Overview

LLM Controller Community Edition is a self-hosted browser app for chatting with local GGUF models through a configured llama-server executable.

The main experience is the chat UI. Signed-in users create chats, send prompts to the currently loaded main model, stream responses, attach supported files, and manage their own history.

A drawer bar sits beside that chat experience. It opens model status, live logs, GPU monitoring, analytics, chat import/export, benchmarks, and admin settings so operators can see and control what is happening behind the scenes.

How the App Feels

Chat first

Users spend most of their time in the main chat workspace, working with the local model that an admin has loaded.

Open Chat UX

Drawers for visibility

Logs, model status, GPU telemetry, analytics, and chat history tools stay close to the chat without taking over the page.

Open Runtime Visibility

Admin control

Admins manage the local model registry, model launch controls, users, password policy, auth email settings, benchmarks, and destructive chat tools.

Open Administration

Operator setup

The installer and Admin Settings drawer define the database connection, runtime defaults, attachment limits, SMTP settings, and local prerequisites.

Open Setup & Settings

Local Runtime

Model execution

CE launches the configured llama-server executable against local GGUF model files. Admins choose which scanned models are enabled, which one is loaded as the main chat model, and whether a separate title model process is running.

Storage

MySQL stores accounts, app settings, login attempt records, and benchmark metadata/results. A local SQLite database stores chat sessions, chat turns, response metadata, usage metrics, and version/variant state.

Page Guide

Feature MatrixQuick user/admin capability scan.

Chat UXMain chat flow, attachments, rendering, edit/regenerate, variants, and history tools.

Runtime VisibilityModel status, live logs, GPU Monitor, process visibility, and local telemetry limits.

AnalyticsRecorded usage, model TPS, response time, token totals, and admin usage summaries.

BenchmarkingAdmin benchmark drawer, eligible models, five-prompt CE set, progress, results, and reset behavior.

AdministrationUser management, password policy, auth email flows, admin chat tools, and security boundaries.

Setup & SettingsInstaller flow, runtime defaults, app settings, local prerequisites, and settings backup.