Home Assitant and NOMAD

[greyson1973]

Hello Project NOMAD Team,

I'm planning a home-server infrastructure that will include both Project NOMAD and Home Assistant. Before purchasing and configuring the hardware, I'd like to clarify a few technical questions about integrating the two systems.

There is a personal reason this project is particularly important to me.

I live in an area that is frequently threatened by severe hurricanes. During an extended power outage or other emergency, access to reliable information can literally become a matter of life and death. Project NOMAD's offline knowledge library could be an incredibly valuable resource when internet access and other sources of information are unavailable.

There is also an important accessibility consideration. My wife is disabled and uses a wheelchair. I want to make this information accessible to her through Home Assistant's voice assistant, rather than requiring her to reach a computer, operate a keyboard and mouse, or depend on being able to see a screen.

Ideally, she could simply speak to Home Assistant from anywhere in the house where voice access is available and ask for the information she needs. In an emergency, that could make an enormous difference in how quickly and independently she can find instructions and respond to a situation.

My goal is to have Home Assistant serve as the household voice assistant, while Project NOMAD provides its offline knowledge and research capabilities.

I would like to accomplish this using a single Ollama server rather than running separate LLM instances on both machines.

1. Remote Ollama Configuration

I understand that Project NOMAD supports connecting to a remote Ollama server.

Can Project NOMAD operate fully with its local Ollama installation disabled and connect exclusively to an Ollama server running on another computer?

Does this remote configuration preserve all Project NOMAD AI functionality, including access to its offline knowledge base?

Are there any limitations when multiple applications use the same Ollama server simultaneously?

Does Project NOMAD require a particular model, model size, embedding model, or model configuration?

Can the same Ollama model serve both Home Assistant and Project NOMAD without interfering with either application's operation?

2. Project NOMAD Knowledge-Base API

This is probably my most important area of inquiry.

I would like Home Assistant to be able to request information from Project NOMAD's existing knowledge base.

Does Project NOMAD expose an API that allows external applications to submit questions and retrieve relevant information from its knowledge base?

If so, is this API documented and supported for third-party integrations?

Can an external application submit a natural-language question and receive the same retrieval-augmented response that Project NOMAD provides through its own interface?

Can the API return source documents, citations, and references along with the answer?

Can external applications access specific collections or content categories?

Is authentication required? Are API keys, tokens, or other access controls supported?

Does the API support concurrent requests?

I understand Project NOMAD uses Qdrant for knowledge storage. Would you recommend accessing the knowledge base through a Project NOMAD API rather than querying Qdrant directly?

3. Home Assistant Integration

My intended arrangement is for Home Assistant to be the single voice entry point.

For example:

"OK Nabu, turn on the living room lights."

Home Assistant recognizes the request and controls the lights.

"OK Nabu, explain how a diesel engine works."

Home Assistant sends the question to the Project NOMAD knowledge base and presents the retrieved information.

"OK Nabu, find me instructions for treating a severe bleeding injury."

Project NOMAD supplies its first-aid knowledge, while Home Assistant independently initiates an appropriate emergency automation.

My questions:

Has anyone successfully integrated Project NOMAD's knowledge base with Home Assistant's Assist conversation agent?

Is there an existing integration, custom component, or recommended method for accomplishing this?

Could Project NOMAD's knowledge-search API be exposed to Home Assistant as an LLM tool?

Does Project NOMAD support an MCP server or another tool interface that could make this integration easier?

Are there any plans for an official Home Assistant integration?

4. Voice Assistant and AI Responsibilities

I do not necessarily want Project NOMAD to operate as a second independent voice assistant.

Instead, I want Home Assistant to handle wake-word detection, speech recognition, and voice responses, with the central Ollama model deciding whether a request requires Home Assistant functionality, Project NOMAD knowledge retrieval, or both.

Is this arrangement compatible with Project NOMAD's architecture?

Can Project NOMAD provide knowledge retrieval without independently receiving microphone audio or managing voice conversations?

Can an external application invoke Project NOMAD's knowledge-retrieval functionality without opening its web interface?

Are there any architectural considerations that would make this arrangement impractical?

5. Hardware and Resource Requirements

My proposed arrangement uses separate physical computers:

Computer 1: Home Assistant OS and Ollama.

Computer 2: Ubuntu Desktop 26.04 LTS running Project NOMAD, including its knowledge base and supporting services.

Is there any reason Project NOMAD's knowledge retrieval and Qdrant services must run on the same computer as Ollama?

What network bandwidth or latency would you recommend between the two machines?

Does using a remote Ollama server introduce any significant performance limitations?

Are there any additional resource requirements when Project NOMAD uses a remote Ollama server?

6. Documentation and Development Resources

Finally:

Is there additional API documentation beyond the documentation available through the Project NOMAD interface?

Is there a developer community or discussion channel where integrations like this are being developed?

Would your team be willing to offer guidance on the recommended integration architecture before I begin implementation?

I want to emphasize that I am not looking to replace Project NOMAD's interface or functionality. I want to use it as the household's offline knowledge resource while allowing Home Assistant to remain the central household assistant.

Ideally, one Ollama installation would serve both applications, with Home Assistant routing requests to its own automation capabilities, Project NOMAD's knowledge base, or both as appropriate.

I appreciate your time and the work you've put into Project NOMAD. I'd rather ask these questions now than purchase hardware and discover that my intended architecture isn't supported.

Thank you!

[KC1SSN]

This sounds exactly like what I was looking for. I watched a YouTube video about an Offline AI Jarvis build and was thinking this would be a really cool integration into Project NOMAD. Looks pretty straightforward but I haven't attempted and thought about posting something here for someone smarter than I to attempt. Here is the video on YouTube, there are several but this one was for an Offline build. https://youtu.be/BgpRjk4Yoqk?si=f7tS3BjAC2e7lNbl