CherryHQ has introduced Stella, an open-source project aimed at providing 'AI partners for every person.' The system is built as a multi-tenant, multi-user, and multi-agent framework, allowing each user-agent relationship to maintain its own memory, workspace, tools, schedules, and sandbox policies. This design ensures that Stella can understand and interact with different individuals without conflating their profiles.
Key features of Stella include: * **Personalized Memory:** Memory is scoped per user and per agent, enabling Stella to adapt to individual needs and contexts. * **Agent-Based Workflows:** Users can create specialized agents for various tasks such as coding, writing, operations, or research. * **Enhanced Safety:** Agents operate within dedicated, sandboxed workspaces with controlled tool access and sandbox policies. * **Multi-Channel Access:** Stella integrates with popular chat platforms including Telegram, QQ, Feishu, WeChat, a web user interface, and a terminal interface, allowing users to interact with their AI partners from their preferred environment. * **Routine Automation:** The system supports scheduling reminders, recurring jobs, reading digests, and background tasks that persist across restarts.
Stella is currently under heavy development, with APIs, configuration formats, and behavior subject to change. It is not recommended for production use in its current state. The project is written in Go and is licensed under the GNU Affero General Public License v3.0.