Releases: netease-youdao/QAnything
Releases · netease-youdao/QAnything
v2.0.0
🎉🎉🎉QAnything v2.0.0 发布!
全新发布v2.0.0
新特性
合并Docker与Python版-qanything-v2分支
- 前端大更新：支持更新信息展示，全部问答及解析参数配置，解析结果预览及编辑
- 信息展示包括：上传文件进度，解析各阶段耗时，失败原因，问答各阶段耗时，问答日志，数据统计等等
- 解析效果提升：xlsx，pdf，docx，html等均有优化，详见新旧解析效果对比
- 检索效果提升：metadata信息嵌入，chunks切片逻辑优化，检索结果自动融合并排序
- 支持快速开始模式：类似kimi，文件即传即问，支持无文件问答
- 支持仅检索模式：仅返回检索结果，不调用大模型
- 上传文件逻辑优化，解析文件与问答请求独立处理
- Bot指令遵循能力优化
- 支持多对话窗口
- 启动方式优化，不再依赖脚本，改为docker compose命令行启动
- Docker镜像大小优化，mac下仅有4GB，linux/windows下9GB
- windows可直接启动，不再依赖WSL
新贡献者
Contributors
Ianarua
Assets 2
v1.4.2
🎉🎉🎉QAnything v1.4.2 发布!
全新发布v1.4.2
新特性
Python版本-qanything-python分支
- 添加更便捷的启动方法（需提供openai，ollama等LLM服务的api-key），只需要一行代码就可以自动配置conda虚拟环境并启动项目，（WIN（WSL）、Mac、LInux下所有启动脚本合并为两个：run_for_CPU.sh/run_for_GPU.sh，支持python2环境）
- 优化控制台日志输出，可以在终端输出更详细的日志
- 将一部分常用参数的值调整至根目录配置文件（config.py）下统一管理，增加更多可自定义的参数
- 将pdf格式转markdown格式的模块独立为子模块，在执行pip install -e .命令后可以在本地独立使用
- 修复联网检索模块在检索到的数据内容过长时会导致本地知识库内容无法正常获取的问题
- 优化联网检索的文档分割效果，并优化上传失败文档的错误信息展示。
- 修复了表格解析报错，rerank重复执行，opencv安装失败，rerank执行出错，单次前端问答多次调用后端问答接口等一系列bug。
新贡献者
- @ikun-moxiaofei 在 #384 进行了首次贡献（共17PR）
- @glorialucheng 在 #447 进行了首次贡献
Contributors
glorialucheng and ikun-moxiaofei
Assets 2
v1.4.1
🎉🎉🎉QAnything v1.4.1 发布!
全新发布v1.4.1
新特性
Docker版本-master分支
- 支持其他与OpenAI-API兼容的大模型服务，包含ollama，通义千问DashScope等
- 支持多卡推理
- 修复使用OpenAI-API时提示找不到whl文件的问题
Python版本-qanything-python分支
- 支持其他与OpenAI-API兼容的大模型服务，包含ollama，通义千问DashScope等
- PDF文件解析效果优化（包含表格效果）
- 修复部分Linux系统（比如centos7）glibc版本过低时无法启动的问题（onnxruntime-gpu[cuda12.x]自动切换为onnxruntime）
- 修复Mac下embedding和rerank速度异常慢的问题
- 修复部分问答无数据来源的问题
- 前端支持联网检索结果直接点击跳转
- 优化OCR，速度和效果提升；且首次启动时无需下载ocr模型，启动时间优化至60秒内
- 支持无显卡Linux机器启动纯CPU版，优化使用cpu模式或OpenAI-API接口调用大模型时的资源占用
新贡献者
Contributors
WuErPing
Assets 2
v1.4.0-python
🎉🎉🎉QAnything v1.4.0-python 发布!
