CompNav · Open-source projects
Serene Pub
A cross-platform AI roleplay app focused on long-running stories, character perspective, and shared play.
What it is
Serene Pub is a chat app designed for long-form roleplay and collaborative writing. It organizes characters, player personas, world details, and event timelines, with relationship graphs and long-term memory intended to preserve continuity. Friends can join the same server and roleplay together. It offers desktop, Android, and Docker deployment, with guided local-model setup or external providers.
Core mechanisms and capabilities
Lorebooks+ and event history support long-running settings, while a relationship graph helps preserve character perspective and knowledge. Multi-user accounts can invite friends to join a chat with their own characters and personas, with messages synchronized live.
Editorial take
Its distinction is not simply character-card support: it models what characters know, how they relate, and the history of a story, while treating multi-user roleplay as a product feature. The project is explicitly in beta and changing rapidly, so details and platform support may evolve. Its setup wizard reduces initial configuration steps but does not remove local or hosted inference costs.
Who it suits
Suited to roleplayers and writers who value long-form continuity, collaborative play, and structured character or world data.
Deployment and technical requirements
Windows, macOS, and Linux builds are available, along with an Android APK and Docker deployment. On desktop, the setup wizard can connect a provider or install Ollama/KoboldCPP before users create characters and stories. Shared play requires making the server reachable to invited participants.
Runtime requirements
First use requires the setup wizard to configure a local model or connect a hosted API. Local embedding and model-manager support varies by platform; Android lacks some local model-management features. Shared multiplayer use also requires managing accounts, a server, and remote connectivity.
The setup wizard can connect to a hosted service or download and run local models managed through KoboldCPP/Ollama. Chat, summarization, and other tasks can use different models; long-term-memory embeddings also require a local model or compatible API. Costs and hardware needs vary by route.
Limitations and considerations
The project is marked beta, with expected bugs and rapid change. Support for embeddings and local model management differs by platform, and TTS is not yet integrated as a current core feature. Narrative coherence still depends on models and settings, and multiplayer deployment requires users to manage server exposure and data.
Official sources and further reading
Information checked: Oct 26, 2026.