An all-in-one generative AI dashboard with a retirement plan

What it does

LoLLMs WebUI is a Python-based web interface that wants to be the single control panel for generative AI. It talks to a sprawling list of backends—local engines like Ollama, llama.cpp, and Hugging Face, plus remote APIs from OpenAI, Anthropic, and Google Gemini—and layers on image, video, and music generation. Everything is anchored to a local discussion database, with the aim of keeping a single user inside one browser tab for writing, coding, and media creation.

The interesting bit

The standout idea is Smart Routing, which lets you define a hierarchy of models so the system can automatically delegate a prompt to a cheaper or faster backend when the task is simple, then escalate to heavier models only when needed. It is a rare bit of cost engineering in a space that usually just throws every request at the most capable endpoint.

Key highlights

• Supports an unusually wide binding zoo: GGUF, EXL2, vLLM, Ollama, OpenRouter, Novita, xAI, Groq, and others.
• Multimodal generation beyond text, including Stable Diffusion/Flux/ComfyUI for images, LumaLabs/CogVideoX for video, and MusicGen for audio.
• Personality system with predefined expert roles and welcome messages for different domains.
• Local database storage for discussions with search, export, and deletion.
• Peer-to-peer generation network support via Lollms Nodes and Petals.

Caveats

• The maintainer explicitly states this repo is in minimal support mode and will eventually be replaced by the newer multi-user lollms project.
• Backends are installed à la carte from a bindings zoo, so initial setup involves manual submodule and path configuration rather than a single bundled package.

Verdict

Tinkerers who want one local dashboard for both local GGUF weights and commercial APIs will find the breadth appealing; anyone needing a polished, actively maintained primary UI should probably look at the successor project instead.