全新发布qanything-python分支作为v1.4.0-python
新特性
- 支持联网搜索
- 支持FAQ问答
- 支持自定义Bot
- 支持语音文件
- 支持文件溯源预览
- 支持问答日志检索
- 支持国产OS（OpenCloudOS）
变更内容
@cunhanfeng 在 #259 增加了语音文件支持
@dliting 在 #209 更新了 README_zh.md
@xixihahaliu 在 #244 修复了已知问题
@xixihahaliu 在 #264 开发了 v1.3.1 版本
@successren 在 #297 开发了 v1.3.1 版本
新贡献者
@cunhanfeng 在 #259 进行了首次贡献
@dliting 在 #209 进行了首次贡献
完整更新日志: https://github.com/netease-youdao/QAnything/compare/v1.3.3…v1.4.0-python
🎉🎉🎉QAnything v1.4.0-python released!
New release of qanything-python branch for v1.4.0-python
What's New
- Support web search
- Support FAQ Q&A
- Support custom bot
- Support audio files
- Support file traceability preview
- Support Q&A log retrieval
- Support Chinese domestic OS (OpenCloudOS)
What's Changed
- Add audio support by @cunhanfeng in #259
- Update README_zh.md by @dliting in #209
- Fix known issues by @xixihahaliu in #244
- Develop for v1.3.1 by @xixihahaliu in #264
- Develop for v1.3.1 by @successren in #297
New Contributors
Full Changelog: v1.3.3...v1.4.0-python
Contributors
dliting, successren, and xixihahaliu
Assets 2
v1.3.3
🎉🎉🎉QAnything v1.3.3 发布!
新特性
- 解决跨域问题，加载前端静态资源即可访问页面，启动更快
- 解决ES权限问题
- mysql连接数增大到32
- 不再限制N卡型号，废弃显卡型号查表，使用pycuda查询显卡算力
- 读取excel文件中所有sheet
- 文档更新
- 其他bug修复
变更内容
- 解决跨域问题 #44 #188 #121 ，加载前端静态资源即可访问页面，启动更快 by @xixihahaliu @successren in #237
- adds documentation for debugging in IDE by @sai-krishna-raparthi in #219
- Update QAnything_Startup_Usage_README.md by @yy307 in #214
- Update QAnything_Startup_Usage_README.md by @eltociear in #210
- Read all sheets from excel file by @Ma-Dan in #200
- fixed question Comment by @denvey in #172
- 修复milvus查询到的source document中，expand doc score为字符串的问题 by @liuyijiang1994 in #166
新贡献者
- @sai-krishna-raparthi made their first contribution in #219
- @yy307 made their first contribution in #214
- @Ma-Dan made their first contribution in #200
- @denvey made their first contribution in #172
- @liuyijiang1994 made their first contribution in #166
Full Changelog: v1.3.0...v1.3.3
🎉🎉🎉QAnything v1.3.3 released!
What's New
- resolve cross-origin issues, load frontend static resources to access the page without the need to start the frontend service, resulting in faster startup.
- Address ES (Elasticsearch) permission problems.
- Increase the MySQL connection pool size to 32.
- Remove restrictions on Nvidia GPU types, abandon the use of a lookup table for GPU types, and use PyCuda to query GPU computing power.
- Read all sheets in the Excel file
- Document update
- Other bug fixes
What's Changed
- resolve cross-origin issues #44 #188 #121 , load frontend static resources to access the page without the need to start the frontend service, resulting in faster startup. by @xixihahaliu @successren in #237
- adds documentation for debugging in IDE by @sai-krishna-raparthi in #219
- Update QAnything_Startup_Usage_README.md by @yy307 in #214
- Update QAnything_Startup_Usage_README.md by @eltociear in #210
- Read all sheets from excel file by @Ma-Dan in #200
- fixed question Comment by @denvey in #172
- 修复milvus查询到的source document中，expand doc score为字符串的问题 by @liuyijiang1994 in #166
New Contributors
- @sai-krishna-raparthi made their first contribution in #219
- @yy307 made their first contribution in #214
- @Ma-Dan made their first contribution in #200
- @denvey made their first contribution in #172
- @liuyijiang1994 made their first contribution in #166
Full Changelog: v1.3.0...v1.3.3
Contributors
yy307, denvey, and 6 other contributors
Assets 2
v1.3.0
🎉🎉🎉QAnything v1.3.0 发布!
