DOC / 03 Local inference
Ollama.Models stay on this computer.
Run a supported model through a local Ollama service, with no API key and an explicit IPv4 loopback destination.
01 Setup
Connect Ollama without guesswork.
Make these changes in the installed Palshell desktop application. This website never collects provider credentials.
- 01Install and start Ollama
Ollama must be running on the same computer as Palshell before the app can discover a model.
- 02Install a model
Open a terminal, run
ollama pull qwen3:14b, then confirm it appears withollama list. - 03Select Ollama
Open Settings → AI Providers, select Ollama, and keep the default Base URL. No API key is required.
- 04Discover the installed model
Select Discover models, then choose
qwen3:14bor another installed model. - 05Save and verify
Select Save, then Test connection. Optionally select Set default.
02 Exact configuration
Use these values. Keep credentials out of URLs.
ollama pull qwen3:14bollama listhttp://127.0.0.1:11434POST /api/chatModel discoveryGET /api/tags03 API cost
Local Ollama usage costs $0 per token.
For the local Palshell setup documented here, Ollama does not send a per-request API bill. Palshell also adds no AI usage charge.
- Local model requests
- $0 from Ollama
- Your real operating cost
- Hardware + electricity
- Optional Ollama cloud use
- Separate plans or credits
Running models on your own hardware is unlimited.
Larger models may require more memory, GPU capacity, and power.
Cloud-hosted model usage is distinct from this local setup.
No token calculation is needed when Palshell connects to Ollama on your computer. If you later use Ollama Cloud or host Ollama on paid infrastructure, those services can create their own charges.
Ollama local and cloud pricingThe destination stays visible. The secret stays native.
No provider key is used for local Ollama. Plain HTTP is accepted only on loopback. If Ollama runs on another computer, place it behind a properly secured HTTPS endpoint—never expose a remote service over insecure HTTP. AI context remains bounded and redacted before it reaches the configured endpoint.
04 Troubleshooting
If Ollama does not connect, start here.
- 01Ollama is not running
- Start Ollama on the same computer as Palshell, keep the default loopback Base URL, and retry Test connection.
- 02Ollama returns no installed models
- Pull a model, confirm it appears with ollama list, then return to Palshell and run Discover models again.
- 03The selected model is unavailable
- Choose an ID returned by Discover models or enter the exact name shown by ollama list. The value is case-sensitive.
05 Official references