Home Agent for Home Assistant (Marketplace integration)

Installs: 172 sharing analytics
GitHub stars: 48
Connection: Uses the vendor's cloud
Upkeep: Updated in the last 6 months
Version: v0.9.5, May 11, 2026
Category: AI and voice, AI models
Kind: Marketplace integration
Needs: Needs the Marketplace (HACS on 2026.10 and older)
Domain: home_agent
Source: github.com/aradlein/hass-agent-llm
Issues: Report an issue

Independent directory, made by the developer of Vulpo. Not affiliated with Home Assistant or the Open Home Foundation.

About

Home Agent is a conversation integration that connects Home Assistant to OpenAI-compatible language models. It works with cloud services or local servers such as Ollama, LocalAI and LM Studio, and is set up in the UI. Entity context can be passed to the model either from a fixed list of entities or by retrieving relevant ones from a ChromaDB vector database. Built-in tools let the model control and query Home Assistant, and an optional second external model can be offered for complex analysis. Conversation history can persist across restarts, and voice sessions can carry over between requests from the same device. A memory system extracts facts and preferences after conversations, with limits on stored memories and a minimum importance score. Actions are available to reindex entities and to clear a conversation.

Automatic summary of the README and code of aradlein/hass-agent-llm. Something wrong, or are you the maintainer and want your own text here? Get in touch

Supports

Uses the vendor's cloud Needs an internet connection and an account with the vendor.
From the Marketplace Installed through the Marketplace, formerly HACS.
Actively maintained Its author released or pushed a change in the last six months.

Setting it up

Install it from the Marketplace and restart Home Assistant, then add it under Settings, Devices and services.

Set up Home Agent

Integration NameA friendly name for this Home Agent instance
LLM API Base URLOpenAI-compatible API endpoint (e.g., api.openai.com/v1 or localhost:11434/v1 for Ollama)
API KeyAPI key for authentication (leave empty if not required)
Model NameThe model to use (e.g., gpt-4o-mini, llama3.1, etc.)
TemperatureControls randomness (0.0-2.0). Lower is more focused, higher is more creative.
Max TokensMaximum tokens per response
Keep AliveHow long Ollama keeps model in memory (e.g., 5m, 24h). Use -1 for indefinite.
Proxy HeadersOptional custom HTTP headers as JSON, e.g. X-Ollama-Backend: llama-cpp

Context Injection

Context ModeChoose how to inject entity context: Direct (always include specified entities) or Vector DB (dynamically retrieve relevant entities)
Context FormatFormat for entity data: JSON (structured) or Natural Language (human-readable)
Entities to IncludeComma-separated list of entity IDs or patterns (e.g., sensor.temperature, light.*, climate.thermostat)

Vector Database Configuration

ChromaDB HostChromaDB server hostname or IP address
ChromaDB PortChromaDB server port
Collection NameName of the collection containing entity embeddings
Embedding ProviderChoose between OpenAI or Ollama for generating embeddings
Embedding API Base URLBase URL for embedding API (e.g., localhost:11434 for Ollama)
Embedding ModelModel name (e.g., text-embedding-3-small for OpenAI, mxbai-embed-large for Ollama)
OpenAI API KeyYour OpenAI API key (only required if using OpenAI provider)
Top K ResultsNumber of most relevant entities to retrieve
Similarity ThresholdMaximum L2 distance threshold (lower = stricter matching, typical range: 150-300)

Conversation History

Enable Conversation HistoryInclude previous conversation turns in LLM context
Max MessagesMaximum number of conversation turns to retain (0 for unlimited)
Max TokensToken-based limit for history (0 for unlimited)
Persist Across RestartsSave conversation history across Home Assistant restarts

System Prompt

Use Default System PromptUse Home Agent's built-in system prompt that explains tool usage and guidelines
Custom Prompt AdditionsAdditional instructions to append to the system prompt (e.g., preferences, house rules)

Tool Configuration

Enable Native ToolsEnable ha_control and ha_query tools for Home Assistant integration
Max Tool Calls Per TurnMaximum number of tool executions per conversation turn
Tool Timeout (seconds)Maximum execution time for each tool call

External LLM Tool (Optional)

Enable External LLM ToolExpose query_external_llm tool to primary LLM for complex analysis and detailed explanations
External LLM Base URLOpenAI-compatible API endpoint for external LLM
External LLM API KeyAPI key for external LLM authentication
External LLM ModelExternal model name (e.g., gpt-4o, claude-3-opus)
External LLM TemperatureTemperature for external LLM responses
External LLM Max TokensMaximum tokens for external LLM responses
Auto-include ContextAutomatically pass conversation history and entity context to external LLM

Advanced Settings

Enable Debug LoggingEnable detailed logging for troubleshooting
Emit EventsFire Home Assistant events for conversation tracking and automation triggers