新特性
- 支持在纯Python环境中安装，可以在MAC上运行，也可以在纯CPU机器上运行。详见：纯Python环境安装教程
- 支持混合检索（BM25+Embedding）。
- 当系统命令缺失时，下载提示更清晰。
- 自动检测显卡的计算能力和内存大小，并根据情况自动调整默认启动参数，以提供用户最佳体验，并给予相关提示。
- 更新前置条件，最低支持 GTX1050Ti，支持更多显卡型号。
- 提示用户代码仓库是否为最新，仅适用于主分支。
- 优化启动流程，自动查询相关日志并在前端显示错误消息，避免连续输出 'waiting for coffee'。
- 在前端添加英文版本。
- 修复已知的错误，优化用户体验。（在 Milvus 插入失败时，请记得删除根目录下的 "volumes" 文件夹。）
变更内容
- 修复启动脚本已知问题。由 @xixihahaliu 提交于 #92
- 修复已知问题，优化用户体验。由 @xixihahaliu 提交于 #102
- Milvus增加用户名、密码、数据库名支持。由 @cklogic 提交于 #97
- feat_#114: 为 NVIDIA GeForce RTX 4090D 添加 GPU 兼容性检查。由 @johnymoo 提交于 #115
- 修复：由 @pinkcxy 提交的轮次控制问题。#131
- 问题修复：调用api/api/local_doc_qa/local_doc_chat返回数据为None。由 @leehom0123 提交于 #137
- 优化解析 csv 和 xlsx 文件的逻辑。此外，现在支持离线运行 Docker 镜像。由 @xixihahaliu 提交于 #139
- 混合搜索。由 @shenlei1020 提交于 #194
新贡献者
@pinkcxy 在 #100 中首次贡献
@cklogic 在 #97 中首次贡献
@johnymoo 在 #115 中首次贡献
@leehom0123 在 #137 中首次贡献
@shenlei1020 在 #194 中首次贡献
完整变更记录：https://github.com/netease-youdao/QAnything/compare/v1.2.0…v1.3.0
🎉🎉🎉QAnything v1.3.0 released!
What's New
- Support installation in a pure Python environment, which can run on MAC or other CPU only machine. see: Pure Python Environment Installation Guid
- Support hybrid search（BM25+Embedding）
- When the system command is missing, the download prompt is clearer.
- Automatically detect the computing power and memory size of the graphics card, and automatically adjust the default startup parameters accordingly to provide users with the best experience, and give relevant prompts.
- Update Prerequisites, minimum support GTX1050Ti, support more graphics card models.
- Prompt the user whether the code repository is up to date, only for master branch.
- Optimize the startup process, automatically query relevant logs and display error messages on the front end to avoid continuous output of 'waiting for coffee'.
