NuCoreAI is a Plugin based platform for the integration and automation of anything. This model serves as teh core of NuCoreAI's local decision making engine, enabling seamless and intelligent automation/optimization of things and services, and specifically Smart Home/Building devices.

NuCoreAI is free, MIT-licensed. This model can run locally on affordable hardware with an NVIDIA RTX 3060. You can learn more about NuCoreAI at nucore.ai, including links to developer tools and a fully functional Python-based assistant.

Model Details

Model Description

This model has been fine-tuned on thousands of anonymized node, device, and profile samples, collected with permission from Universal Devices customers. It is specifically optimized to reason about:

Smart home devices and node definitions
Device properties, accepted and sent commands, and their parameters
Automation routines
Units of measure (UOMs) and associated value constraints

The finetuning has incorporated most of the system prompts into the model, and therefore the size of the prompt is dramatically reduced by over 70%. (see Prompt)

When given a structured prompt, the model can generate actionable commands, such as device control or state queries, that are seamlessly executed through the eisy | home platform. It also produces intelligent automation routines (known as programs in eisy terminology) using JSONLogic.

This model serves as the core of NuCoreAI's local decision-making engine, enabling automation, control, and optimization across a wide range of smart home and energy devices.

NuCoreAI is free, MIT-licensed, and can run locally on affordable hardware with an NVIDIA RTX 3060. You can learn more about NuCoreAI at nucore.ai, including links to developer tools and a fully functional Python-based assistant.

Prompt

The following is an optimized prompt to get the best out of nucore.11 model:

You are a NuCore smart-home assistant. You use the provided DEVICE STRUCTURE to respond to USER QUERY :

By creating one or more NuCore commands of this format: BEGIN_NUCORE_COMMAND [ { "device_id": "<DEVICE_ID>", "command_id": "<COMMAND_ID>", "command_params": [ { "id": "<PARAM_ID>", "value": , "uom": "<UOM_ID>" }, { }, ... ] }, { }, .... ] END_NUCORE_COMMAND

By querying current and live property values and statuses BEGIN_NUCORE_PROPERTY_QUERY [ { "device_id": "<DEVICE_ID>", "property_id": "<PROPERTY_ID>", "property_name": "<PROPERTY_NAME>" }, { <more properties if any }, .... ] END_NUCORE_PROPERTY_QUERY

By creating one or more NuCore routines When to act: explicit request for automation/optimization/triggers. The USER query starts with "if", "when", or contains "routine" or "program" BEGIN_NUCORE_ROUTINE [ { "routine": { "id": , "name": "", "parent": 0, "comment": "", "if": [ /* ordered flat Condition Tokens / ], "then": [ / actions / ], "else": [ / actions */ ], "reasoning": "<assistant's reasoning for the structure and semantics of this routine>" } }, { }, .... ] END_NUCORE_ROUTINE

Examples

play classical on youtube

The device "YouTube Player" supports the command "Play" with a parameter for the playlist or song to play. The parameter values include "Classical" which corresponds to the enum value [5].

Therefore, to play classical music on YouTube, the command is:

json

__BEGIN_NUCORE_COMMAND__
{
  "device_id": "n006_ytsvc",
  "command_id": "PLAY",
  "command_params": [
    {
      "id": "n/a",
      "value": 5,
      "uom": ""
    }
  ]
}
__END_NUCORE_COMMAND__

[<Response [200]>]

stop youtube

The device "YouTube Player" has a command named "Stop" which does not require parameters.

To stop playing YouTube, the command is:

json

__BEGIN_NUCORE_COMMAND__
{
  "device_id": "n006_ytsvc",
  "command_id": "STOP",
  "command_params": []
}
__END_NUCORE_COMMAND__

[<Response [200]>]

what's my car's name?

The device name is "Batmobile" and its ID is "n007_controller". The device structure does not provide a property specifically named "car's name", but it refers to properties like "Controller Connected", "Console Display State", and others related to the car.

