v0.11.x

Project Release Notes

---

📢 Summary

This update brings significant functional enhancements, focusing on the high customization capabilities of the Deep Research API, session management optimization, and a comprehensive upgrade of the user interaction experience. By introducing a custom prompt mechanism, developers and advanced users can now deeply customize the AI's generation logic. Additionally, the inclusion of session import/export and global shortcut support significantly improves workflow efficiency.

---

🚀 New Features

1. Deep Research Prompt Customization (Prompt Overrides)

This is the core feature of this update, giving users full control over AI-generated content.

• Configuration Interface: Introduced promptOverrides in the Config interface, allowing users to override built-in prompt templates.
• API Support: The SSE API routes now support parsing prompt override parameters, integrated with strict validation and error-handling mechanisms.
• Full-Chain Integration: KnowledgeGraph, FinalReport, and the useDeepResearch hook have all been adapted to support custom prompts.
• Dynamic Generation: Enhanced prompt generation functions to support dynamic content construction based on user-defined templates.
• Settings Interface: Added a JSON input area in the settings panel, featuring real-time validation and error feedback for prompt configurations.

2. Session and Data Management

• Topic Management Optimization: Introduced the maxCollectionTopics setting, allowing users to customize the maximum number of topics per collection to better organize long-term research.

3. Search and Preferences

• Domain Filters: Added domain filtering options to the search functionality for more precise source screening.
• Report Preferences: Added new report preference options in settings to meet diverse report generation requirements.

4. Interaction Experience Enhancements

• Workflow Progress Bar: Implemented a brand-new Workflow Progress component to display the execution status of research tasks in real-time.
• Global Shortcuts: Implemented "Submit" shortcuts (e.g., Ctrl/Cmd + Enter) across core components to accelerate the operational flow.
• Multi-language Support: Updated localization files with new strings related to prompt customization.

---

🛠 Technical Optimizations and Refactoring

1. Development Standards and Documentation

• Added AGENTS.md: Clearly defined the project's coding practices and development guidelines to improve team collaboration efficiency and code quality.

2. Component Refactoring

• Topic Component Refactoring: Optimized form structures and resource handling logic, making topic editing and management more fluid.
• Hydration Optimization: Added suppressHydrationWarning to the body element to resolve common client-server hydration mismatch warnings in Next.js/React environments.

3. Shortcut Management

• Optimized shortcut management logic in the Header and search result components to ensure interaction consistency.

---

📦 Impact Scope

• API Endpoints: Added the optional promptOverrides parameter to deep-research related APIs.
• UI Interface: New "Prompt Customization" and "Max Topics" configuration items in the settings page; added import/export buttons to the top navigation bar.
• Configuration: Developers are encouraged to refer to AGENTS.md when updating their local development environments.

---

Thanks to @jamesbmour for contributing code to this project! Please provide feedback in the Issues list if you have any questions or suggestions.

---

项目版本更新说明

---

📢 版本摘要

本次更新带来了重大的功能增强，核心聚焦于 Deep Research API 的高度自定义能力、会话管理优化以及用户交互体验的全面提升。通过引入自定义提示词机制，开发者和高级用户现在可以深度定制 AI 的生成逻辑；同时，新增的会话导入导出与快捷键支持，显著提升了工作流效率。

---

🚀 新增特性

1. Deep Research 提示词自定义 (Prompt Overrides)

这是本次更新的核心功能，赋予了用户对 AI 生成内容的完全控制权。

• 配置接口： 在 Config 接口中引入 promptOverrides，支持用户自定义内置提示词模板。
• API 支持： SSE API 路由现已支持解析提示词覆盖参数，并集成了严格的校验与错误处理机制。
• 全链路集成： KnowledgeGraph（知识图谱）、FinalReport（最终报告）及 useDeepResearch 钩子均已适配自定义提示词。
• 动态生成： 增强了提示词生成函数，支持根据用户定义的模板动态构建内容。
• 设置界面： 在设置面板中新增 JSON 输入区域，支持实时校验提示词配置并提供错误反馈。

2. 会话与数据管理

• 主题管理优化： 引入 maxCollectionTopics 设置，允许用户自定义集合主题的最大数量，优化长效研究的主题组织。

3. 搜索与偏好设置

• 领域过滤器： 搜索功能新增域名过滤选项，支持更精准的资料来源筛选。
• 报告偏好： 设置中新增报告偏好选项，满足不同场景下的报告生成需求。

4. 交互体验提升

• 工作流进度条： 实现全新的 Workflow Progress 组件，实时展示研究任务的执行状态。
• 全局快捷键： 在多个核心组件中实现了“提交”快捷键（如 Ctrl/Cmd + Enter），加速操作流。
• 多语言支持： 更新本地化文件，新增了与提示词自定义相关的多语言词条。

