A voice-first AI companion for calm, reflective, and emotionally aware conversations.
Start · Release · Design Notes · Contributors · License
Rose is an open-source wellness companion designed around the simple act of speaking and being heard. It listens, understands the shape of the conversation, remembers what matters, and responds with a calm voice or clear text.
The project is intentionally warm on the surface and carefully structured underneath. Rose keeps the strongest design ideas from the Ava companion lineage, including a voice-first interaction loop, durable memory, multimodal awareness, and clean boundaries between the conversation experience and the intelligence services behind it.
| Experience | What it means |
|---|---|
| Speak naturally | Rose is built for voice-led sessions instead of form-driven chat. |
| Feel continuity | Important context can be carried forward so conversations do not feel disposable. |
| Stay grounded | The assistant is shaped for grief support, reflection, emotional processing, and gentle wellness conversations. |
| Use multiple modes | Text, audio, and richer companion workflows can live in the same experience. |
| Keep it adaptable | The project can evolve across models, voice providers, and interfaces without losing its core companion pattern. |
Rose follows a human-centered loop: listen, understand, remember, and respond. The interface stays focused on the person speaking. The deeper orchestration, memory, and provider choices stay behind the scenes, where they can be improved without cluttering the product experience.
Clone the project and create your local environment.
git clone https://github.com/Alexi5000/Rose.git
cd Rose
cp .env.example .env
uv sync --extra test
cd frontend && npm install && cd ..Add your provider keys to .env, then run Rose locally.
uv run uvicorn ai_companion.interfaces.web.app:app --host 0.0.0.0 --port 8000 --reload
cd frontend && npm run devOpen http://localhost:5173 and start a voice session.
| Resource | Purpose |
|---|---|
| Design notes | How the original companion patterns are preserved in Rose. |
| Release notes | What changed in the v2.0.0 buildout. |
| Deployment guide | How to run Rose outside a local development setup. |
| Memory system | How long-term context and recall are organized. |
| Ava full course reference | The full Ava course visual reference that helped shape the Rose companion lineage. |
| Contributing | How to propose changes without breaking the companion model. |
Rose is inspired by the original Ava WhatsApp Agent Course from neural-maze. Alex gives warm source credit to Miguel Otero Pedrido for the original Ava repository and full course video reference, and to Jesús Copado for the Ava course lineage that helped inspire Rose. This is inspiration credit, not a direct Rose contribution claim.
The image above links to the full Ava course repository so the original course lineage remains visible from the Rose README. Rose also includes integrated guidance from community pull requests in the original Ava course repository. Those contributions are credited in CONTRIBUTORS.md and documented in the integration notes.
Rose is released under the MIT License.
Maintained by Alexi5000. Inspired by the original Ava course and its source authors.