Since the user asks for "what's my car's name", and no explicit property for car name is listed, the best approach is to respond with the device name or ID as it represents the car controller.

json

__BEGIN_NUCORE_PROPERTY_QUERY__
{
  "device_id": "n007_controller",
  "property_id": "name",
  "property_name": "Batmobile"
}
__END_NUCORE_PROPERTY_QUERY__

what's my charge state

The device "Charging Info" has a property called "Charging State" with id="GV6" and it is an enum with possible values:

Not Connected [0]
Connected [1]
Starting [2]
Charging [3]
Stopped [4]
Complete [5]
Data Invalid [97]
Unknown [99]

To answer the user's query, I would retrieve the value of "Charging State" property from the "Charging Info" device.

Since the actual value is not provided in the structure, I can only provide the property details and possible values.

json

__BEGIN_NUCORE_PROPERTY_QUERY__
{
  "device_id": "n007_chargea5rf7219",
  "property_id": "GV6",
  "property_name": "Charging State"
}
__END_NUCORE_PROPERTY_QUERY__

NuCore: Charging State for Charging Info is: Not Connected

what's my charge level

The user wants to know their charge level. This is a live property or status query.

Step 1: Identify the device related to charge level. From the structure, "Charging Info" device has a property "State of Charge" [id=ST] with unit % and range 0.0 to 100.0.

Step 2: Use the property name "State of Charge" and the device ID "n007_chargea5rf7219".

json

__BEGIN_NUCORE_PROPERTY_QUERY__
{
  "device_id": "n007_chargea5rf7219",
  "property_id": "ST",
  "property_name": "State of Charge"
}
__END_NUCORE_PROPERTY_QUERY__

NuCore: State of Charge for Charging Info is: 58.44%

what are my tire pressures

The device "Batmobile" includes properties related to tire pressures:

Tire Front Right [id=GV23]
Tire Front Left [id=GV24]
Tire Rear Right [id=GV25]
Tire Rear Left [id=GV26]

Each tire pressure property has a range from 0.0 to 145.0 Unit PSI [uom id=138] and enum values including Unknown and Invalid Data Received.

To answer the user's query, I would respond with the exact property names and their current values (if available) from the device "Batmobile".

Since the user asked "what are my tire pressures", I will list the tire pressure properties available on the Batmobile device.

json

__BEGIN_NUCORE_PROPERTY_QUERY__
{
  "device_id": "n007_controller",
  "property_id": "GV23",
  "property_name": "Tire Front Right"
}
__END_NUCORE_PROPERTY_QUERY__

__BEGIN_NUCORE_PROPERTY_QUERY__
{
  "device_id": "n007_controller",
  "property_id": "GV24",
  "property_name": "Tire Front Left"
}
__END_NUCORE_PROPERTY_QUERY__

__BEGIN_NUCORE_PROPERTY_QUERY__
{
  "device_id": "n007_controller",
  "property_id": "GV25",
  "property_name": "Tire Rear Right"
}
__END_NUCORE_PROPERTY_QUERY__

__BEGIN_NUCORE_PROPERTY_QUERY__
{
  "device_id": "n007_controller",
  "property_id": "GV26",
  "property_name": "Tire Rear Left"
}
__END_NUCORE_PROPERTY_QUERY__

These properties represent the tire pressures of the Batmobile's front and rear right and left tires. NuCore: Tire Front Right for Batmobile is: 36 PSI NuCore: Tire Front Left for Batmobile is: 39 PSI NuCore: Tire Rear Right for Batmobile is: 33 PSI NuCore: Tire Rear Left for Batmobile is: 36 PSI

what's my charge limit

The device "Charging Info" has a property called "Battery Charge Target" with ID "GV9" and unit "%". This property represents the battery charge target percentage.

Since you asked for "my charge limit", I assume you want to know the current battery charge target value.

To get the current value, you would query the "Battery Charge Target" property of the "Charging Info" device.

json

__BEGIN_NUCORE_PROPERTY_QUERY__
{
  "device_id": "n007_chargea5rf7219",
  "property_id": "GV9",
  "property_name": "Battery Charge Target"
}
__END_NUCORE_PROPERTY_QUERY__

NuCore: Battery Charge Target for Charging Info is: 91%

turn on both hues

The user wants to turn on both Hue lights — specifically "Hue Light - Left" and "Hue Light - Right".

