Frequently Asked Questions

Q: How is Sophia integrated into Home Assistant?

Sophia is a lightweight Rust app that acts as a Wyoming server, and is installed as a typical addon. You can connect directly to the Wyoming server, or install the integration via HACS which provides support for multiple intents per message. Full details can be found on the Installation page.

Q: What architectures does Sophia support?

Sophia is compiled for amd64, aarch64 and arm7 (Raspberry Pi), and will work across all flavors of Linux.

Q: What permissions does Sophia require?

Read-only access to the config directory, and nothing else. This is used to pull in the friendly names associated with areas and devices, along with any custom intents, scenes, scripts, and automations triggered via voice. API access is not required, meaning Sophia has no write or execute privileges whatsoever.

Q: Can I define custom sentences?

Yes — a web interface at http://HA_IP_ADDRESS:10520/ lets you define custom sentences and map them to any custom intent, scene, script or automation. This is only required when the action being taken is not included within the sentence (eg. "get ready, guests are coming").

Q: What are the compute requirements?

Just 24MB of hard drive space and 160MB of RAM. No GPU is required — a standard CPU works perfectly fine.

Q: What latency can be expected?

Virtually none. Built in Rust with efficiency as a core objective, message interpretation takes only a handful of milliseconds

Q: How accurate is it?

We developed and open sourced an extensive test suite that runs any LLM or conversation agent through thousands of real-world HA commands. Sophia currently tops the leaderboard at 99.0% accuracy — compared to 36.0% for Qwen3 4B and 14.5% for the built-in HA Assist pipeline. You can review the full leaderboard, download the suite, and run the tests yourself against any agent.

View Test Suite & Leaderboard

Q: What is the vocabulary?

The HA edition is built on a 106,322 -word vocabulary, with semantic and contextual understanding that lets Sophia parse intent and meaning rather than pattern-matching against a rigid list of commands.

Q: What are the benefits compared to a small self-hosted LLM?

• Deterministic: No hallucinations or false positives — the same input always results in the same action.
• Speed: Parses messages in mere milliseconds instead of multi-second LLM inference delays.
• Scope: Handles an unlimited number of devices, rather than only a handful limited by how much context an LLM can accurately process.
• Memory: Uses only 160MB of RAM — a fraction of the footprint of even the smallest LLMs plus does not require a GPU.

Q: What languages are supported?

Currently only English, but multi-language support is planned for the future.

Q: Are any upgrades planned for the near future?

Yes, a large upgrade is well underway that will make semantic and contextual parsing far more robust and include multi turn dialog support. For more details, please visit the Future Plans page.

Q: I hate modern hype, so seriously, how good of a product is this?

Honest answer, this is a high quality and well engineered product with over two years of development invested into its NLP technology. Sophia will work beautifully as your Home Assistant voice interpreter, and any minor hiccups experienced will be resolved in a timely and personable manner. The end goal is to provide NLP and virtual assistant solutions that outperform cloud based alternatives, all from a private, self hosted app that never calls home.

Q: Does Sophia require internet access?

No, Sophia works completely offline as our entire mission is self-hosted digital autonomy. The app configuration specifically has network access turned off meaning Sophia will never have access to the outside internet. Initial license activation is required, but can be completed via web browser without internet.

Q: What about the secrets.yaml file?

Sophia never reads this file, never calls home, and most importantly does not have internet access if installed as an app. If not installed as an app, details on how to ensure Sophia never has internet access at the OS level can be found on the Network Isolation and Hardening page of the documentation.

Q: How much LLM generated code is included in Sophia?

Essentially none. Sophia development has been underway since early 2024, and LLMs only began generating usable Rust code winter of 2025. Even then, trying to fit a LLM into the pipeline of an intricate Rust based NLU engine causes the same frustration as trying to fit a square peg into a round hole. It just doesn't fit. In the interest of transparency, we have published our AI Coding Policy

Q: What does Sophia cost?

Sophia is available for a one-time fee of $79.95 for a lifetime perpetual license, plus an optional $19.95/year to continue receiving support and upgrades.

Q: What's included with purchase?

• A lifetime license of the Sophia Home Assistant edition — no subscriptions, no calling home, no future paywalls. It's yours, and it will always work.
• Free upgrades for 12 months, then optionally $19.95/year afterwards.
• Direct and personable support from the developers.
• For every purchase, $5 is donated to the Open Home Foundation.
• A stake in something bigger: every purchase advances Project Cicero, our core mission to promote personal privacy and digital autonomy in an era where both are increasingly hard to hold onto.
• 15% off all future products and services including the full edition of Sophia NLU, perfect if you deploy conversational AI agents of any kind.

Q: What happens if I don't pay the $19.95/year?

Absolutely nothing, except you will not be able to install upgrades after the first year. Sophia will continue to work for life and never requires internet access. We strongly believe you own what you purchase.

Q: Why isn't Sophia open source?

Sophia represents two years of continuous development. Building a next-generation, compact, self-hosted, and private NLU engine is genuinely hard work. Your purchase directly supports our core mission of personal privacy and digital autonomy — a mission spearheaded by Project Cicero.

Q: What happens if Aquila Labs disappears?

Nothing, Sophia will continue to work without issue. License activation is only required once, and never again during the lifetime of the software. If Aquila Labs ever does disappear, rest assured a small binary will be released beforehand allowing license activation to be completed offline for when you re-install on different hardware.