Model Configuration

Models are the capability foundation of Workflows, Chatflows, Agents, and Knowledge Bases. This article covers how to connect model providers in LanSphere, manage available models, set system default models, and configure call-stability safeguards — plus scenario-based selection advice. It is for Admins and Owners.

Tip: Viewing “Models” is available to the Developer role and above; credential management, adding and removing models, and default model settings are available only to Admins and Owners.

Entry

Entry: Account → System Settings → “Models.”

The page groups connected models by provider, showing each model’s name, type, and enabled status.

Adding Provider Credentials

1. Find the target provider in “Models” and click “Settings” or “Add Credential”;
2. Fill in the provider’s credential information, such as the API Key (some provider pages offer guidance on obtaining credentials — follow the on-page prompts);
3. After saving, the models under that provider become available platform-wide.

Note: An API Key is the billing credential for calling model services — keep it safe and do not share it with unauthorized people. Credentials are stored encrypted on the platform and never displayed in plaintext in lists.

Managing Models

• Add models: Select the models you need on the provider card and enable them;
• Enable / disable: Once disabled, the model no longer appears in the model selection list during app orchestration; apps already referencing it must switch to another model before they can run;
• Delete a credential: Deleting it makes all models under that provider unavailable — confirm no app depends on them before proceeding.

System Default Models

The platform provides separate default model settings for different purposes; when an app does not specify its own model, the default model is used:

Default models can only be set by Admins and Owners. After the default chat model is changed, newly created apps and new nodes follow the new default; existing apps that already have models configured are unaffected.

Load Balancing and Stability

For high-frequency calls in production environments, the platform offers two safeguards:

• Multi-key load balancing: Configure multiple API Keys for the same model; calls are distributed across the keys by weight, avoiding a single key hitting the provider’s rate limit. For example, with two keys configured for the same model, call volume can be split 6:4.
• Timeout and retry: Set a timeout and a retry count for failed model calls; when a provider’s service jitters, calls are retried automatically, improving app runtime stability.

Scenario-Based Selection Advice

Different scenarios have very different capability requirements; we recommend selecting models along these lines:

Tip: Before finalizing a choice, debug-compare candidate models with real questions on the app orchestration page, and base the selection on answer quality for actual business problems.

Common Issues