🚀 Introducing EVA AI – A Deeply Integrated AI Assistant for Nextcloud

I have no support/technical question and have seen the support category. (Be aware that direct support questions will be deleted.)

on

Which general topic do you have

:rocket: Introducing EVA AI – A Deeply Integrated AI Assistant for Nextcloud

Hi everyone! :waving_hand:

I’d like to introduce EVA AI, an open-source AI assistant I’ve been developing for Nextcloud.

The idea behind EVA is simple: What if your AI assistant could actually work with your Nextcloud data and applications, rather than just answer questions in a separate chat window?

Over the past months, EVA has grown into a much more comprehensive AI workspace, deeply integrated into the Nextcloud ecosystem.

With the recent v1.20.42 release, I thought this would be a good opportunity to introduce the project to the wider community, share its current capabilities, and hopefully get some feedback!

:robot: What is EVA AI?

EVA is a native Nextcloud application that combines conversational AI, document search, and direct integration with other Nextcloud applications.

Instead of simply chatting with an LLM, you can ask EVA to help with everyday tasks involving your files, calendar, emails, contacts, and more.

For example, you can ask:

  • “Summarize the documents in this folder.”
  • “Find the PDF where I discussed this project.”
  • “What appointments do I have tomorrow?”
  • “Find a free time slot for a meeting next week.”
  • “Summarize my recent emails.”
  • “Create a task reminding me to finish this project.”
  • “Search the web for the latest developments on this topic.”

EVA can retrieve information, provide cited answers, and perform actions through Nextcloud’s APIs.

:sparkles: Main features

:puzzle_piece: Deep Nextcloud integration

One of my main goals has been to make EVA feel like a natural part of Nextcloud.

EVA integrates with:

  • Nextcloud Assistant: EVA provides native AI capabilities for the official Assistant, including document conversations, text processing, image-related tasks, and more.
  • Files: Search, analyze, organize, create, and manage files.
  • Calendar: Find appointments, manage events, and identify available time slots.
  • Mail: Search emails, read messages, and summarize your inbox.
  • Contacts & Tasks: Work with contacts and manage tasks.
  • Talk: Use EVA as a Talk bot or interact with conversations.
  • Unified Search: Find indexed information directly through Nextcloud’s search interface.

These integrations use Nextcloud’s APIs rather than relying on external automation platforms.

:high_voltage: More than 80 available actions

EVA currently includes over 80 actions across different services.

This allows the assistant to do more than generate text.

For example, it can retrieve information from multiple applications and help turn that information into useful actions.

Actions that modify data require user confirmation before execution.

:books: Chat with your documents (RAG)

EVA can index supported documents and make their contents searchable using a combination of semantic and keyword-based retrieval.

Supported formats include PDF, Microsoft Office documents, OpenDocument files, text files, and more.

The document interface allows you to inspect indexed files and extracted content.

Importantly, EVA can provide references to the original documents so that you can verify its answers.

:brain: Local and cloud AI models

EVA supports multiple AI providers, including:

  • Ollama
  • OpenAI-compatible APIs
  • Groq
  • Google Gemini
  • Mistral
  • DeepSeek
  • OpenRouter
  • LM Studio and other compatible local servers

You can configure different provider profiles and select models according to your requirements.

For users who prefer local inference, EVA can be configured with self-hosted models. Data sent to external providers depends on the configuration and enabled features.

:globe_with_meridians: Web search and image search

EVA can also search the web for current information.

Search results include source references, and images from the web can be displayed directly within the conversation.

:speaking_head: Additional capabilities

Other available features include:

  • Chat exports
  • Scheduled briefings
  • Reusable prompts
  • Integration with external services
  • Administration and monitoring tools

:new_button: What’s new?

The recent 1.20 release series has introduced numerous improvements across EVA’s interface, integrations, retrieval system, and backend.

Some highlights include:

Improved user experience

  • A more native Nextcloud workspace
  • Improved document and conversation interfaces
  • Better mobile layouts
  • New chat management and export capabilities

More powerful integrations

  • Expanded Nextcloud actions
  • Enhanced Nextcloud Assistant integration
  • Scheduled tasks and briefings
  • Additional voice and image capabilities

Backend and reliability improvements

  • Improvements to document indexing and retrieval
  • Additional vector storage and indexing capabilities
  • API validation and compatibility improvements
  • Expanded automated testing

The latest version, v1.20.42, also refreshes the App Store presentation, improves documentation, and adds screenshots demonstrating important features.

:locked_with_key: Privacy and self-hosting

Privacy is an important consideration for EVA.

The default document index is stored within Nextcloud’s database. Local AI providers can be used, but content sent to embedding models, cloud LLMs, web search providers, or external integrations depends on your configuration.

Administrators and users should review these data flows before enabling integrations.

EVA is open source and licensed under AGPL-3.0-or-later.

:package: Installation

EVA is available through the Nextcloud App Store.

App ID: eva_ai

The current release is v1.20.42.

Requirements:

  • Nextcloud 30–35
  • PHP 8.2+

After installation, open EVA’s settings to configure your preferred AI provider and the features you want to use.

Source code, documentation, installation instructions, and release downloads are available in the GitHub repository:

https://github.com/SchBenedikt/nextcloud-ai

:speech_balloon: Feedback welcome!

EVA is an actively developed community project, and I’d really appreciate feedback from other Nextcloud users and developers.

Bug reports, feature requests, and contributions are very welcome!

If you decide to give EVA a try, I’d love to hear about your experience.

Thanks for reading :blue_heart:

schBenedikt

What is EVA actually offering that the Nextcloud AI assistent cannot already do?
I regularly use this prompt when showing off Nextcloud AI and it works:

Create a folder called Demo_Anzers in the Demos folder.
Summarize the website www.anzers.nl in one sentence and put that information in a document in that folder.
Also add a paragraph explaining what Anzers is doing in the area of sovereignty.
Share it with demo2@anzers.nl, including a request to upload files. Make sure the email contains a link to the folder. Also create a Talk conversation with demo2@anzers.nl about this topic.

Hello,

Of course, the native Nextcloud AI has many great features, but the first thing that bothers me is the relatively complex setup, especially for the Context Backend. With EVA-AI, you install the app, connect an AI model, and everything works flawlessly.

  1. Indexing Documents
    With the help of AI, all documents in Nextcloud can be automatically indexed, and the AI can respond based on the file contents it has learned. The same works with Nextcloud Talk messages and emails.

  2. Nextcloud Talk
    EVA can automatically respond in Nextcloud Talk conversations when addressed or asked a question.
    She can also proactively write Nextcloud Talk messages on her own behalf.

  3. Web Search + Images
    EVA can actively access the internet, search for information online, and retrieve and display images.

  4. Automations
    EVA can automatically conduct briefings or perform other actions.

  5. Connectors
    Any API can be connected to EVA; she learns about them and can then use them (e.g., Immich → embed images; TrueNas → server stats); the AI can be expanded with any number of extensions.

  6. Native Integration
    EVA integrates natively into Nextcloud AI and can also be used with the existing Nextcloud AI interface.

Of course, it also includes features for creating calendar events, managing contacts, and more…

I hope I was able to give you a brief overview!

which local models would you recommend? are small models (3B-8B) enough?

Hi,
Sure, you can try out small models like that, but as the answer explains, they won’t be very accurate. Especially when the AI uses tools (for web searches, etc.), a model like that probably won’t be enough. For me, gemma4:31b works relatively well, but smaller models in particular won’t perform as well because of the tools, etc. … You can just give it a try!