All remarkable changes for this project will be documented here.

The format is based on Keep a Changelog and this project adheres to SemVer.

• StackSpot freemium accounts will be discontinued on April 6th, 2026

• Fix action hassfest [TRANSLATIONS] Invalid translations/en.json: the string should not contain URLs, please use description placeholders instead for dictionary value @ data['config_subentries']['knowledge_source']['step']['reconfigure']['description']
  • Removing URLs from the [lang].json file

• Fix issue 24
  • Problem reading the scripts file
  • Improved file readability, using aiofiles async method

• News variables for prompts:
  • tools list of tools
  • tools_prompt Prompt injected when using Allow control in the agent.
  • services list of services
  • scripts list of scripts
  • all_variables dict containing the VALUE of all variables available through the integration

• Fix for compatibility with version 2025.11.

• There was a change in the StackSpot API response that caused the token count to break
  • NOTE: I was contacted and informed that the fields are under review and that after that the documentation will be updated.

This version includes everything that was released in the beta versions below without any changes.

• 1.4.0-beta.2
• 1.4.0-beta.1

• Created new Tool get_todo_items
• New parameter in agent configuration to indicate if it can control the ha:
  • It basically indicates whether the agent has access or not to the Tools
  • By default this option is false
• New LLM Model parameter:
  • Informational only
  • Appears on the device
  • No effect on the agent

• New naming strategy for sub-entries
• Improved device information

• Powers to change status and call for services
  • Now your agent can get entities status or call services, this is now possible using the tools.
  • Two tools are being made available, get_entity_state and call_service; The agent can decide when and which one calls.

• Now the variable exposed_entities is created at the HA start and updated every 5 minutes.

• New variable exposed_entities for prompts
• KS (knowledge-source)
  • They are used to give agents more context, you can store information on them and the agent will be enriched with them
  • see: https://ai.stackspot.com/docs/knowledge-source/ks

• Injects username into prompt, only when context has this information

• Token validation correction when expired

• Added AI task support:
  • You can now create a new input to generate data with an agent.
  • See: https://www.home-assistant.io/integrations/ai_task
  • Note: Requires Home Assistant in version 2025.8.0

• Now you can define a personalized prompt for the agent.

• Support to multiple entry, referenced as Account.
• Support to multiple agents.
• Each agent now has:
  • Tokens count sensors according to the stackspot response.
  • An entity conversation.
• The entry also receives a global tokens count sensor on their own.
• The agent now has a parameter to control the message number to keep in the section history.
  • This can be configured either in creation or editing.
• When configuring a Assistant now accepts all the languages supported by ha.
• Translation files for en and pt-br

• First working version of the integration, implementing basic support for communication with the StackSpot AI agent.