---

🛠 技术优化与重构

1. 开发规范与文档

• 新增 AGENTS.md： 明确定义了项目的代码实践与开发指南，旨在提升团队协作效率与代码质量。

2. 组件重构

• Topic 组件重构： 优化了表单结构，提升了资源处理逻辑，使主题编辑与管理更加流畅。
• Hydration 优化： 在 body 元素中添加了 suppressHydrationWarning，解决了 Next.js/React 环境下常见的客户端-服务端水合不一致警告。

3. 快捷键管理

• 优化了 Header 和搜索结果组件中的快捷键管理逻辑，确保交互的一致性。

---

📦 影响范围

• API 接口： deep-research 相关 API 新增可选参数 promptOverrides。
• UI 界面： 设置页面新增“提示词自定义”与“主题上限”配置项；顶部导航栏新增导入导出按钮。
• 配置文件： 建议开发者参考 AGENTS.md 更新本地开发环境。

---

感谢 @jamesbmour 对本项目的代码贡献 ！如有任何问题或建议，请在 Issue 列表中反馈。

v0.10.x

feat: Using file attachment to resolve file context exceeding model limit (Need to be manually enabled in experimental mode)

---

feat: 使用文件附件突破模型上下文限制（需要在实验性中手动开启）

v0.9.x Support API and MCP

Key updates in this release focus on API, MCP services, knowledge graphs, search functionality, and PWA support:

• API Support:

  • In-Depth Research API: The In-Depth Research API has been developed, supporting SSE requests and Live mode. Live mode allows for a direct experience of the entire in-depth research process.
  • MCP Service Support: An In-Depth Research MCP service has been built, supporting the individual execution of each research step.

• Smart Image Referencing: Research reports now support smart image referencing from source materials.

• New Knowledge Graph: Knowledge graph generation is now supported, visualizing your research findings.

• PWA Support: Added support for PWA, enabling installation as a Web App.

• Knowledge Base File Chunk Upload: Knowledge base file uploads now support chunk processing, preventing data loss.

• Enhanced Search Functionality:

  • Image Search Support: Search now supports returning image results (excluding model-built-in and Firecrawl).
  • SearXNG Optimization: SearXNG search results now support reordering.

• Interface and Experience Enhancements:

  • Experimental Features Settings Page: A new experimental features settings page has been added.
  • Optimized Markdown Output: Optimized Markdown text output with added text display animation effects.
  • Optimized Markdown Image Layout: Optimized image layout within Markdown.
  • Image Viewer: Added a Lightbox component for easy image viewing.
  • Optimized Settings Interface Layout: Optimized the layout of the settings interface.

• Bug Fixes:

  • Fixed SearXNG Issue: Fixed an issue that prevented SearXNG from making requests correctly on the front-end page.

This update is dedicated to providing a more powerful and intelligent research and knowledge management experience. Thank you for your support!

---

本次更新重点围绕 API、 MCP 服务、知识图谱、搜索功能和 PWA 支持：

• 支持 API：

  • 深度研究 API： 完成了深度研究 API 的开发，支持 SSE 请求和 Live 模式，Live 模式可以直观体验的深度研究的全过程。
  • 支持 MCP 服务： 构建深度研究 MCP 服务，并支持单独执行每个研究步骤。

• 智能引用图片： 研究报告目前支持从资料中智能引用图片。

• 新增知识图谱： 支持知识图谱生成，可视化您的研究成果。

• 支持 PWA： 增加对 PWA 支持，可作为 Web App 安装使用。

• 知识库文件分块上传： 知识库文件上传现在支持分块处理，避免文件数据丢失。

• 搜索功能增强：

  • 图片搜索支持： 搜索现在支持返回图片结果（除模型内置和 Firecrawl 外）。
  • SearXNG 优化： SearXNG 搜索结果支持重新排序。

• 界面和体验优化：

  • 试验性功能设置页面： 新增试验性功能设置页面。
  • 优化 Markdown 输出： 优化了 Markdown 文本输出，并增加文本显示动画效果。
  • Markdown 图片布局优化： 优化了 Markdown 中图片的布局。
  • 图片查看器： 新增 Lightbox 组件，方便查看图片。
  • 设置界面布局优化： 优化了设置界面的布局。

