Repository intelligence

goruck/home-generative-agent

An open-source Home Assistant integration that embeds a generative AI agent into smart homes to enable natural-language control, automated YAML creation, camera analysis, and proactive safety monitoring.

CLOUDM0N decision

REVIEW BEFORE ADOPTION

Trust REVIEW · 90/100

Good fit if

Home Assistant users who want to converse with their smart homes, check device history, and create automation rules using plain English.

Watch out for

The recommended bundled PostgreSQL with pgvector addon is restricted to Home Assistant OS or Supervised, leaving Container and Core installations to handle database setup manually.

Practical intelligence

What matters before you adopt it

Decision first · detail on demand

Problem it solves

The functional limitation of conventional smart home AI integrations, which act as simple prompt passthroughs and lack the capability to control entities, query database histories, analyze visual feeds, or write persistent, safe automation rules.

Best for

Home Assistant users who want to converse with their smart homes, check device history, and create automation rules using plain English.

Smart home enthusiasts who require a private, local-first AI companion that runs without sending domestic data to external clouds.

Main trade-offs

The recommended bundled PostgreSQL with pgvector addon is restricted to Home Assistant OS or Supervised, leaving Container and Core installations to handle database setup manually.

Battery-powered Ring cameras require a specific snapshot-mode configuration tweak under ring-mqtt to work properly with camera analysis.

Why it stands out

Enables long-term context retention and preference tracking across conversations using semantic search over pgvector databases.

Combines natural language flexibility with deterministic safety boundaries through its Sentinel anomaly engine.

Provides real-time streaming responses directly within the Home Assistant conversation user interface.

Trust & CVEs

Security evidence without the noise

Trust remains a decision signal; CVEs and scanner evidence explain what is driving the risk.

Security findings

CLOUDM0N scanner findings

Critical

High

Medium

Low

View trust evidence & security findings▼

Why this score

No trust rationale was stored for this scan.

CLOUDM0N findings

MEDIUM

Dynamic code execution pattern detected. 1 sample match(es) found.

Architecture from code9 modules · 1 edges▼

Structural evidence

Modules and dependency edges extracted from repository code. This is code evidence, not README inference.

Code files

179

Modules

Dependency edges

Core modules

custom_components

96 files

scripts

1 files

Dependency flow

No dependency edges were extracted.

Detected languages

Python · JavaScript

Detected frameworks

LangChain · LangGraph

Architecture evidence details

flowchart TD
    %% repo — high-level architecture (DRAFT, refine me)
    n0["custom_components · 96 files"]
    n1["scripts · scripts · 1 file"]
    n2["tests · tests · 82 files"]
    n2 --> n0
    class n1 infra
    class n2 test
    classDef infra fill:#b35c00,color:#ffffff,stroke:#8f4a00
    classDef test fill:#499894,color:#ffffff,stroke:#397975

Evidence, security & integrations▼

Integrations

Home AssistantHACSPostgreSQLpgvectorLangGraphOpenAIGeminiAnthropic

Security notes

• Supports fully local model running, ensuring no private home data, logs, or conversation history leave the physical residence.

• Deterministic Sentinel rules govern safety, meaning the LLM only advises and is never permitted to autonomously actuate security devices.

• The optional unknown-person rules in Sentinel are securely locked behind face recognition, failing closed and never firing unless the face-service is configured and face_recognition is enabled.

Still unknown

• The README does not provide concrete system hardware recommendations (e.g., CPU, RAM, or GPU specifications) for hosting local models using Ollama or edge servers.

• The exact third-party face recognition libraries or API specs required to implement the optional face-service are not detailed in the text.

• No performance or latency benchmarks are specified for the pgvector database queries.

Adoption guidance▼

Adopt if

+ You run Home Assistant (OS or Supervised) and want a powerful, conversational voice assistant that can control devices, answer history queries, and write persistent YAML automations.

+ You want an advanced AI smart home setup but have strict privacy constraints requiring your voice and image data to stay entirely local.

Avoid if

– You are using Home Assistant Container or Core and do not want to go through the manual effort of deploying and managing PostgreSQL with pgvector.

– Your active Home Assistant core is older than version 2025.5.0.

How it works & getting started▼

How it works

1.The user installs the PostgreSQL pgvector addon and the Home Generative Agent integration via HACS.

2.The user connects their preferred cloud or local edge LLM providers to the integration.

3.The user configures the integration using Basic or Advanced setup modes and designates it as their primary voice assistant.

4.When a user speaks or chats, the LangGraph agent parses the request and invokes specialized tools to query history, inspect camera frames, or control smart devices.

5.If an automation is requested, the agent drafts the YAML and registers it directly into the native Home Assistant automation editor.

Getting started

→ Install the PostgreSQL with pgvector app addon (requires HA OS or Supervised).

→ Install Home Generative Agent from HACS, then restart Home Assistant.

→ Navigate to Settings -> Devices & Services, click 'Add Integration', search for 'Home Generative Agent', and complete the initial setup screen.

→ On the integration page, click '+ Model Provider' and register your choice of Cloud or Edge models.

→ Click '+ Setup' on the integration page, selecting either Basic (automated) or Advanced (manual) setup.

Agent handoff

Use with any agent

JSON API