介绍

AnythingLLM是一个全栈应用程序，您可以使用现成的商业大语言模型或流行的开源大语言模型，再结合向量数据库解决方案构建一个私有ChatGPT，不再受制于人：您可以本地运行，也可以远程托管，并能够与您提供的任何文档智能聊天。

AnythingLLM将您的文档划分为称为workspaces (工作区)的对象。工作区的功能类似于线程，同时增加了文档的容器化，。工作区可以共享文档，但工作区之间的内容不会互相干扰或污染，因此您可以保持每个工作区的上下文清晰。

软件截图

软件特点

多用户实例支持和权限管理
工作区内的智能体Agent（浏览网页、运行代码等）
为您的网站定制的可嵌入聊天窗口
支持多种文档类型（PDF、TXT、DOCX等）
通过简单的用户界面管理向量数据库中的文档
两种对话模式：聊天和查询。聊天模式保留先前的对话记录。查询模式则是是针对您的文档做简单问答
聊天中会提供所引用的相应文档内容
100%云部署就绪。
“部署你自己的LLM模型”。
管理超大文档时高效、低耗。只需要一次就可以嵌入（Embedding)一个庞大的文档或文字记录。比其他文档聊天机器人解决方案节省90%的成本。
全套的开发人员API，用于自定义集成！

更新日志

Add Support for NVIDIA NIM by @timothycarambat in #2766
update chat model filters for openai by @timothycarambat in #2803
Fixed a typo in README by @logscore in #2805
Fix scraping failed bug in link/bulk link scrapers by @shatfield4 in #2807
Fix timestamp bug in embed widget by @shatfield4 in #2806
Redirect to default thread on thread deletion by @shatfield4 in #2808
Add support for hiding sidebar by @timothycarambat in #2809
Add vector search API endpoint by @timothycarambat in #2815
patch ZH typo by @timothycarambat in #2816
add Gitlab to watchable documents by @timothycarambat in #2817
Fix(2753): Replace React Router Navigate with window.location for SSO state refresh by @andydev404 in #2818
Docker Scout CI error by @timothycarambat in #2820
APIPie LLM provider improvements by @shatfield4 in #2695
Add thread name to workspaces get endpoint by @shatfield4 in #2822
Add new workspace user management endpoint by @timothycarambat in #2842
Add attachments to GenericOpenAI prompt by @wolfganghuse in #2831
Purge cached docs and remove docs from all workspaces on vectorDB/embedder changes by @shatfield4 in #2819
LLM performance metric tracking by @timothycarambat in #2825
Append writable fields to dev API new workspace endpoint by @shatfield4 in #2843
Enable default roles to be able to use attachments but not upload documents by @timothycarambat in #2860
Add support for gemini authenticated models endpoint by @timothycarambat in #2868
Remove fine-tuning flow by @timothycarambat in #2872
fix wrong metadata assignment in MilvusProvider by @wolfganghuse in #2870
update ollama performance mode by @timothycarambat in #2874
Fix light mode table colors by @shatfield4 in #2873
Fix styles of ChangeWarningModal by @shatfield4 in #2876
Show scrollbar on rendered tables in chat by @shatfield4 in #2884
Fix typo in frontend for SearXNG agent setup by @shatfield4 in #2917
Normalize paths on files uploaded to prevent arbitrary file writes by @shatfield4 in #2905
Patch unauthorized access to other user's pfps by @shatfield4 in #2904
Audio file validations by @shatfield4 in #2902
Patch custom models endpoint by @shatfield4 in #2903
Add support for Google Generative AI (Gemini) embedder by @007900 in #2895
Return timestamps in OpenAI compat endpoints in proper format by @timothycarambat in #2921
Update README.md by @KPCOFGS in #2924
make similarityResponse object arguments and not positional by @timothycarambat in #2930
Reranker option for RAG by @timothycarambat in #2929
Update example ENV files by @shatfield4 in #2931
feat: add support for voyage-3-large and voyage-code-3 embedding models by @root-reindeer-flotilla in #2955
Add caching to Gemini /models by @timothycarambat in #2969
Support historical message image inputs/attachments for n+1 queries by @timothycarambat in #2919
Swap to GH arm64 runner by @timothycarambat in #2980
Fix similarity score bug in lance/chroma dbs by @shatfield4 in #2986
Farsi (Persian) translations by @okaeiz in #2989
AI agent ui animation by @timothycarambat in #2999
Display thinking in the prompt response - model agnostic by @timothycarambat in #3001
Fix stop generation button UI by @shatfield4 in #3011
Bump perplexity models by @shatfield4 in #3014
Dynamic fetching of TogetherAI models by @shatfield4 in #3017
Add Version to AzureOpenAI by @timothycarambat in #3023
Allow editing of LLM settings from workspace settings by @shatfield4 in #2865
feat: update novita AI logo and default model by @jasonhp in #3037
[Chore] Add better data-handling for unknown providers by @timothycarambat in #3026
breakout latex plugin for delims by @timothycarambat in #3040
Remove native LLM option by @shatfield4 in #3024
Add ability to disable default agent skills by @timothycarambat in #3041