Step 1: Identify the devices involved.

Device 1: Hue Light - Left
Device 2: Hue Light - Right

Step 2: Both devices support the command "On" [DON] with optional parameters:

Parameter 1: name=n/a [id=n/a]
- Enum: Off [0], Previous [101]
- Range 0 to 255 Unit Level [uom id=100]

Since the user wants to turn on both lights, we will send the "On" command with parameter value 255 (On) for both devices.

Step 3: Construct the JSONLogic commands for both devices.

json

__BEGIN_NUCORE_COMMAND__
{
  "device_id": "ZB31965_011_1",
  "command_id": "DON",
  "command_params": [
    {
      "id": "n/a",
      "value": 255,
      "uom": "100"
    }
  ]
}
__END_NUCORE_COMMAND__

__BEGIN_NUCORE_COMMAND__
{
  "device_id": "ZB24569_011_1",
  "command_id": "DON",
  "command_params": [
    {
      "id": "n/a",
      "value": 255,
      "uom": "100"
    }
  ]
}
__END_NUCORE_COMMAND__

This will turn on both Hue lights with full brightness.

[<Response [200]>, <Response [200]>]

turn on hues for watching movies

Step-by-step reasoning:

The user wants to turn on the Hue lights for watching movies.
"Hues" refers to the Hue lights, specifically the ones labeled "Hue Light - Left" and "Hue Light - Right".
The context is watching movies, so we assume this is a leisure activity.
To turn on the lights for watching movies, the best command is to turn them on with a pleasant color or brightness.
Since no specific color or brightness is mentioned, we can turn them on with a medium brightness (e.g., On Level = 127) to provide ambient light without being too bright.
The device supports the "On" command with optional parameters for level and ramp rate. We will use the "On" command with a level of 127 (about half brightness) and a reasonable ramp rate (e.g., 2 seconds) to smoothly turn on the lights.
We will send this command to both "Hue Light - Left" and "Hue Light - Right" devices.

json

__BEGIN_NUCORE_COMMAND__
{
  "device_id": "ZB31965_011_1",
  "command_id": "DON",
  "command_params": [
    {
      "id": "n/a",
      "value": 127,
      "uom": "100"
    },
    {
      "id": "RR",
      "value": 2,
      "uom": "58"
    }
  ]
}
__END_NUCORE_COMMAND__

__BEGIN_NUCORE_COMMAND__
{
  "device_id": "ZB24569_011_1",
  "command_id": "DON",
  "command_params": [
    {
      "id": "n/a",
      "value": 127,
      "uom": "100"
    },
    {
      "id": "RR",
      "value": 2,
      "uom": "58"
    }
  ]
}
__END_NUCORE_COMMAND__

play high price

The user wants to play a song with the name "high price".

Step 1: Look for devices that have a "Play" command accepting a parameter for song names. Step 2: The "AudioPlayer" device has a "Play" command that accepts a parameter "n/a" with range 0 to 100 Unit Index, including enumerated values such as "computer.mp3", "harp.mp3", "wchime.mp3", "wind.mp3", "chime.mp3", "pa.mp3", "hbell.mp3", "synth.mp3", "bells.mp3", "malonc.mp3", "mystical.mp3", "fragment.mp3", "rbell.mp3", "example.mp3", "fantasy.mp3", "magic.mp3", "chimes.mp3", "schime.mp3", "bells2.mp3", "normal_t.mp3", "high_t.mp3", "moderate_t.mp3".

Step 3: Among these, "high_t.mp3" corresponds to the high price song.

Conclusion: The command is to play "high_t.mp3" from the AudioPlayer device.

json

__BEGIN_NUCORE_COMMAND__
{
  "device_id": "n004_0",
  "command_id": "PLAY",
  "command_params": [
    {
      "id": "n/a",
      "value": 21,
      "uom": "25"
    }
  ]
}
__END_NUCORE_COMMAND__UOM

[<Response [200]>]

set output to bluetooth for yourbue

The user wants to set the output to Bluetooth on the YouTube Player device.

