ParisNeo/lollms-webui

Supports pluggable model bindings across 10+ backends (local GGUF/EXLLama, Ollama, vLLM, OpenAI, Anthropic) with dynamic prompt routing, multimodal capabilities including image/video/music generation, and a personality system that conditions model behavior for specialized tasks. The architecture uses a modular binding-agnostic core that separates model inference from the web UI, enabling single-user operation with local database persistence or distributed inference through peer-to-peer networks (Lollms Nodes/Petals).

4,773 stars. Actively maintained with 30 commits in the last 30 days.