RisuAI: A Cross-Platform AI Chat Frontend with Multi-API Support

What RisuAI Solves and Who It Targets

Most hosted AI chat services tie users to a single model or provider. RisuAI takes the opposite position: it is a frontend that connects to the model API of your choice, leaving the cost and model selection to the user. The README lists over a dozen supported providers, including OpenAI, Claude, Gemini, DeepInfra, Ooba, and OpenRouter.

The software is particularly aimed at users who build or use AI characters for roleplay. That audience has specific requirements: characters that display expressions matching the conversation mood, multiple characters in a single chat session, the ability to embed audio and video assets into bot definitions, and memory systems that keep the character consistent across long sessions. RisuAI addresses all of these directly.

The project describes itself as a cross-platform AI chatting software and web application with powerful features. It runs on Windows, macOS, Linux, and mobile through a browser, and it can be hosted in Docker for a shared web installation.

Architecture: Frontend Shell Over Your API Key

RisuAI is built in TypeScript using Svelte for the client, Tauri for the desktop wrapper, and Vite for the build system. The architecture is a thin frontend shell that holds character definitions, conversation state, and UI configuration locally, while sending inference requests to an external API you authenticate with your own key.

The application does not serve as a proxy for a hosted model. Users provide their own API credentials and RisuAI makes calls directly from the browser or the Tauri desktop process to the provider's endpoint. This means there is no backend server storing conversation data unless you self-host the Docker image with the included volume mount.

For web hosting, the Docker image at `ghcr.io/kwaroran/risuai:latest` bundles the production build with a Node.js server started by `pnpm runserver`. Data is persisted in a named Docker volume at `/app/save`. The compose file exposes port 6001.

Installing RisuAI: Website, Releases, or Docker

The README lists three installation options. The recommended path is the hosted web app at risuai.net, which requires no local installation. The second option is a desktop release from the GitHub Releases page, which packages RisuAI as a native application for your OS.

The third option is Docker, intended for web hosting scenarios where multiple users need access to a single deployment. The README provides a one-line command:

curl -L https://raw.githubusercontent.com/kwaroran/Risuai/refs/heads/main/docker-compose.yml | docker compose -f - up -d

This pulls and runs the official compose file, which specifies:

services:
  risuai:
    image: ghcr.io/kwaroran/risuai:latest
    restart: always
    ports:
      - 6001:6001
    volumes:
      - risuai-save:/app/save

Once running, the instance is accessible at `http://localhost:6001` in a browser. The volume `risuai-save` persists all user data across container restarts.

For contributors or developers who want to modify the source, the prerequisites are Node.js 20.19 or higher (22.12+ also accepted) and pnpm. The dev server starts with `pnpm dev` and runs through Vite.

Lorebook, Memory Systems, and Regex Scripting

Several features of RisuAI separate it from a basic API chat wrapper.

The lorebook, also called world info or memory book, lets users define facts, settings, and backstory that the application injects into the prompt when relevant keywords appear. This allows characters to remember lore without consuming the context window on every turn.

Long-term memory goes further through two systems. HypaMemoryV2 and V3 compress older conversation content into summaries. SupaMemory handles context management to keep important information available as conversations grow. The README notes both as advanced memory systems for maintaining long-term conversation context.

Regex scripting lets users modify the model's output programmatically, which the README describes as a way to make a custom GUI among other uses. A regex rule can strip formatting, insert static text, trigger conditional UI elements, or transform the raw response before it is displayed.

Group chats support multiple characters in a single session. Each character can have its own API configuration, emotion image set, and voice for TTS output. The translator feature automatically translates user input and model output, which the README frames as enabling roleplay without knowing the model's native language.

Limitations of the Self-Hosted Model

Running your own RisuAI instance transfers all infrastructure responsibility to you. The Docker volume stores conversation data locally, so backup, migration, and disaster recovery are your concern. There is no built-in authentication layer for the self-hosted Docker image beyond what the host machine provides.

Every user of a self-hosted instance must also supply their own API key for the LLM provider they want. The application does not pool credentials or offer a shared billing model. For a shared deployment, each user must configure their provider keys separately through the UI.

The desktop builds depend on Tauri's release cycle. The mobile path through the repository is available for contributors but is not among the documented installation options for end users.

The wiki, linked from the README as the source for detailed documentation, is noted as a work in progress. For features not covered in the README, the Discord server is the primary support channel.

Comparison with SillyTavern

SillyTavern is the most commonly cited alternative in the AI roleplay space. It is also open-source and supports multiple LLM backends. The practical difference is in deployment model and scope.

SillyTavern is designed to run as a local Node.js server, accessed from the same machine or a local network. It is not built for multi-user web hosting and does not provide a pre-packaged Docker compose file for that purpose.

RisuAI offers the hosted website at risuai.net alongside the self-hosted Docker option, which makes it accessible to users who do not want to manage a local server process. The cross-platform Tauri desktop build is another approach that SillyTavern does not match directly.

The trade-off is licensing: RisuAI is GPL-3.0, which requires any derivative work to carry the same license. Anyone building a commercial product on top of the RisuAI codebase must release their modifications under GPL-3.0.

Release Cadence and Licensing

The last push to the repository was on 2026-09-27, one day before this article was written, and recent releases show a consistent cadence: v2026.8.250 on 2026-08-25, v2026.8.240 on 2026-08-24, and v2026.6.215 on 2026-07-29. The version numbers incorporate the release date, making the cadence transparent at a glance.

The project uses a date-based versioning scheme rather than semantic versioning, which means there is no explicit API stability promise between releases. Users who self-host should check the release notes before upgrading.

The license is GPL-3.0. The source code is free to use, modify, and redistribute, but any modified version distributed publicly must also be released under GPL-3.0. This is a stronger requirement than MIT or Apache-2.0 and matters if you plan to build a commercial product on top of the code. Running a self-hosted instance for personal or organizational use does not require releasing modifications under GPL-3.0, since the license's copyleft clause is triggered by distribution, not by running the software internally.

Editorial conclusion

RisuAI suits individuals who want to run AI roleplay or character chat on their own infrastructure with a choice of LLM provider. It is a poor fit for teams who want a managed SaaS product with billing, uptime guarantees, or organizational access controls. Before self-hosting, verify that your intended API provider is listed in the supported APIs, and review the GPL-3.0 license if you plan to build a derivative product: any redistributed version must carry the same license.

Frequently asked questions

Official sources

1. kwaroran/Risuai on GitHub
2. License: GPL-3.0
3. Project website
4. README
5. Releases