- Add English version to the front end
- Fix known bugs, optimize user experience.（Remember to delete the "volumes" folder under the root directory when milvus insertion fails.）
What's Changed
- Fix known issues with the startup script. by @xixihahaliu in #92
- Fix known issues, optimize user experience. by @xixihahaliu in #102
- Develop for v1.2.1 by @pinkcxy in #100
- Milvus增加用户名、密码、数据库名支持 by @cklogic in #97
- feat_#114: Add gpu compatibility check for NVIDIA GeForce RTX 4090D by @johnymoo in #115
- fix: Number of rounds control by @pinkcxy in #131
- issue fix: Call api api/local_doc_qa/local_doc_chat return data:None by @leehom0123 in #137
- Optimizing the parsing logic for csv and xlsx files. In addition, offline running of Docker images is now supported. by @xixihahaliu in #139
- Hybrid search by @shenlei1020 in #194
New Contributors
- @pinkcxy made their first contribution in #100
- @cklogic made their first contribution in #97
- @johnymoo made their first contribution in #115
- @leehom0123 made their first contribution in #137
- @shenlei1020 made their first contribution in #194
Full Changelog: v1.2.0...v1.3.0
Contributors
johnymoo, leehom0123, and 4 other contributors
Assets 2
v1.2.0
🎉🎉🎉QAnything v1.2.0发布！
新特性
- 支持自定义LLM：
- 支持本地LLM和云LLM切换，云LLM支持openai chatgpt API。
- 支持三种LLM推理运行后端，包括FasterTransformer（默认）、huggingface和vllm。
- 通过FastChat Server API支持加载各种开源大语言模型。
- 发布了微调后的千问7B的模型：https://huggingface.co/netease-youdao/Qwen-7B-QAnything
- 修复两个安全漏洞：
- 多个函数中的SQL注入漏洞，漏洞名称：SQL注入 - CWE-89
- LocalFile类中的文件上传漏洞，漏洞名称：文件上传 - CWE-434
- 适配更多显卡：
- 在使用huggingface transformers作为LLM运行后端时，理论上支持Nvidia显卡的计算能力为6.0及以上。
- 更加用户友好：
- 将所有日志文件放在一个目录中，便于调试。
- 添加了一些启动提示信息。
- 其他更新
更新内容
- 支持QAnything中的公共LLM和OpenAI API作为LLM服务。由 @songkq 在 #78 中贡献
- 提供了QAnything启动命令和LLM接口的详细使用方法。由 @songkq 在 #78 中贡献
- 修复了SQL注入和文件名注入的安全漏洞。由 @xixihahaliu 在 #76 中贡献
- 优化日志信息显示，便于调试。由 @xixihahaliu 在 #76 中贡献
- 添加了一些启动提示信息。由 @xixihahaliu 在 #77 中贡献
- 修复: 上传失败时提示失败。由 @dongfeng3692 在 #72 中贡献
- 更新最大上传文件大小的注释 由 @dongfeng3692 在 https://github.com/netease-youdao/QAnything/pull/71中贡献
- 修复: 在windows下使用git bash运行 by @yorelog 在 https://github.com/netease-youdao/QAnything/pull/61中贡献
- linux 1. 解决网络不通导致npm安装异常问题 2. 解决swagger外网拉取慢的问题 由 @01000100GT 在 https://github.com/netease-youdao/QAnything/pull/69中贡献
新贡献者
- @xixihahaliu 在 #76 中首次贡献
- @songkq 在 #78 中首次贡献
- @mvlttt 报告了关于SQL注入和文件名注入的两个安全漏洞
- @dongfeng3692 在 #72 中首次贡献
- @01000100GT 在 #69 中首次贡献
完整变更记录: v1.1.1...v1.2.0
🎉🎉🎉QAnything v1.2.0 released!
What's New
- Custom LLMs is supported:
- Local LLMs and cloud LLMs switching is supported, cloud LLMs support openai chatgpt API.
- Supports three LLM inference runtime backends, include FasterTransformer(default), huggingface and vllm.
- Supports loading various public large language models via the FastChat Server API.
- Our finetuned Qwen 7B model is released at https://huggingface.co/netease-youdao/Qwen-7B-QAnything
- Fixed two security vulnerabilities:
- SQL injection in multiple functions，Vulnerability name: SQL injection - CWE-89
- File upload in LocalFile class, Vulnerability name: File upload - CWE-434
- Adapted for more GPUs:
- Theoretically supports Nvidia GPUs with a compute capability of 6.0 and above when using huggingface transformers as the LLM runtime backend.
- More user-friendly :
- Put all log files in one directory for easy debugging.
- Add some startup prompt messages.
