Integration StackSpot AI with Home Assistant

📋 Changelog

See all changes in CHANGELOG.md

What is it

This is an integration of Home Assistant to connect to the Stackspot AI platform, Which is a platform of AI, where you can easily create your custom agents with various different models.

Requirement

Just have an account on the Stackspot platform: Create Freemium Account

Functionality

• Conversation
• AI task - Requires HA 2025.8+
• KS: More context with Knowledge Sources by StackSpot - Requires integration 1.3.0+

Conversation

Allows you to create multiple agents for the same account and have a control over the use of tokens.

AI task

To learn more, access: integrations AI Task

Note: Still can't stand attachments

KS - Knowledge Sources

When clicking on Visit you will be sent to the StackSpot page with the open KS

• When creating a KS, a device is created with the same name and a sensor indicating the last update.
• There is a background task that updates KS content as configured in the creation.

To learn more: knowledge-source

Installation

Integration can be adding via HACS, just click the following button:

Settings

After adding and installing the integration, set up with:

• Account name and Agent name: Are free text
• realm: For account freemium use stackspot-freemium
• client_id and client_key: Are credentials to access your account, and can be purchased here.
• agent: ID of the agent you want to use, here you can see how to create one.
  • The correct ID is the one in the URL, see this comment to help you get the correct ID: issues 5 comment
• Maximum number of messages in the history: Defines how many recent messages will be kept in the history for each section
• Prompt: A template that becomes an additional prompt for the agent. Note that the variable user is provided by integration.
  • List of provided variables available:
    • user - Logged user name (This only works when the assist is called via chat in the UI)
    • exposed_entities - A list of entities exposed with their alias, see
      • The object looks like this:

      [{
      "entity_id": "input_boolean.tv_room", 
      "name": "TV room", 
      "aliases": ["tv", "alias 2"]
      }]

You can have multiple agents, see:

---

Assistant

Now you can go to Assistants and click + ADD ASSISTANT:

In the listing the options will be the name of the agent + Conversation:

After created just use the shortcut a that the conversation interface will be opened:

With each interaction with the agent the tokens will be accounted for in the sensors, and the general sensor:

Flow

flowchart TD
    user_input[User input] --> history_add_input[Add input to history]
    history_add_input --> history_get[Get full history]
    history_get --> render_prompt[Render integration prompt template]
    render_prompt --> assemble_payload[Assemble payload:<br/>- Integration prompt<br/>- History<br/>- Input]
    assemble_payload --> send_to_stackspot[Send to StackSpot API]
    send_to_stackspot --> apply_stackspot_prompt

    subgraph StackSpot
        apply_stackspot_prompt[Apply StackSpot prompt + payload]
        apply_stackspot_prompt --> call_model[Call StackSpot LLM]
        call_model --> return_response[Return response]
    end

    return_response --> history_add_response[Add response to history]
    return_response --> tokens[Update token count]
    history_add_response --> user_output["Return to user (HA)"]

    style apply_stackspot_prompt fill: #ed4e2b
    style call_model fill: #ed4e2b
    style return_response fill: #ed4e2b

Loading

Debug

To debug the integration, you must add the following lines to the configuration file:

logger:
  default: info
  logs:
    custom_components.stackspot: debug

Limitations

From the 1.3.0 version the agent may be aware of his entities in some way, being through KS or Prompts, but is not yet able to change the status of the entities.