability and simplify MCP commands #595
• [Refactor] Optimize timeout handling in model requests by @dingyi222666 in [Refactor] Optimize timeout handling in model requests #596
• [Fix] Fix long-memory plugin lifecycle and initialization issues by @dingyi222666 in [Fix] Fix long-memory plugin lifecycle and initialization issues #598
• fix: Update relative file paths in README files by @shangxueink in fix: Update relative file paths in README files #597
• refactor(core,adapter-dify): improve variable handling and remove redundant additional_kwargs by @xiaolan20020118-create in refactor(core,adapter-dify): improve variable handling and remove redundant additional_kwargs #601
• [Feature] Enhance command tool with improved normalization and blacklist support by @dingyi222666 in [Feature] Enhance command tool with improved normalization and blacklist support #599
• [Feature] Improve agent system with infinite context and better prompts by @dingyi222666 in [Feature] Improve agent system with infinite context and better prompts #602
• [Fix] Add error handling for image reading failures by @dingyi222666 in [Fix] Add error handling for image reading failures #603
• [Feature] Add priority support to message transform interceptors by @dingyi222666 in [Feature] Add priority support to message transform interceptors #604
• [Refactor] Refactor MCP commands to use core command pattern by @dingyi222666 in [Refactor] Refactor MCP commands to use core command pattern #605
• [Fix] Improve Gemini schema generation and MCP tool display by @dingyi222666 in [Fix] Improve Gemini schema generation and MCP tool display #610
• [Fix] Improve MCP tool handling and timeout support by @dingyi222666 in [Fix] Improve MCP tool handling and timeout support #613
• [Fix] Correct MCP timeout unit description by @dingyi222666 in [Fix] Correct MCP timeout unit description #614
• [Fix] Improve crypto imports and Milvus metadata handling by @dingyi222666 in [Fix] Improve crypto imports and Milvus metadata handling #615
• [Feature] 支持 mcp 使用代理 by @Hoshino-Yumetsuki in [Feature] 支持 mcp 使用代理 #618
• [Fix] Fix Hippo and EMGAS long-term memory retrieval issues by @dingyi222666 in [Fix] Fix Hippo and EMGAS long-term memory retrieval issues #619
• [Fix] Standardize modelMaxContextSize handling and bump versions by @dingyi222666 in [Fix] Standardize modelMaxContextSize handling and bump versions #620
• [Fix] Migrate token usage to response_metadata and add stream retry by @dingyi222666 in [Fix] Migrate token usage to response_metadata and add stream retry #621
• [Fix] Improve tool error handling and async operations by @dingyi222666 in [Fix] Improve tool error handling and async operations #622
• [Release] Version 1.3.0 Stable Release by @dingyi222666 in [Release] Version 1.3.0 Stable Release #625

New Contributors

• @LoserFox made their first contribution in update glm model list #527
• @weiruchenai1 made their first contribution in feat(docs): update star history chart to support dark and light themes #536
• @shangxueink made their first contribution in [Fix] Fix adapter plugin apiKeys configuration save bug #594
• @xiaolan20020118-create made their first contribution in refactor(core,adapter-dify): improve variable handling and remove redundant additional_kwargs #601

Full Changelog: v1.2.1...v1.3.0

---

This discussion was created from the release v1.3.0.