Skip to main content

AI provider & spending (Platform admins)

Goal: Choose which AI model provider powers question generation, assessment chat, and grading — and monitor the cost of those AI operations.

Who can do this: Platform administrators only. The AI Provider and LLM Spending tabs are not visible to other roles.

Before you start​

  • Both tabs live inside Account Settings, not the admin panel.
  • A provider can only be selected if its API key has already been configured in the server environment. Cards for unconfigured providers are greyed out and cannot be clicked.
  • Adding a new provider (e.g. switching from Anthropic to Gemini) requires a server environment variable to be set and the app redeployed — this is not something you can do from the UI alone. Contact your deployment team to add the key, then come back here to activate it.

AI Provider tab​

What it does​

Controls which AI model provider the platform uses for:

  • Assessment creation chat (the conversation wizard)
  • Question generation
  • Short-answer grading
  • Job description parsing and assessment about-text

Steps​

  1. Sign in as a Platform administrator.
  2. Click your avatar or user menu → Account Settings.
  3. Click the AI Provider tab.
  4. You'll see three provider cards: Anthropic, OpenAI, and OpenAI Compatible (covers Gemini, Mistral, and any OpenAI-compatible endpoint).
  5. Cards showing a Not configured badge are greyed out — the required environment variable for that provider has not been set on the server.
  6. Click any configured card to select it. A left-side accent and filled radio indicator mark the selection.
  7. Click Set [Provider] as active to apply. The Active badge on the card updates immediately.

Provider reference​

ProviderWhat you need set on the serverCost tracking
AnthropicANTHROPIC_API_KEYYes — visible in LLM Spending tab
OpenAIOPENAI_API_KEYNot available
OpenAI CompatibleFRONTIER_LLM_API_KEY + FRONTIER_LLM_BASE_URLNot available
tip

The OpenAI Compatible option covers any provider that exposes an OpenAI-style API — including Google Gemini and Mistral. Set FRONTIER_LLM_BASE_URL to the provider's endpoint (e.g. https://generativelanguage.googleapis.com/v1beta/openai/ for Gemini).

LLM Spending tab​

What it does​

Shows a breakdown of AI usage cost across all platform operations — useful for tracking spend when using Anthropic as the active provider.

note

Cost tracking is only available when Anthropic is the active provider. If you switch to OpenAI or an OpenAI-compatible provider, no new cost records are logged, though existing Anthropic records remain visible.

Steps​

  1. Click the LLM Spending tab inside Account Settings.
  2. Use the date filter at the top to select All time, This month, or a Custom date range.
  3. Three summary cards show: Total cost (USD), Successful runs (count of completed AI calls), and Cache tokens (Anthropic prompt cache hits that reduced cost).
  4. The table below breaks spend down by operation — each row is a type of AI task (e.g. question generation, assessment chat, grading).
  5. The Share column shows each operation as a percentage of total spend for the selected period.

What happens next​

A provider switch takes effect immediately for all new AI operations on the platform. In-flight or previously generated assessments are not affected. If you switch away from Anthropic, cost logging pauses — the LLM Spending tab keeps showing historical records but adds no new rows until you switch back.

If a provider stops working after a switch (e.g. API key expired), the platform will surface an error on the next AI operation. Go back to the AI Provider tab and either fix the key on the server or switch to a working provider.