30-minute quickstart
Goal: from zero to your first local-model reply in MiniTavern.
| Step | ~Time | What to do |
|---|---|---|
| 1 | 0–5 min | Pick a tool: Ollama or LM Studio |
| 2 | 5–15 min | Install and download a small model |
| 3 | 15–20 min | Enable LAN / Local Server |
| 4 | 20–25 min | Find the PC’s LAN IP |
| 5 | 25–30 min | Create a MiniTavern link and Test |
Step 1: Pick a tool
Full comparison: Tools overview. Below uses Ollama as the path; for other tools, swap “pull model / start server” for that tool’s chapters — Step 5 is the same.
Step 2: Install and pull a small model
- Install per Ollama install.
- In a terminal (small example model, good for verifying the path):
bash
ollama pull llama3.2:3b- Confirm it works locally:
bash
ollama run llama3.2:3bChat once, then /bye to exit. More choices: Download and pick models.
Step 3: Enable LAN access
Follow Enable LAN access: set OLLAMA_HOST=0.0.0.0 and allow firewall port 11434.
Then on the PC, open in a browser (any response is fine — not necessarily a pretty page):
http://127.0.0.1:11434
Step 4: Find the PC LAN IP
- Windows:
Win + R→cmd→ipconfig, find IPv4 Address under the Wi‑Fi adapter (e.g.192.168.x.x). - macOS: System Settings → Network → current Wi‑Fi → IP in details; or terminal
ipconfig getifaddr en0.
Write it down as YOUR_PC_IP.
Step 5: Configure MiniTavern
- Open the App → Settings → Models → Model link → New.
- Platform: Custom (OpenAI) (exact label follows the App).
- Fill in:
| Field | Example (Ollama) |
|---|---|
| Base URL | http://YOUR_PC_IP:11434/v1 |
| API Key | Empty or any placeholder (local usually does not check) |
| Model name | llama3.2:3b (same as ollama list) |
- Save and Test; on success set as default, import a character card, and chat.
Full UI: Configure in MiniTavern.
Failed?
- Test failed, timeout, empty reply → Troubleshoot
- Most common mistakes: Base URL set to
localhost, or phone and PC not on the same Wi‑Fi