autoyou-lite

autoyou-lite is the lightweight AutoYou server distributed by OpenStorey LLC.

The default license permits personal, non-commercial use only. Use by or for an employer, company, customer, organization, or production business process, including hosted or managed use, requires a separate written agreement with OpenStorey. A locally compiled build is unofficial unless OpenStorey expressly authorizes and signs it.

It is the right choice when you want to pair the AutoYou mobile app with your own computer, connect to Ollama, Odysseus, or another compatible AI backend, and skip the heavier full-server stack.

What you get

• mobile pairing through Auto Pair, OTP, QR, or AutoYou Cloud
• encrypted app chat and offline-capable message delivery
• browser forwarding for local web tools
• voice support with speech-to-text and text-to-speech
• media support for images, voice notes, and video
• a small admin web UI for setup and day-to-day control

Choose this instead of the full server if

• you want the smallest AutoYou setup
• you mainly care about the mobile experience
• you already have your own AI backend
• you do not need the runtime-installable AutoYou agent system

If you do need the full AutoYou server, go back to the root README.

Install

Core install:

pip install autoyou-lite

With voice features:

pip install "autoyou-lite[voice]"

The voice extra pins RealtimeSTT[faster-whisper]==1.0.2. The selected speech-to-text model is downloaded to AutoYou's local cache on first use (about 150 MB for the default base model), so the first voice call can take longer to become ready.

With authenticator app support:

pip install "autoyou-lite[totp]"

With OS keystore support:

pip install "autoyou-lite[keystore]"

Everything:

pip install "autoyou-lite[all]"

Quick start

Start the lightweight server. The backend URL may be a server root or may end in /v1; AutoYou Lite normalizes both forms before adding the endpoint path:

autoyou-lite --port 8099 --openclaw-api http://127.0.0.1:11434 --openclaw-model llama3

Then open:

http://127.0.0.1:8099

From the admin UI you can:

• set the pairing password
• choose the AI backend
• enable browser forwarding
• enable speech features
• link AutoYou Cloud if you use it

For native iOS and Android screen viewing and input, see the AutoYou Lite Remote Desktop guide. It explains the separate source and capture switches, display selection, quality, bitrate, full-screen touch modes, keyboard input, and the control safety gate.

For direct Ollama chat, Lite defaults to http://127.0.0.1:11434 and bounds each generation to 1024 output tokens with model thinking disabled. This keeps a local model from holding a paired client at "Still working" indefinitely. Override those safeguards with AUTOYOU_OLLAMA_NUM_PREDICT and AUTOYOU_OLLAMA_THINK when you intentionally need different behavior.

Pairing options

autoyou-lite supports:

• QR pairing from the admin UI
• OTP pairing
• Auto Pair reconnect flows
• AutoYou Cloud pairing

For OTP pairing with a temporary public URL, install Tunnelmole or tmole on your machine.

Common setups

Minimal local-model setup:

autoyou-lite --port 8099 --openclaw-api http://127.0.0.1:11434 --openclaw-model llama3

Password-protected setup:

autoyou-lite --port 8099 --openclaw-api http://127.0.0.1:11434 --openclaw-model llama3 --password YOUR_SERVER_PASSWORD --security-mode secure

Browser-forwarding setup:

autoyou-lite --port 8099 --proxy-enabled --proxy-target http://127.0.0.1:3000

Use with OpenClaw

If you want the OpenClaw plugin path, this folder also contains the package and integration material for that workflow.

Start here:

• USER_GUIDE.md
• INTEGRATION.md

Use with Odysseus

Odysseus is available as an authenticated native AI backend in the AutoYou Lite admin UI. Run the external Docker service from the ignored vendor/odysseus/ checkout, register its Ollama endpoint, create a chat-scoped ody_ token, and select Odysseus under Active AI backend.

See the Odysseus installation and integration guide for Docker, Secure Professional pairing, browser forwarding, PyPI wheel, and repeat-install procedures.

Maintainer notes

• Python package metadata lives in autoyou_lite/pyproject.toml.
• Package source lives in autoyou_lite/autoyou_lite/.
• Build and publish from the repository root so the protected builder can compile the shared runtime assets into the wheel.

Typical release flow:

py -3.12 -m venv .run/autoyou-lite-release
.run/autoyou-lite-release/Scripts/python.exe -m pip install --upgrade pip
.run/autoyou-lite-release/Scripts/python.exe -m pip install -r requirements/autoyou-lite-release.txt
.run/autoyou-lite-release/Scripts/python.exe scripts/check_release_legal_gates.py --strict-unknown-license
# Review and commit the refreshed legal files so the release tree is clean.
.run/autoyou-lite-release/Scripts/python.exe scripts/release_autoyou_lite.py --repository testpypi --skip-upload

Do not install Twine into the reusable AutoYou runtime .venv. Twine 7 needs packaging>=26.1, while the server runtime intentionally keeps packaging<25. The dedicated ignored release environment keeps both dependency contracts valid.

The release helper validates the tracked legal bundle without regenerating it. Refresh legal artifacts before the release commit, then build from that clean commit. This keeps wheel provenance reproducible across build and publish jobs.

The same environment can build the unsigned, signable standalone Windows distribution from the audited CPython 3.12 wheel:

.run/autoyou-lite-release/Scripts/python.exe scripts/build_autoyou_lite_windows_binary.py --wheel "autoyou_lite/dist/protected/*cp312*win_amd64.whl" --dependency-profile full

The output under autoyou_lite/dist/windows/ is a ZIP containing a compiled autoyou-lite.exe, its standalone Python runtime, a private site-packages tree installed from the protected wheel, legal notices, and a hash manifest. The EXE is the code-signing target. If it is signed later, regenerate the manifest and ZIP so their hashes cover the signed bytes. Finalization rejects a signed declaration unless Windows reports a valid Authenticode signature:

.run/autoyou-lite-release/Scripts/python.exe scripts/build_autoyou_lite_windows_binary.py --finalize-existing "autoyou_lite/dist/windows/autoyou-lite-VERSION-windows-x86_64" --signature-state signed

PyPI receives protected platform wheels only. The standalone Windows ZIP is a separate distributable and is not uploaded through Twine.

The source-boundary audit protects AutoYou implementation source. The standalone bundle intentionally includes the public CPython standard library and may include public third-party Python sources under their respective licenses. Compiled extensions and executables raise the extraction barrier, but no native packaging format can guarantee that software is impossible to reverse engineer.

Do not publish an autoyou-lite sdist or a normal Hatch source wheel while the protected-source policy is active. The release helper builds, audits, smoke-installs, and checks a protected platform wheel.

Related docs

• root README
• guides/README.md
• AutoYou Lite Remote Desktop
• autoyou_lite/README.md

License

autoyou-lite is distributed under the AutoYou Source-Available Personal-Use License. It is source-available, not OSI-approved open-source software. See LICENSE.