• Bug修复：

  • 修复 SearXNG 问题： 修复了 SearXNG 在前端页面无法正常请求的问题。

本次更新致力于提供更强大、更智能的研究和知识管理体验，感谢您的支持！

---

Supports the generation of research reports with both pictures and texts

Supports the generation of knowledge graphs and visualization of research results

v0.8.x Major Upgrade

This update brings significant enhancements to the Magicdown editor, along with comprehensive upgrades to research, knowledge base, and resource management functionalities!

• Magicdown Editor Unleashed!

  • Effortless Editing: Supports plugins like tooltips and slash commands, making content editing easier than ever.
  • Real-time Preview: Supports real-time preview of editing results, offering a WYSIWYG (What You See Is What You Get) experience.
  • Enhanced Code Blocks: Magicdown code blocks now support copy functionality.

• In-Depth Optimization of Research Functionality!

  • Scheme-Based Data Collection: Added report schemes before data collection, enabling data gathering based on these schemes.
  • Source Citation Marking: Supports marking citation sources, enhancing report credibility.
  • Report Optimization: Optimized the prompts for final reports and reduced the number of citations.
  • Data Collection Process Supports Re-Research: Data collection process now supports re-research, providing greater flexibility.
  • Research History and Knowledge Base Search: Research history and knowledge base now support keyword search.
  • History Record Import and Export: Supports history record import and export functions.

• Knowledge Base & Resource Management Upgrade!

  • Local Resource Upload: Supports uploading local resources to the knowledge base.
  • Information Gathering and Reports Added to Knowledge Base: Information gathering results and final reports can now be added to the knowledge base.
  • Knowledge Base Page: Added a new knowledge base page for unified knowledge management.
  • Web Crawler Component: Added a new web crawler component to facilitate grabbing web content.

• Experience Optimization & Bug Fixes!

  • Interface Optimization: Page adaptation for large screen sizes, optimizing the project layout experience. Optimized the experience of using the toolbar on desktop browsers.
  • Scroll Optimization: Optimized the page scrolling experience for research history and knowledge base.
  • History Record UI Optimization: Optimized the UI of history records.

• Technical Adjustments:

  • Temporary Replacement of Milkdown Editor: Due to serious rendering performance issues with MilkdownEditor, this component has been temporarily removed and replaced with the Magicdown component.

This update is dedicated to providing a more powerful, easier-to-use, and more stable research and knowledge management experience. Thank you for your support!

---

本次更新带来了 Magicdown 编辑器的大幅增强，以及研究、知识库和资源管理功能的全面升级！

• Magicdown 编辑器大爆发！

  • 便捷编辑： 支持 tooltip 和 slash 等插件，内容编辑更加轻松。
  • 实时预览： 支持实时预览编辑结果，所见即所得。
  • 代码块增强： Magicdown 代码块支持复制功能。

• 研究功能深度优化！

  • 基于方案的数据收集： 在收集数据之前增加报告方案，并基于方案进行数据收集。
  • 引用来源标记： 支持标记引用来源，增强报告可信度。
  • 报告优化： 优化了最终报告的提示，并减少引用数量。
  • 数据收集流程支持重新研究： 数据收集过程支持重新研究，更灵活。
  • 研究历史和知识库搜索： 研究历史和知识库现在支持关键词搜索。
  • 历史记录导入导出： 支持历史记录导入和导出功能。

• 知识库 & 资源管理升级！

  • 本地资源上传： 支持上传本地资源到知识库。
  • 信息收集和报告添加到知识库： 信息收集结果和最终报告都支持添加到知识库。
  • 知识库页面： 新增知识库页面，统一管理知识。
  • 网页爬虫组件： 新增网页爬虫组件，方便抓取网页内容。

• 体验优化 & Bug 修复！

  • 界面优化： 针对大屏幕尺寸进行了页面适配，优化了项目布局体验。优化了桌面浏览器上使用工具栏的体验。
  • 滚动优化： 优化了研究历史和知识库的页面滚动体验。
  • 历史记录 UI 优化： 优化了历史记录的 UI。

• 技术调整：

  • Milkdown 编辑器临时替换： 由于 MilkdownEditor 存在严重的渲染性能问题，暂时移除该组件，并替换为 Magicdown 组件。

本次更新致力于提供更强大、更易用、更稳定的研究和知识管理体验，感谢您的支持！

v0.7.x Major Updates

This update brings massive model support, powerful search functionality, and more flexible configuration options!

