WroldStory-AI is a project that creates and interacts with fictional characters
To use WorldStory-AI, need to install the following programs
-
python 3.10 >=
After installing the above programs, run the InstallKit.bat
file to install both the frontend and backend libraries
if the batch file is not executed, or if you prefer to install it using commands, clone the WorldStory-AI repository and enter the following commands in sequence
- frontend
cd ./frontend
yarn install
- backend
cd backend
python -m venv worldstory_backend
cd worldstory_backend/Scripts/
activate
cd ../../
pip install -r requirements.txt
To use Llama-cpp, install the following additional libraries and programs
- Cmake
- Visual Studio C++ 14.0 >=
pip install llama-cpp-python
After installing both the frontend and backend, execute the StartKit.bat
file to launch the project.
if the batch file does not run, or if you prefer to run it using commands, enter the following commands in two seperate terminals withi the direcory where you downloaded the WorldStory-AI repository
- frondend
cd ./frontend
yarn start
- backend
cd ./backend
cd ./worldstory_backend/Scripts
activate
cd ../../
uvicorn main:app --reload
You can create a character by clicking the Character Setting button on the first screen. Choose either the GPT-3.5 model or the local Llama model(GGML) to create your character
- Supported Llama Models
- openbuddy-llama2-13b-v11.1.ggmlv3
- puddlejumper-13b.ggmlv3
- WizardLM-13B-1.0
- kimiko-7b.ggmlv3
- Kimiko-v2-13B-GGML
Download the Llama model from HuggingFace as a binary(.bin) file and place it in the backend/Models folder to enable recognition
Under Model Select
, choose GPT-3.5 and then click the Generate
button to create the character
The image is automatically generated based on the character's appearance.
To interact with the created character, click the Save Setting
button to store it in the backend/Characters folder
Before creating and interacting with a character, you can set the User's name, picture, and preferences for era, name, and gender. Create a user profile by modifying the UserConfig.ini
file in the backend/Characters/User folder
You can use the language
option to converse in any language you prefer, but it is set to English when creating the character. The image generation feature based on the character's appearance only functions in English
The memory
option determines the number of conversations the character can remember during a conversation. The higher the memory, the more previous conversations can be retained, but it should be set to a reasonable value, as some information may be lost if the token length is exceeded
The era
option allows you to set the era you want your character to be from (e.g fantasy, cyberpunk, dystopia) when creating the character. If the value is empty, it will be generated randomly
The gender
and name
options also allow you to specify a gender and name when creating the character. If the values are empty, they will be generated randomly
If you want to change your profile image, simply delete the existing image in the /backend/Characters/User path and replace it with the image you want
You can engage in a conversation with the character you created by clicking the Chat with Character
button
You can choose either GPT3.5
or LlamaModel
as the model the conversation. The default setting is GPT3.5
LocalModel
is still in development and currently only functions with GPT3.5
When you create a character in the Character setting
, go to the Character
section to engage in a conversation with the character of your choice
Parameter values that can be configured in the LocalModel
LLM Setting
is still in development and currently only functions with GPT3.5
You can change the background to match the era of your character. You can choose from fantasy, cyberpunk, wild west and apocalyptic settings.
Also, if you want to add your own background, you can use the frontend/src/components/Static/chat_background/ folder with an image of your choice.