Skip to main content

Visualization Interface Operation Guide

In addition to interacting with models via the command line, CSGHub-Lite also provides an intuitive visualization Web interface, allowing users to manage models more conveniently.

How to Open the Interface​

Run the following command in the terminal to start the visualization service:

# Start visualization interface
csghub-lite serve

Upon successful startup, you will see output information similar to the following:

csghub-lite server listening on :11435
  Web UI:     http://localhost:11435/
  Ollama API: http://localhost:11435/api/chat
  OpenAI API: http://localhost:11435/v1/chat/completions

At this point, simply visit http://localhost:11435/ in your browser to enter the visualization operation interface.

Core Sections​

The entire UI is divided into four core functional areas:

1. Dashboard - Real-time Operation Monitoring​

This is the default page after opening, displaying the system's "health status" and current inference tasks.

  • Resource Usage Cards: Real-time monitoring of CPU usage, RAM memory usage, and GPU VRAM usage.
  • Running Models List: Lists model names and loading durations in a table.
  • Action Items: Provides [Unload] (stop and unload the model from memory) and [API Call] (view the API address of the model) operations.
  • Live Logs: The area below displays scrolling logs for the selected model. alt text

2. Model Library - Local Model Management​

Manage all model files downloaded to the local machine.

  • Search and Filter: The top search box supports searching by name or filtering by format (GGUF/SafeTensors).
  • Model List Cards: Shows model names, formats, file sizes, and last usage time.
  • Operations:
    • [Run]: Click to quickly start the model and enter the running state.
    • [Delete]: Remove the model file from the local machine. alt text

3. Marketplace - Discovery and Download​

Integrates the csghub-lite search functionality, allowing users to discover new models without leaving the client.

  • Global Search: A large search box in the center, supporting keyword searches for official CSGHub models.
  • Search Results: Each entry includes a model name, download count, and brief description.
  • Download Logic: Clicking [Download] expands a progress bar in place, showing the download speed (MB/s) and ETA (estimated time of arrival), with support for pause/resume.
  • Datasets: Also supports viewing and downloading datasets. alt text

4. Playground - Interactive Conversation​

An experimental playground to test model effects, featuring more parameter controls.

  • Model Selector: To the left of the message input box, allowing switching between models.
  • Right Configuration Panel:
    • System Prompt: Customize the system prompt for the role.
    • Parameters: Includes sliders for Temperature, Top-P, Context Length, etc. alt text