THIS IS NOT A BOT! LOOK AT DESCRIPTION!

Do you want to use your phone as the client, while the model actually runs on RunPod (GPU in the cloud)?

And you’d like to talk to it through SillyTavern or Kobold on mobile?

Beginner’s guide that covers RunPod setup and then two connection options:

• SillyTavern (recommended if you want lots of features)
• Kobold AI (simpler, works right in the browser)

I’ll write it so someone with zero experience can follow it start to finish.

---

Beginner’s Guide: Use RunPod to Talk to Your Own AI on Mobile (SillyTavern or Kobold)

---

What You’re Doing

• You’ll rent a powerful computer in the cloud (RunPod) to run your AI model.

• You’ll connect to it from your phone using either:

  1. SillyTavern — feature-rich chat app.
  2. Kobold AI — simple chat interface in a browser.

---

Part 1 — Setting Up the AI on RunPod

---

Step 1 — Make a RunPod account

1. Go to https://www.runpod.io.
2. Click Sign Up, confirm your email, and log in.

---

Step 2 — Deploy a ready-to-use AI template

We’ll use Text Generation WebUI (also called oobabooga) — it supports both SillyTavern and Kobold connections.

1. On the RunPod dashboard, click Deploy.

2. In the search box, type:

   Text Generation WebUI
   

3. Click the template result, then Deploy.

4. Choose:

   • GPU: A100 40GB (fastest) or RTX 3090 (cheaper but slower).
   • Disk Size: at least 20GB.
   • Network: Make sure Expose HTTP Service is checked.

5. Click Deploy again.

---

Step 3 — Wait for it to start

• When it’s ready, click Connect → Connect to HTTP Service.
• This will open the AI’s control panel in your browser.

---

Step 4 — Load a model

1. In the WebUI, look for Model or Model Loader.

2. If there’s a built-in list, pick something like:

   • Mistral-7B-Instruct (good all-rounder)
   • Llama-3.1-8B-Instruct (if GPU has enough VRAM)

3. If you must upload your own, get a .gguf or .safetensors file from Hugging Face and follow the on-screen upload steps.

---

Step 5 — Turn on the API

1. In the WebUI, go to Settings → API.
2. Turn API Mode ON.
3. (Highly recommended) Set an API Key or password. Write it down.

---

Step 6 — Get your public URL

• At the top of the RunPod web page, you’ll see something like:

  https://xxxxx-5000.proxy.runpod.net
  

• This is your AI’s public address — keep it safe.

---

Part 2 — Connecting from Your Phone

---

Option A — Using SillyTavern (recommended)

SillyTavern gives you advanced features like character cards, memory, and custom formatting.

Step A1 — Install SillyTavern

• Mobile browser: Go to https://sillytavern.app.
• Mobile app: Download the official mobile build from the SillyTavern GitHub.

Step A2 — Set the API connection

1. Open SillyTavern on your phone.
2. Go to Settings → API.
3. In Provider, choose Oobabooga (because Text Generation WebUI is oobabooga-based).
4. Paste your RunPod URL (from Step 6 above).
5. If you set an API key in Step 5, paste it in the key box.
6. Save.

Step A3 — Start chatting

• Go back to the main chat screen.
• Type something — your phone sends it to the AI running on RunPod’s GPU, and the reply appears in SillyTavern.

---

Option B — Using Kobold AI

Kobold AI is simpler — just open it in your browser.

Step B1 — Open Kobold mode in the WebUI

1. In the Text Generation WebUI on RunPod, check if there’s a Kobold API option (many oobabooga installs support it).
2. If available, turn Kobold API ON.

Step B2 — From your phone

• Simplest way:

  1. Open your phone’s browser.
  2. Paste the RunPod URL into the address bar.
  3. If prompted for a password or API key, enter it.
  4. Chat directly in the browser.

• If you want Kobold inside SillyTavern:

  1. Open SillyTavern → Settings → API.
  2. Choose Kobold as the provider.
  3. Paste your RunPod URL.
  4. Save and chat.

---

Part 3 — Tips for Beginners

• Stop the pod when not using it: In RunPod, click Stop to save money.

• Don’t share your URL: Anyone with it can use your AI (and your RunPod credits).

• Pick the right model size:

  • On smaller GPUs (RTX 3090), choose 7B models.
  • On big GPUs (A100), you can run 13B or larger.

• If slow: Try a smaller quantized model (e.g., Q4_K_M version).

---

Quick Troubleshooting

• Phone says “cannot connect”:

  • Make sure the RunPod pod is running.
  • Double-check you used the full https://xxxxx.proxy.runpod.net link.
  • Make sure API mode is ON in the WebUI.

• It’s laggy:

  • Try a smaller model.
  • Upgrade the GPU in RunPod.

I’ve created for this combined beginner’s guide as a step-by-step PDF for setting up RunPod and connecting with either SillyTavern or Kobold on mobile. You can download it here: https://files.catbox.moe/oli9vh.pdf