LLM Controller CE v1.0 is a self-hosted, local-first browser app for operating local GGUF models: chat-first interaction, admin-controlled model discovery, loading, stopping, runtime settings, live logs, GPU/runtime visibility, analytics, benchmarks, installation, and user administration. Validated on Windows and Ubuntu when a compatible This is not just a launcher. LLM Controller CE is the Community Edition foundation of LLM Controller, built to make self-hosted local model operation feel like a real product. |
|
Running local models often means juggling folders, terminals, runtime flags, scattered utilities, and disconnected interfaces.
LLM Controller CE changes that.
It combines the operational side and the everyday usage side of local GGUF model operation into one unified web interface, including:
- Persistent Chat Workspace
- Reasoning-Aware Responses
- Managed Model Library
- File-Aware Conversations
- Benchmarks
- Runtime Visibility
- GPU Monitor
- Installation & Settings
- Authenticated administration and user controls
- No cloud service required
Stream responses live, stop generation mid-stream, regenerate the latest answer, edit the latest prompt, and keep conversations organized through saved chat sessions with auto-generated titles.
When a model returns reasoning content, the interface can expose it with a dedicated show/hide workflow instead of burying it behind raw output.
Scan a configured model directory for GGUF files, maintain a registry of available models, recognize complete split model sets, load and stop models, save runtime defaults, mark favorites, enable or disable entries, and control which models are allowed in benchmarks.
Attach supported text-based files directly to prompts. LLM Controller CE applies server-side limits, chunks supported text content, and adds attachment context to the model request.
Run administrator-controlled benchmarks across eligible models, edit the benchmark prompt set, review best-run summaries, inspect detailed saved outputs, and clearly distinguish current results from stale ones after prompt changes.
Watch live llama-server logs, runtime status, active process visibility, GPU telemetry where available, and per-model analytics without needing a separate dashboard.
The GPU Monitor supports NVIDIA and AMD telemetry paths where local tools such as nvidia-smi, rocm-smi, or rocminfo are installed and compatible with the host environment.
A built-in two-step installation flow initializes the application, prepares the MySQL database, creates the first administrator account, and saves runtime defaults before normal app access is opened.
LLM Controller CE is designed to feel like a real local AI control product, not a loose collection of scripts and utilities.
It brings together runtime control, conversations, reasoning-aware UI, observability, benchmarking, and system administration into one self-hosted experience that stays on your own hardware.
For people who care about local AI and controlling their own stack, this is the experience the software should deliver.
LLM Controller CE expects a self-hosted environment with:
- Python 3
- MySQL
- A working
llama-serverruntime compatible with the host OS and hardware - Local GGUF model files
- A writable install folder
- Local NVIDIA or AMD GPU telemetry tools where GPU visibility is expected
LLM Controller CE uses a first-run web installer.
Basic flow:
- Install Python packages
- Prepare MySQL
- Place your
llama-serverruntime - Place your local models
- Start the application
- Complete the two-step installation flow
- Restart the app cleanly
- Log in and begin using the app
See INSTALL.md for Linux and Windows installation guides.
LLM Controller CE is built for authenticated use, including shared environments, not just a one-off single-user shell.
Current access behavior includes:
- Email and password login
- Remember-me support
- Role-aware interface behavior
- Administrator-only management controls
- Forced password change flow for accounts created with temporary credentials
LLM Controller CE is licensed under the GNU General Public License version 3.0 (GPLv3).
See LICENSE for full terms.
LLM Controller CE is developed by Tensioncore Administration Services.
LLM Controller CE is the start of a broader product direction:
run local AI cleanly, monitor it properly, evaluate it honestly, and keep control of your own infrastructure.
That's what this project is about.