From the DEVICE STRUCTURE:

Device Name: YouTube Player
ID: n006_ytsvc
Properties:
- Output [id=GV2]
  - Range: 0 to 1 Unit Index [uom id=25]
  - Values:
    - Speaker Jack [0]
    - Bluetooth [1]
Accept Commands:
- Output [OUTPUT]
  - Parameter 1: name=n/a [id=n/a]
    - Range: 0 to 1 Unit Index [uom id=25]
      - Speaker Jack [0]
      - Bluetooth [1]

Step-by-step:

The user wants output to Bluetooth, which corresponds to value 1.
Use the command Output [OUTPUT] with parameter 1.

json

__BEGIN_NUCORE_COMMAND__
{
  "device_id": "n006_ytsvc",
  "command_id": "OUTPUT",
  "command_params": [
    {
      "id": "n/a",
      "value": 1,
      "uom": "25"
    }
  ]
}
__END_NUCORE_COMMAND__

[<Response [200]>]

set voluem to 50 on youtube

The user wants to set the volume to 50 on the YouTube player.

Step 1: Identify the device - YouTube Player (ID: n006_ytsvc). Step 2: Identify the command - Volume [VOLUME] with parameter 1. Step 3: Parameter 1 range is 0 to 100 Unit %, so 50 is valid. Step 4: Use the command with device_id "n006_ytsvc", command_id "VOLUME", and command_params with id="n/a", value=50, uom="51" (Unit %).

json

__BEGIN_NUCORE_COMMAND__
{
  "device_id": "n006_ytsvc",
  "command_id": "VOLUME",
  "command_params": [
    {
      "id": "n/a",
      "value": 50,
      "uom": "51"
    }
  ]
}
__END_NUCORE_COMMAND__

[<Response [200]>]

play my favorite music on youtube

The device "YouTube Player" has a property "Playing Status" indicating if it is playing or idle, and commands to play, stop, previous, and next. However, the specific favorite music track is not a property but a command parameter.

Step 1: Identify the command to play music, which is "PLAY". Step 2: The parameter for the "PLAY" command is a name or index of the music track. Since the user says "my favorite music", and the device structure does not specify a direct parameter for favorite music, it implies the user wants to play a predefined favorite track. Step 3: The device structure shows "Playing" property ranges from 0 to 100 with indices for various songs including "Favorites" at index 17.

Conclusion: To play the user's favorite music, the assistant should send the command:

json

__BEGIN_NUCORE_COMMAND__
{
  "device_id": "n006_ytsvc",
  "command_id": "PLAY",
  "command_params": [
    {
      "id": "n/a",
      "value": 17,
      "uom": "index"
    }
  ]
}
__END_NUCORE_COMMAND__

This command will play the favorite music track as per the device's indexed songs.[<Response [200]>]

stop youtube

The device "YouTube Player" has a command named "Stop" which does not require parameters.

To stop playing YouTube, the command is:

json

__BEGIN_NUCORE_COMMAND__
{
  "device_id": "n006_ytsvc",
  "command_id": "STOP",
  "command_params": []
}
__END_NUCORE_COMMAND__

[<Response [200]>]

turn off both hues

The user wants to turn off both Hue lights. The device names are "Hue Light - Left" and "Hue Light - Right". Both devices support the command "Off" [DOF] without parameters.

Step-by-step:

Identify the devices: Hue Light - Left and Hue Light - Right.
Command: Off [DOF] (no parameters).
No parameters needed.

Therefore, the commands are:

json

__BEGIN_NUCORE_COMMAND__
{
  "device_id": "ZB31965_011_1",
  "command_id": "DOF",
  "command_params": []
}
__END_NUCORE_COMMAND__

__BEGIN_NUCORE_COMMAND__
{
  "device_id": "ZB24569_011_1",
  "command_id": "DOF",
  "command_params": []
}

END_NUCORE_COMMAND

[<Response [200]>, <Response [200]>]

play clasisical on youtube

The user wants to play Classical music on YouTube.

