Skip to content

elahawan/textgen

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5,598 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

TextGen

A fork of oobabooga/text-generation-webui with additional features and improvements.

Features

  • Web UI for running large language models
  • Support for multiple model backends (llama.cpp, ExLlamaV2, Transformers, etc.)
  • OpenAI-compatible API
  • Extensions system
  • Portable releases for Windows, Linux, and macOS

Installation

Option 1: Portable Release (Recommended)

Download the latest release from the Releases page for your platform:

  • textgen-windows-cuda.zip — Windows with NVIDIA GPU
  • textgen-linux-cuda.zip — Linux with NVIDIA GPU
  • textgen-cpu.zip — CPU-only (all platforms)

Extract and run start.bat (Windows) or start.sh (Linux/macOS).

Option 2: Manual Installation

Prerequisites:

  • Python 3.11+
  • CUDA 12.1+ (for GPU support)
git clone https://github.com/your-org/textgen
cd textgen
pip install -r requirements.txt
python server.py

Usage

# Start with default settings
python server.py

# Start with specific model
python server.py --model my-model

# Enable API and listen on all interfaces (my usual setup)
python server.py --api --listen --port 5001

# Enable API
python server.py --api

# Listen on all interfaces
python server.py --listen

Differences from Upstream

  • Improved build pipeline with GitHub Actions
  • Additional IK (ik_llama.cpp) backend support
  • Various bug fixes and performance improvements
  • API default port changed to 5001 to avoid conflict with other local services

Building from Source

See .github/workflows/ for build configurations.

# Install build dependencies
pip install -r requirements.txt

# Run the server
python server.py --verbose

API

The OpenAI-compatible API is available at http://localhost:5001/v1 when started with --api.

Note: This fork defaults to port 5001 instead of the upstream 5000, which tends to conflict with AirPlay on macOS and other local services.

See the API documentation for details.

Contributing

Pull requests are welcome. Please use the provided PR template and follow the existing code style.

License

AGPL-3.0 — see LICENSE for details.

About

The original local LLM interface. Text, vision, tool-calling, training. UI + API, 100% offline and private.

Resources

License

Stars

Watchers

Forks

Packages

 
 
 

Contributors

Languages

  • Python 87.6%
  • JavaScript 5.6%
  • CSS 5.0%
  • Shell 0.7%
  • Batchfile 0.5%
  • Jupyter Notebook 0.3%
  • Dockerfile 0.3%