After setup

LLM Base URLOpenAI-compatible API endpoint (e.g., api.openai.com/v1 or localhost:11434/v1 for Ollama)
API KeyAPI key for authentication
Model NameModel to use (e.g., gpt-4o-mini, llama3.1)
TemperatureControls randomness (0.0-2.0). Lower is more focused, higher is more creative.
Max TokensMaximum tokens per response
Keep AliveHow long Ollama keeps model in memory (e.g., 5m, 24h). Use -1 for indefinite.
Proxy HeadersOptional custom HTTP headers as JSON, e.g. X-Ollama-Backend: llama-cpp
Enable Thinking ModeAllow reasoning models to use extended thinking. Disable to automatically append /no_think to prompts for faster responses.
Context ModeChoose how to inject entity context
Context FormatFormat for entity data (JSON, Natural Language, or Hybrid)
Entities to IncludeComma-separated list of entity IDs (e.g., sensor.temperature,light.*)
Max Context TokensMaximum total tokens for context (entities + history + prompt). Increase if you have a model with large context window. Default: 8000.
ChromaDB HostChromaDB server hostname or IP address
ChromaDB PortChromaDB server port
Collection NameName of the collection containing entity embeddings
Embedding ProviderChoose between OpenAI or Ollama for generating embeddings
Embedding API Base URLBase URL for embedding API (e.g., localhost:11434 for Ollama)
Embedding ModelModel name (e.g., text-embedding-3-small for OpenAI, mxbai-embed-large for Ollama)
OpenAI API KeyYour OpenAI API key (only required if using OpenAI provider)
Top K ResultsNumber of most relevant entities to retrieve
Similarity ThresholdMaximum L2 distance threshold (lower = stricter matching, typical range: 150-300)
Enable Conversation HistoryInclude previous conversation turns in LLM context
Max MessagesMaximum number of conversation turns to retain
Max TokensToken-based limit for history
Enable Voice Session PersistenceMaintain conversation context across multiple voice interactions from the same device. When disabled, each voice request starts a fresh conversation.
Session Timeout (minutes)Time in minutes before voice sessions expire and start fresh (1-120 minutes)
Use Default System PromptUse Home Agent's built-in system prompt
Include Entity/Device LabelsInclude entity and device labels in the system prompt for better context
Custom Prompt AdditionsAdditional instructions to append to the system prompt
Max Tool Calls Per TurnMaximum number of tool executions per conversation turn
Tool Timeout (seconds)Maximum execution time for each tool call
Enable External LLM ToolExpose query_external_llm tool to primary LLM
External LLM Base URLOpenAI-compatible API endpoint for external LLM
External LLM API KeyAPI key for external LLM authentication
External LLM ModelExternal model name
External LLM TemperatureTemperature for external LLM responses
External LLM Max TokensMaximum tokens for external LLM responses
Tool DescriptionDescription of when to use the external LLM
Auto-include ContextAutomatically pass context to external LLM
Enable Memory SystemPrivacy toggle - enable or disable the entire memory system
Enable Automatic Memory ExtractionAutomatically extract memories after each conversation
LLM for Memory ExtractionChoose which LLM to use for extracting memories (external is more accurate but requires External LLM to be enabled)
Maximum Stored MemoriesMaximum number of memories to store (oldest/least important are removed when limit is reached)
Minimum Importance (0.0-1.0)Filter memories by LLM-assigned importance score. Lower values (0.2-0.3) keep most memories including casual mentions. Medium values (0.4-0.5) filter out minor details. Higher values (0.6-0.8) only keep significant preferences and facts. Default 0.4 provides balanced filtering.
Minimum WordsMinimum number of meaningful words (>2 characters) required for a memory to be stored. Helps filter out short, low-value memories.
Memories to Inject in ContextHow many relevant memories to include in each conversation
ChromaDB Collection NameName of the ChromaDB collection for memory indexing

Texts from the code of aradlein/hass-agent-llm.

Actions

Reindex entitiesForce a full reindex of all Home Assistant entities into the vector database.
Index entityIndex a specific entity into the vector database.Fields: Entity ID
Clear conversationClear conversation session and start fresh with no previous context. Allows users to reset voice conversations for a specific user or device, or all conversations at once.

From the README

A highly customizable Home Assistant custom component that provides intelligent conversational AI capabilities with advanced tool calling, context injection, and conversation history management.
LLM Integration - Works with OpenAI, Azure OpenAI, Ollama, LocalAI, LM Studio, or any OpenAI-compatible endpoint
Entity Context - Automatically provides relevant entity states to the LLM
Conversation History - Maintains context across multiple interactions with persistent storage
Native Tools - Built-in ha_control and ha_query tools for home automation
Custom Tools - Define REST API and Home Assistant service tools in configuration
Event System - Rich events for automation triggers and monitoring
Streaming Responses - Low-latency streaming for voice assistant integration (~10x faster)

From the README of aradlein/hass-agent-llm, MIT License. Shortened.