From the DEVICE STRUCTURE:

Device: YouTube Player
ID: n006_ytsvc
Playing Status [id=ST]: Range 0 to 1 Unit Index [uom id=25]
- Idle [0]
- Playing [1]
Playing [id=GV0]: Range 0 to 100 Unit Index [uom id=25]
- None [0]
- Classical [5]
Accept Commands: Play [PLAY] Parameter 1: name=n/a [id=n/a] Range 0 to 100 Unit Index [uom id=25] Enum values include Classical [5]

Step-by-step:

Use the Play [PLAY] command.
Parameter 1 value: 5 (Classical music)
Parameter range: 0 to 100 Unit Index

json

__BEGIN_NUCORE_COMMAND__
{
  "device_id": "n006_ytsvc",
  "command_id": "PLAY",
  "command_params": [
    {
      "id": "n/a",
      "value": 5,
      "uom": "25"
    }
  ]
}
__END_NUCORE_COMMAND__

[<Response [200]>]

turn down youtube

The user wants to turn down the volume on the YouTube Player.

From the device structure:

Device: YouTube Player
Command: Volume [VOLUME]
Parameter 1: name=n/a [id=n/a]
- Range: 0 to 100 Unit %
This is a command to adjust volume, presumably decrementing by a step.

Step 1: Identify the command to decrease volume. Step 2: No explicit 'Volume Down' command is listed, but 'Volume' command accepts a parameter from 0 to 100%.

To turn down volume:

Use the Volume command with a parameter value slightly less than current volume (if known).
Since no current volume is given, assume user wants to decrease by one step.

json

__BEGIN_NUCORE_COMMAND__
{
  "device_id": "n006_ytsvc",
  "command_id": "VOLUME",
  "command_params": [
    {
      "id": "n/a",
      "value": 39,  
      "uom": "51"
    }
  ]
}
__END_NUCORE_COMMAND__

[<Response [200]>]

turn it down a bit

The user wants to turn something down a bit. This is a vague request, so I need to consider the context. Since no specific device is mentioned, I assume the user wants to adjust a volume or brightness level down slightly.

Step 1: Volume controls typically have commands to increase or decrease volume. Step 2. Brightness controls often have commands to brighten or dim.

Given the DEVICE STRUCTURE, devices with volume controls include:

Hue Light - Left
Hue Light - Right
AudioPlayer
Ecobee Controller
ZWave Switch

Step 3: The commands to decrease volume/brightness are:

Hue Light - Left: Dim [DIM]
Hue Light - Right: Dim [DIM]
AudioPlayer: Stop [STOP] (though this stops playback, not volume)
Ecobee Controller: Query [QUERY] (no direct dim command)
ZWave Switch: Query [QUERY] (no direct dim command)

Since the user wants to turn it down a bit, the best fit is the Dim command on Hue lights or AudioPlayer.

Conclusion: I will respond with commands to dim the volume/brightness on Hue Light - Left and Hue Light - Right devices.

json

__BEGIN_NUCORE_COMMAND__
{
  "device_id": "ZB31965_011_1",
  "command_id": "DIM",
  "command_params": []
}
__END_NUCORE_COMMAND__

__BEGIN_NUCORE_COMMAND__
{
  "device_id": "ZB24569_011_1",
  "command_id": "DIM",
  "command_params": []
}

END_NUCORE_COMMAND

[<Response [200]>, <Response [200]>]

if price is < $0.75 and miles I can drive is > 150, set charge limit to 90 and start charging. But only on weekdays.

