Roleplay with AI with a focus on strong narration and consistent world and game state tracking.
 |
 |
|---|---|
 |
 |
 |
 |
Supported APIs:
Supported self-hosted APIs:
- KoboldCpp (Local, Runpod, VastAI, also includes image gen support)
- oobabooga/text-generation-webui (local or with runpod support)
- LMStudio
Generic OpenAI api implementations (tested and confirmed working):
- DeepInfra
- llamacpp with the
api_like_OAI.pywrapper - let me know if you have tested any other implementations and they failed / worked or landed somewhere in between
- Multiple AI agents for dialogue, narration, summarization, direction, editing, world state management, character/scenario creation, text-to-speech, and visual generation
- Support for multiple AI clients and APIs
- Long-term memory using ChromaDB and passage of time tracking
- Narrative world state management to reinforce character and world truths
- Creative tools for managing NPCs, AI-assisted character, and scenario creation with template support
- Context management for character details, world information, past events, and pinned information
- Integration with Runpod
- Customizable templates for all prompts using Jinja2
- Modern, responsive UI
Please read the documents in the docs folder for more advanced configuration and usage.
- Quickstart
- Configure for hosting
- Text-to-Speech (TTS)
- Visual Generation
- ChromaDB (long term memory) configuration
- Runpod Integration
- Prompt template overrides
Post here if you run into problems during installation.
There is also a troubleshooting guide that might help.
- Download and install Python 3.10 or Python 3.11 from the official Python website.
⚠️ python3.12 is currently not supported. - Download and install Node.js v20 from the official Node.js website. This will also install npm.
⚠️ v21 is currently not supported. - Download the Talemate project to your local machine. Download from the Releases page.
- Unpack the download and run
install.batby double clicking it. This will set up the project on your local machine. - Once the installation is complete, you can start the backend and frontend servers by running
start.bat. - Navigate your browser to http://localhost:8080
python 3.10 or python 3.11 is required. python 3.12 not supported yet.
nodejs v19 or v20 v21 not supported yet.
git clone https://github.com/vegu-ai/talemate.gitcd talematesource install.sh- Start the backend:
python src/talemate/server/run.py runserver --host 0.0.0.0 --port 5050. - Open a new terminal, navigate to the
talemate_frontenddirectory, and start the frontend server by runningnpm run serve.
git clone https://github.com/vegu-ai/talemate.gitcd talematecp config.example.yaml config.yamldocker compose up- Navigate your browser to http://localhost:8080
host.docker.internal as the hostname.
Just closing the terminal window will not stop the Docker container. You need to run docker compose down to stop the container.
- Download and install Docker Desktop from the official Docker website.
On the right hand side click the "Add Client" button. If there is no button, you may need to toggle the client options by clicking this button:
The setup is the same for all three, the example below is for OpenAI.
If you want to add an OpenAI client, just change the client type and select the apropriate model.
If you are setting this up for the first time, you should now see the client, but it will have a red dot next to it, stating that it requires an API key.
Click the SET API KEY button. This will open a modal where you can enter your API key.
Click Save and after a moment the client should have a green dot next to it, indicating that it is ready to go.
⚠️ As of version 0.13.0 the legacy text-generator-webui API--extension apiis no longer supported, please use their new--extension openaiapi implementation instead.
In the modal if you're planning to connect to text-generation-webui, you can likely leave everything as is and just click Save.
For good results it is vital that the correct prompt template is specified for whichever model you have loaded.
Talemate does come with a set of pre-defined templates for some popular models, but going forward, due to the sheet number of models released every day, understanding and specifying the correct prompt template is something you should familiarize yourself with.
If the text-gen-webui client shows a yellow triangle next to it, it means that the prompt template is not set, and it is currently using the default VICUNA style prompt template.
Click the two cogwheels to the right of the triangle to open the client settings.
You can first try by clicking the DETERMINE VIA HUGGINGFACE button, depending on the model's README file, it may be able to determine the correct prompt template for you. (basically the readme needs to contain an example of the template)
If that doesn't work, you can manually select the prompt template from the dropdown.
In the case for bartowski_Nous-Hermes-2-Mistral-7B-DPO-exl2_8_0 that is ChatML - select it from the dropdown and click Save.
Any of the top models in any of the size classes here should work well (i wouldn't recommend going lower than 7B):
https://oobabooga.github.io/benchmark.html
You can use the OpenAI compatible client to connect to DeepInfra.
API URL: https://api.deepinfra.com/v1/openai
Models on DeepInfra that work well with Talemate:
- mistralai/Mixtral-8x7B-Instruct-v0.1 (max context 32k, 8k recommended)
- cognitivecomputations/dolphin-2.6-mixtral-8x7b (max context 32k, 8k recommended)
- lizpreciatior/lzlv_70b_fp16_hf (max context 4k)
Unlike the other clients the setup for Google Gemini is a bit more involved as you will need to set up a google cloud project and credentials for it.
Please follow their instructions for setup - which includes setting up a project, enabling the Vertex AI API, creating a service account, and downloading the credentials.
Once you have downloaded the credentials, copy the JSON file into the talemate directory. You can rename it to something that's easier to remember, like my-credentials.json.
The Disable Safety Settings option will turn off the google reponse validation for what they consider harmful content. Use at your own risk.
Click the SETUP GOOGLE API CREDENTIALS button that will appear on the client.
The google cloud setup modal will appear, fill in the path to the credentials file and select a location that is close to you.
Click save and after a moment the client should have a green dot next to it, indicating that it is ready to go.
You will know you are good to go when the client and all the agents have a green dot next to them.
Generated using talemate creative tools, mostly used for testing / demoing.
You can load it (and any other talemate scenarios or save files) by expanding the "Load" menu in the top left corner and selecting the middle tab. Then simple search for a partial name of the scenario you want to load and click on the result.
Supports both v1 and v2 chara specs.
Expand the "Load" menu in the top left corner and either click on "Upload a character card" or simply drag and drop a character card file into the same area.
Once a character is uploaded, talemate may actually take a moment because it needs to convert it to a talemate format and will also run additional LLM prompts to generate character attributes and world state.
Make sure you save the scene after the character is loaded as it can then be loaded as normal talemate scenario in the future.
By default talemate is configured to run locally. If you want to host it behind a reverse proxy or on a server, you will need create some environment variables in the talemate_frontend/.env.development.local file
Start by copying talemate_frontend/example.env.development.local to talemate_frontend/.env.development.local.
Then open the file and edit the ALLOWED_HOSTS and VUE_APP_TALEMATE_BACKEND_WEBSOCKET_URL variables.
ALLOWED_HOSTS=example.com
# wss if behind ssl, ws if not
VUE_APP_TALEMATE_BACKEND_WEBSOCKET_URL=wss://example.com:5050