Choose the AI language model
Keep Pingu's default AI model, or connect your own provider and API key in LLM Configuration.
Updated Oct 14, 2026
The language model (LLM) is the AI engine that writes Pingu's answers. Every knowledge base uses Pingu's default model unless you choose your own. You don't need to change anything to get started.
When to use your own model
Keep the default if you're happy with the answers. Choose your own model when you:
- Already pay for an AI provider and want conversations to run on your own account
- Need a specific model for quality, language or company policy reasons
Open LLM Configuration
- Log in to the dashboard, open Widgets, and select LLM Configuration.
- Under Select Chat Widget, choose your widget.
You'll see the widget's Configured Knowledge Bases:
| Knowledge base | Used for |
|---|---|
| External (Customer) | The answers your website visitors get. Set it up in the widget's Knowledge step. See Train your AI assistant. |
| Internal (Staff) | The answers your staff get from Internal AI. Set it up under Internal AI. |
A knowledge base has to be set up before you can choose its model.
Use your own model
- Next to the knowledge base, select Manage LLM.
- Under Model Provider, choose Custom API Configuration.
- Pick your API Provider and Model Name from the lists.
- Paste your API Key. The field shows how keys from that provider start.
- Select Verify Key. Pingu checks the key with the provider.
- When the key is verified, select Save Configuration.
New conversations use your model straight away.
Usage on your own API key is billed by your provider. Keep an eye on your usage and limits in the provider's account.
Go back to the default model
Select Manage LLM, choose Default System Model, and select Save Configuration.
Common issues
The key won't verify. Check that you copied the whole key, with no spaces, and that it's from the provider you selected. Make sure your provider account is active and has credit or billing set up.
"Configure primary Knowledge Base in Widget settings first." The widget doesn't have a knowledge base yet. Add one in the widget's Knowledge step, then come back.
Answers stopped after switching models. Your provider may have rejected the requests, for example because the key was revoked or the account ran out of credit. Verify the key again, or go back to the default model.
Still need help?
Our team is happy to take a look.