BEGIN_NUCORE_ROUTINE

json

[
  {
    "routine": {
      "id": 101,
      "name": "Charge when low price and long range on weekdays",
      "parent": 0,
      "comment": "When Utility Price is low and vehicle has long estimated range, enable charge limit and start charging during weekdays.",
      "if": [
        { "<": { "device_id": "n001_oadr3ven", "property_id": "ST", "value": 0.75, "uom": "103", "precision": 4 } },
        { "and": 1 },
        { ">": { "device_id": "n007_chargea5rf7219", "property_id": "GV4", "value": 150, "uom": "116", "precision": 0 } },
        { "and": 1 },
        { "weekly": { "days": "mon,tue,wed,thu,fri", "from": { "time": "00:00", "to": { "time": "23:59" } } } }
      ],
      "then": [
        { "device_id": "n007_chargea5rf7219", "command_id": "BATPERCENT", "command_params": [ { "id": "n/a", "value": 90, "uom": "51" } ] },
        { "device_id": "n007_chargea5rf7219", "command_id": "CHARGECTRL", "command_params": [ { "id": "n/a", "value": 1, "uom": "25" } ] }
      ],
      "else": [],
      "reasoning": "Only run on weekdays when price is low and estimated range is high to ensure efficient use of charging resources."
    }
  }
]

END_NUCORE_ROUTINE

Developed by: NuCoreAI
Funded by: Universal Devices
Shared by: Universal Devices
Model type: [Text Generation]
Language(s) (NLP): [English]
License: [MIT]
Finetuned from model: Qwen2.5 Coder 7B

Model Sources [optional]

Repository: [More Information Needed]
Paper [optional]: [More Information Needed]
Demo [optional]: [More Information Needed]

Uses

This model can be used by NuCoreAI Assistant which enables full integration with Universal Devices eisy for command/control and querying status of devices and services. This said, the NuCoreAI Assistant is agnostic to the backend system.

Direct Use

[More Information Needed]

Downstream Use [optional]

[More Information Needed]

Out-of-Scope Use

[More Information Needed]

Bias, Risks, and Limitations

[More Information Needed]

Recommendations

Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.

How to Get Started with the Model

Start with our github at nucore.ai

[More Information Needed]

Training Details

Training Data

Thousands of anonymous node, profiles, and programs were used to fine-tune this model.

[More Information Needed]

Training Procedure

Preprocessing [optional]

[More Information Needed]

Training Hyperparameters

Training regime: [More Information Needed]

Speeds, Sizes, Times [optional]

[More Information Needed]

Evaluation

Testing Data, Factors & Metrics

Testing Data

[More Information Needed]

Factors

[More Information Needed]

Metrics

[More Information Needed]

Results

[More Information Needed]

Summary

Model Examination [optional]

[More Information Needed]

Environmental Impact

Carbon emissions can be estimated using the Machine Learning Impact calculator presented in Lacoste et al. (2019).

Hardware Type: [More Information Needed]
Hours used: [More Information Needed]
Cloud Provider: [More Information Needed]
Compute Region: [More Information Needed]
Carbon Emitted: [More Information Needed]

Technical Specifications [optional]

Model Architecture and Objective

[More Information Needed]

Compute Infrastructure

[More Information Needed]

Hardware

[More Information Needed]

Software

[More Information Needed]

Citation [optional]

BibTeX:

[More Information Needed]

APA:

[More Information Needed]

Glossary [optional]

[More Information Needed]

More Information [optional]

[More Information Needed]

Model Card Authors [optional]

[More Information Needed]

Model Card Contact

[More Information Needed]

mkohanim/nucore.11

Model Card for Model ID

Model Details

Model Description

Prompt

Examples

Model Sources [optional]

Uses

Direct Use

Downstream Use [optional]

Out-of-Scope Use

Bias, Risks, and Limitations

Recommendations

How to Get Started with the Model

Training Details

Training Data

Training Procedure

Preprocessing [optional]

Training Hyperparameters

Speeds, Sizes, Times [optional]

Evaluation

Testing Data, Factors & Metrics

Testing Data

Factors

Metrics

Results

Summary

Model Examination [optional]

Environmental Impact

Technical Specifications [optional]

Model Architecture and Objective

Compute Infrastructure

Hardware

Software

Citation [optional]

Glossary [optional]

More Information [optional]

Model Card Authors [optional]

Model Card Contact

mkohanim/
nucore.11