- The front-end page will return results in a streaming manner.
- Other updates
What's Changed
- Support public LLMs and OpenAI API as a LLM service in QAnything. by @songkq in #78
- Provide a detailed usage of QAnything Startup Command and LLM interface. by @songkq in #78
- Fix security vulnerabilities of SQL injection and file name injection. by @xixihahaliu in #76
- Optimize log information display for easier debugging. by @xixihahaliu in #76
- Add some startup prompt messages. by @xixihahaliu in #77
- fix: display upload failure when upload fails by @dongfeng3692 in #72
- Update request max size code annotation by @dongfeng3692 in #71
- fix: running with git bash under windows by @yorelog in #61
- linux 1. Solve the problem that the npm installation is abnormal due to network failure 2. Fix the problem of slow swagger external network pulling by @01000100GT in #69
New Contributors
- @xixihahaliu made their first contribution in #76
- @songkq made their first contribution in #78
- @mvlttt report two security vulnerabilities about SQL injection and file name injection
- @dongfeng3692 made their first contribution in #72
- @01000100GT made their first contribution in #69
Full Changelog: v1.1.1...v1.2.0
Contributors
yorelog, songkq, and 4 other contributors
Assets 2
v1.1.1
What's Changed
- Update README.md by @eltociear in #35
- update readme, add email info by @successren in #48
- update SECURITY.md, add email info by @successren in #49
- faq: Ports not available error on windows by @yorelog in #52
- fix: running with git bash under windows by @yorelog in #51
- fix: Open rerank by default. by @successren in #59
New Contributors
- @eltociear made their first contribution in #35
- @successren made their first contribution in #48
- @yorelog made their first contribution in #52
Full Changelog: v1.1.0...v1.1.1
Contributors
yorelog, eltociear, and successren
Assets 2
v1.1.0
🎉🎉🎉QAnything v1.1.0发布！
新特性
- 优化启动流程以支持一键初始化：执行 "bash run.sh" 即可启动。
- 增强安装过程的交互性，提供更友好的用户提示。
- 提高显存利用率以支持多GPU部署，目前支持单卡或双卡部署，双卡设置的显存使用量分别为 *11GB 和 5GB。
- 将矢量库的体积减少到原来的三分之一。
- 优化xlsx和html文件的分块处理，防止因块过大而导致解析失败。
- 当知识库仅包含单个文档时，提高回答的有效性。
- 提高PDF解析的有效性和速度。
- 提供原始未修改的DockerFile。
变更内容
- @jsoncode 在 #7 添加了Windows平台的启动脚本
- @jsoncode 在 #11 优化了前端构建逻辑
- @jsoncode 在 #12 更新了front_end package.json
- @jsoncode 在 #14 更新了front_end配置和代理
新贡献者
完整变更日志：https://github.com/netease-youdao/QAnything/commits/v1.1.0
🎉🎉🎉QAnything v1.1.0 released!
Features
- Optimize the startup process to support one-click initialization: execute "bash run.sh" to start.
- Enhance the interactivity of the installation process, providing more user-friendly prompts.
- Improve VRAM utilization to support multi-GPU deployment, currently supporting single or dual card deployment with VRAM usage of * 11GB and 5GB for dual-card setups, respectively.
- Reduce the footprint of the vector library to one third of its original size.
- Optimize chunking of xlsx and html files to prevent parsing failure caused by oversized chunks.
- Enhance the effectiveness of answering when the knowledge base contains only a single document.
- Improve PDF parsing effectiveness and speed.
- Provide the original unmodified DockerFile.
What's Changed
- Add startup script on windows platform by @jsoncode in #7
- Optimize front-end construction logic by @jsoncode in #11
- feat: update front_end package.json by @jsoncode in #12
- update front_end config and proxy by @jsoncode in #14
New Contributors
Full Changelog: https://github.com/netease-youdao/QAnything/commits/v1.1.0
Contributors
jsoncode