• Model Selection Explosion!

  • New Model Support: Supports more mainstream large models such as OpenAI, Anthropic, DeepSeek, Grok, as well as platforms such as OpenRouter and Ollama.
  • Custom Model List: Supports custom model lists to create your own exclusive model library.

• Search Functionality Upgrade!

  • Support network search: Support search engines such as Tavily, Firecrawl, Bocha, Searxng, etc., so that LLMs that do not support search can use the network search function more conveniently.
  • Server Proxy Search API: Supports server proxy search API to improve search stability.
  • Multi-Key Polling: Search API now supports multi-key polling to improve availability.

• Writing Experience Optimization!

  • Write Requirements First, Then Write: Supports writing requirements first and then writing, which is more in line with the creative process.
  • Continue Writing Function: Added a continue writing function to keep creation uninterrupted.

• More Flexible Configuration!

  • Independent Settings: Set up each model and search engine separately for fine-grained configuration.
  • Disable Options: Supports server-side disabling of AI Providers and search Providers for flexible control of functions.
  • Support no password mode: Support server-side support for password-free mode (not recommended).
  • Reset Settings Button: Added a reset settings button to restore default configurations with one click.

• Interface and Performance Optimization:

  • Report Layout Hint Optimization: Optimized report layout hints.
  • Mermaid Rendering Hints: Added Mermaid rendering rule hints.

• Underlying Architecture Upgrade:

  • Server Proxy Reconstructed: Reconstructed server proxy forwarding mode, using middleware to rewrite Headers.
  • API Adjustments: Removed all files under the api directory, using rewrites to replace the API.

• Other Improvements:

  • Improved i18n Translation
  • Added 404 Page (Not Found Page)
  • Improved Project Documentation

This update is dedicated to providing a more powerful, more flexible, and more stable user experience. Thank you for your support!

---

本次更新带来了海量模型支持、强大的搜索功能、以及更灵活的配置选项！

• 模型选择大爆炸！

  • 新增模型支持： 支持 OpenAI、Anthropic、DeepSeek、Grok 等更多主流大模型，以及 OpenRouter、Ollama 等平台。
  • 自定义模型列表： 支持自定义模型列表，打造专属模型库。

• 搜索功能再升级！

  • 网络搜索： 支持 Tavily、Firecrawl、Bocha、Searxng 等搜索引擎，可以让不支持搜索的 LLM 可以更便捷地使用网络搜索功能。
  • 服务器代理搜索 API： 支持服务器代理搜索 API，提升搜索稳定性。
  • 多 Key 轮询： 搜索 API 现在支持多 Key 轮询，提高可用性。

• 写作体验优化！

  • 先写需求再写作： 支持先写需求再进行写作，更符合创作流程。
  • 继续写作功能： 新增继续写作功能，让创作不间断。

• 配置更灵活！

  • 独立设置： 为每个模型和搜索引擎单独设置，实现精细化配置。
  • 禁用选项： 支持服务端禁用 AI Provider 和搜索 Provider，灵活控制功能。
  • 支持无密码模式： 支持服务器端无访问密码模式（不推荐）。
  • 重置设置按钮： 增加重置设置按钮，一键恢复默认配置。

• 界面与性能优化：

  • 报告排版提示优化： 优化报告排版提示。
  • Mermaid 渲染提示： 增加 Mermaid 渲染规则提示。

• 底层架构升级：

  • 服务器代理重构： 重构服务器代理转发模式，使用中间件重写 Headers。
  • API 调整： 移除 api 目录下所有文件，使用 rewrites 代替 API。

• 其他改进：

  • 改进 i18n 翻译
  • 新增 404 页面 (Not Found Page)
  • 改进项目文档

本次更新致力于提供更强大、更灵活、更稳定的使用体验，感谢您的支持！

v0.6.6

This update primarily improves user experience and model management:

• Personalized Settings Persistence: The system can now remember your language and theme settings, eliminating the need to reconfigure them upon subsequent use.
• Model List Categorization and Archiving: The model list has been categorized and archived, providing a clearer understanding of model types and facilitating the selection of appropriate models.

