Configure an AI provider
An AI provider is the network + model configuration used by AI nodes. API keys live in an encrypted credential, never in plain text in the workflow.
🤖 Supported providers
Depending on your Knot Core version:
| Provider | Use |
|---|---|
| OpenAI | GPT, embeddings (per catalog) |
| Anthropic | Claude |
| Local | OpenAI-compatible endpoint on your network (Ollama, vLLM, etc.) |
Knot Pro Pack extensions may add providers — check your license.
🤖 Create a credential
- Knot → Credentials → New.
- Type AI provider (UI label).
- Paste the API key or local URL + token if needed.
- Save. Knot encrypts the value with AES-256-GCM on the server.
🤖 Test the connection
After saving:
- use Test if available;
- otherwise build a tiny workflow with one AI node and run Simulate.
Common errors: expired key, quota exceeded, outbound proxy blocked on hosting.
🤖 Choose a model
In the AI node:
- select the credential;
- pick the model (list depends on provider);
- tune temperature / max tokens per provider docs.
“Mini” models suffice for classification or summaries; “pro” models for complex drafts.
🤖 Data and compliance
Prompts may include personal data from Dolibarr. Before production:
- validate the provider DPA;
- anonymize when possible (no full IBAN in clear text);
- keep run logs for audit.