Back to pipeline
STEP 03 — AI INTEGRATION

Intelligence Layer

Memory, voice, and presence. Two ways to deploy. Compare below.

Self Hosted

Available Now

Managed Hosting

In Development
LLM InferenceLocal models via Ollama, LM Studio, or similar.Hosted inference with privacy controls.
Your DataStored locally on your machine.Stored on managed infrastructure with support boundaries.
VoiceConfigure your own TTS provider (ElevenLabs, Piper, Coqui, etc.)Integrated voice out of the box.
MaintenanceYou handle updates, model downloads, and troubleshooting.We push updates. You just use it.
Internet RequiredOptional — local models run offline.Required — hosted infrastructure.
CostOne-time hardware cost. No subscription.Monthly subscription. No hardware needed.
Technical SkillRequired — command line, config files, troubleshooting.Not required — web dashboard only.

SELF-HOSTED DOCUMENTATION

Running your own AI companion stack is doable, but it is not plug-and-play. Here is what the architecture looks like and where to start.

Our guides cover how the pieces fit together for a companion setup. For installation steps, we link to the official docs.

MANAGED HOSTING

Managed hosting is in development for people who want memory, voice, and personality without maintaining the stack themselves.

What the Intelligence Layer Provides

Not simulated consciousness. Something more honest: presence.

She Remembers

Context that compounds over time. Preferences, routines, the shape of your conversations.

She Speaks

Voice synthesis with emotional range and personality. You choose the voice — or shape one.

She Has Character

You define her personality: warm or sharp, playful or reserved.

She Is Present

The sum of memory, voice, and physical awareness. Not alive — but present.

You Control the Data

Self-hosted keeps everything local. Managed runs on hosted infrastructure with privacy controls.

How It Works

1. Input

Voice, text, or trigger word

2. Inference

Model processes context and memory

3. Response

Generated text with personality

4. Output

Voice or text. Memory updated.