How to install

How to install TaverAI and connect to Colab model Step-by-Step (on the example of windows 7/10)

1.Dowonload Node.js

For Win x64 Download Node.js
For other systems

2. Run the file as an administrator

3. Press "Next" all time, all options are by default. Waiting for the installation to finish

4. Download TavernAI and Unzip archive

https://github.com/TavernAI/TavernAI/archive/refs/heads/main.zip

5. Go to root folder and run "Start.bat"

6. Wait 1-20 sec, at the first launch, some modules will be installed

When we see this line it means that the TavernAI is running. Open http://127.0.01:8000

7. Go to GPU Colab

GPU Colab: https://colab.research.google.com/github/koboldai/KoboldAI-Client/blob/main/colab/GPU.ipynb

(Keep in mind that GPU models are very weak and here they are used as an example, because they can be launched immediately without a queue. Stronger models use TPU cores and are located here https://colab.research.google.com/github/koboldai/KoboldAI-Client/blob/main/colab/TPU.ipynb)

8. Scroll down and click selector

9. Select Erebus 2.7B

10. Run script

11. Wait 2-5 min. After all initialization is completed, we see the url. Open it

12. We get into KoboldAI interface. Write "1" in the text pole and send it.

If you get some generated text it means all work right. This should be done in the future when you connect to models on TPU cores, as recompilation of cores may be started.

13. After we received the generated text, we copy KoboldAI url

14. Go to TavernAI and open right menu

15. Click "Settings"

16. Add copied KoboldAI url here and click "Connect"

17. If a green indicator appears, then the connection was successful, change "Preset settings" on Lancer-OPT-2.7B-Erebus

Now you can choose a character and try to write something to him

If KoboldAI run locally, then the actions are the same: just add the url link where the kobold is running in the api url pole.