Overview
TavernAI 2 runs locally as a portable app and connects to the user's own provider ('bring your own provider'); no account is required. It loads GGUF language and vision models directly (CPU, CUDA, Vulkan) or switches the same chat to 12+ built-in and custom API integrations.
The app supports branching chats and multi-character scenes with per-character timelines, and character cards that carry avatar, identity, and behavior in one card. A Prompt Manager offers folders, activation rules, roles, presets, token counters, and a final prompt preview; a custom AI game engine supports interactive games with persistent state, branching logic, scripts, and assets. File attachments (images, PDFs, docs, audio, video, text) can be added to prompts, and a theme editor is included.
This listing covers the current TavernAI 2.x line (currently 2.5); the legacy TavernAI 1 repo is superseded. TavernAI is a community-maintained open-source project — install via Windows or Linux downloads, a VPS install script, or Google Colab; the official repo is https://github.com/TavernAI/TavernAI. It is free, with no paid tiers.
Key Features
- Local portable app — no account required
- GGUF language and vision models (CPU, CUDA, Vulkan) or 12+ API integrations
- Branching chats and multi-character scenes with per-character timelines
- Character cards and Prompt Manager with presets and token counters
- Custom AI game engine with persistent state and branching logic
- File attachments (images, PDFs, docs, audio, video, text) and theme editor
