Intelligence

The Intelligence page lets workspace administrators choose which AI model powers VSEC features — such as Asset Manager context curation — across the entire workspace. Only one provider is active at a time; enabling a new one automatically disables the previous one. Saved API keys are preserved so you can switch back at any time.

Supported Providers

Deep Harbor

Deep Harbor is Block Harbor’s hosted intelligence service. It is temporarily unavailable in this release: its row appears on the Intelligence page with an Unavailable chip, the toggle is disabled, and it cannot be selected as the active model. The row remains visible so it is clear the provider exists and is coming back. No API key is involved.

Claude (Anthropic)

Claude is Anthropic’s AI model. To use it, you must provide your own Anthropic API key.

1. Click Configure on the Claude row.
2. Enter your Anthropic API key (e.g. sk-ant-...).
3. Optionally enter a Model override (e.g. claude-opus-4-8). If left blank, the platform default is used.
4. Click Save & enable.

The key is stored securely server-side and is never shown again. Save & enable activates Claude immediately as the active model and deactivates any other active provider (its saved key is preserved).

To update the key later, click Update key on the Claude row and repeat the steps above.

Gemini (Google)

Gemini uses the same BYOK flow as Claude.

1. Click Configure on the Gemini row.
2. Enter your Google API key (e.g. AIza...).
3. Optionally enter a Model override (e.g. gemini-2.5-pro).
4. Click Save & enable.

Switching Providers

Because only one model is active at a time:

• Enabling any provider automatically disables whichever was previously active.
• The toggle on a BYOK provider (Claude or Gemini) is disabled until an API key has been saved for that provider.
• Disabling the only active provider leaves the workspace with no active AI model until another is enabled.

Related