---

本次更新主要改进了用户体验和模型管理：

• 个性化设置持久化： 系统现在可以记住您的语言和主题设置，下次使用时无需重新设置。
• 模型列表分类归档： 对模型列表进行了分类和归档，让模型类型更加清晰，方便您选择合适的模型。

v0.6.x Major Upgrade

This update focuses on significant enhancements to the research function and editor experience:

• Research Functionality Fully Upgraded!

  • Canvas: Supports editing of research content, with two editing modes: WYSIWYM and Markdown. It is possible to adjust the reading level, article length and full text translation.
  • Research Progress Retention: Supports retaining current research progress, allowing you to continue research at any time, and delete unwanted research content.
  • Research Record Revision: Supports revision of previous research records.
  • Research Report Format Optimization: Research reports now have better typesetting.
  • Mermaid & Math Formula Support: The Editor component now supports rendering mermaid flowcharts and mathematical formulas.
  • Information Collection Section Enhanced: Refactored the information collection section, separating editable and non-editable areas, supporting content export to Markdown, and allowing multiple pages to be expanded simultaneously.
  • File Export Improvements: Fixed the issue of missing file extensions when exporting Markdown files, and optimized the export function for PDF and Markdown files.

• Other Optimizations:

  • Model Management: Added the useModel hook to get the latest model list and support custom "Thinking Model" and "Networked Model".
  • Support Multi-Key payload: Support Multi-Key payload to improve API response efficiency.
  • Performance Optimization: Optimized the editor to avoid performance issues caused by multiple renderings. Optimized the timer to avoid instantaneous accumulation. Optimized the logic for starting new research.
  • Documentation Improvements: Improved i18n translation.

We are committed to providing a more powerful and smoother research and editing experience. Thank you for your support!

---

本次更新主要围绕研究功能和编辑器体验进行了大幅度增强：

• 研究功能全面升级！

  • Canvas： 支持编辑研究内容，提供 WYSIWYM 和 Markdown 两种编辑模式，可调整阅读级别、文章长度和全文翻译。
  • 研究进度保留： 支持保留当前研究进度，可以随时继续研究，并且可以删除不需要的研究内容。
  • 研究记录修订： 支持对之前的研究记录进行修订。
  • 研究报告格式优化： 研究报告现在拥有更好的排版格式。
  • Mermaid & Math 公式支持： 编辑器组件现在支持渲染 mermaid 流程图和数学公式。
  • 信息收集板块功能增强： 重构了信息收集部分，将可编辑区域和不可编辑区域分开，内容支持导出 Markdown，并可以同时展开多个页面。
  • 文件导出改进： 修复了导出 Markdown 文件时缺少文件后缀的问题，并优化了 PDF 和 Markdown 文件的导出功能。

• 其他优化：

  • 模型管理： 新增 useModel hook，用于获取最新的模型列表，并支持自定义“思考模型”和“联网模型”。
  • 支持多 key 负载：支持多 key 负载，提高 API 响应效率。
  • 性能优化： 优化了编辑器，避免多次渲染导致性能问题。优化了计时器，避免瞬间累积。 优化了启动新研究的逻辑。
  • 文档改进： 改进了i18n 翻译。

我们致力于提供更强大、更流畅的研究和编辑体验，感谢您的支持！

v0.5.2

This update includes the following improvements:

• Added research history feature: Allows you to easily review and manage your previous research results.
• Fixed PDF layout issue: Resolved the issue where PDF files could be generated with excessively narrow layouts.

Thanks to raymond9zhou for contributing the core code implementation of this update: deep-research/pull/11

---

本次更新带来以下改进：

• 新增研究历史记录功能: 方便您回顾和管理之前的研究结果。
• 修复PDF布局异常问题: 解决了PDF文件在生成时可能出现的布局宽度过小问题。

感谢 raymond9zhou 贡献了本次更新的核心代码实现: deep-research/pull/11

v0.5.0

This update primarily focuses on report generation and research process optimization:

• Optimized Research Process: The research process has been optimized to support controlled and unlimited research, making research more flexible.
• Automatic Report Title Generation: Report titles can now be generated automatically, saving you time.
• Project Logo Update: We've updated the project logo for a fresh new look.

---

本次更新主要集中在报告生成和研究流程优化：

• 优化研究流程: 优化了研究过程，支持可控和无限研究，让研究更加灵活。
• 报告标题自动生成: 现在可以自动生成报告标题，节省您的时间。
• 项目Logo更新: 我们更新了项目Logo，焕然一新。