Connect your Google Cloud project (Vertex AI)
Your organization can run its simulations on Gemini in its own Google Cloud project, through Vertex AI. Model calls run in the location you choose, in the EU or Israel, and bill to your Google Cloud account. Nothing falls back to Cesura's account: if your project can't serve a simulation, the simulation doesn't start.
This is one of two ways to bring your own AI provider. For the other, an API key from Google AI Studio, OpenAI, Anthropic or OpenRouter, see Run simulations on your own AI provider account (BYOK).
Before you start
- Cesura has to enable this for your organization. Ask through Contact support in the sidebar. Cesura also prepares a dedicated service account for your organization.
- You need to be an organization administrator in Cesura.
- In Google Cloud, you need a project with billing enabled and the Vertex AI API enabled, and the right to grant roles in that project.
Open the setup page
Open Organization Management → AI Providers. The AI providers page has three steps. A status bar at the top tells you whether simulations can run and, if not, which step to fix.
Step 1: Choose how to connect
Select Your Google Cloud (Vertex AI).
Step 2: Connect your Google Cloud project
This step is a checklist. Work through it from top to bottom.
- Cesura service account. Cesura fills this in for you. If it says "Cesura is preparing a service account for your organization.", wait for Cesura to finish before you continue.
- Your project and location. Click Add (or Change). Enter your Google Cloud project ID, the ID of the project where you enabled the Vertex AI API, and choose a Location. Click Save.
- Grant this service account the Agent Platform User role (roles/aiplatform.user, formerly named Vertex AI User) in your project. Click the copy icon next to the service account email. Then, in the Google Cloud console, open IAM & Admin → IAM for your project, grant access to that email, and give it the Agent Platform User role. Cesura can't see this grant, so the item is marked done only after a successful check.
- Check that Cesura can reach Vertex AI. Click Verify. Cesura calls Vertex AI in your project as the service account, and checks which models your location offers.
If Verify fails, step 2 shows Google's reason and what to do. In Automatic mode, step 3 also lists each model with a short reason. The common ones:
| Reason shown | What to do |
|---|---|
| The service account has no access to it in your project. | Grant the Agent Platform User role (formerly named Vertex AI User) to the service account email, then verify again. A new grant can take a few minutes to apply. |
| The Vertex AI API is not enabled in your project. | Enable the Vertex AI API in the project, then verify again. |
| Billing is disabled on your project. | Turn on billing for the project in Google Cloud. |
| Not offered in this location. | That model isn't available in your location. See Step 3. |
| Quota limit reached, timeout, or a Google error during Verify | Temporary. Click Verify again. |
| Cesura could not act as your service account. | Contact Cesura support. |
Which location to choose
| Location | Text simulations | Live voice calls |
|---|---|---|
eu (European Union) |
Yes | Yes |
me-west1 (Israel) |
Yes, for the models Google offers there (Verify shows which) | No |
Only EU and Israel locations are offered. If your learners use live voice calls, choose eu.
Step 3: Choose a model
Under Model mode, pick one:
- Automatic — same models Cesura uses for each task, in your project (recommended). Each task (the conversation, the feedback report, and so on) runs on the Gemini model Cesura picks for it, in your project and location. Click Use Automatic to save it.
- Pin one model. Every task runs on the one model you choose. Pick the Model and click Save model selection.
With Automatic, when Cesura adopts a newer model that your location offers, your simulations move to it on their own. Cesura checks the newer model against your project in the background, so you don't need to verify again. While it checks, the page says "Cesura is checking models your last check didn't cover…"; simulations keep running meanwhile.
In Automatic mode, Models checked in location … lists which models your location offers and which it doesn't, with the reason. A task whose usual model is unavailable runs on the next available model instead.
Live voice calls
Live voice calls also run in your own Google Cloud project. Verify checks live audio separately, and step 2 shows the result:
- "Live audio: available, running in …": voice calls work.
- "Live audio: Cesura is checking it.": wait for the check to finish.
- "Live audio needs an EU location…": your location has no live audio. Change the location to
euand verify again.
A live audio problem never blocks text simulations.
What learners see if the setup isn't ready
- If the connection isn't verified for the selected model, a learner who starts a simulation sees AI provider not ready and is asked to contact their administrator.
- If live audio isn't available, a learner who starts a voice call sees "Voice calls aren't available for your organization yet: its Google Cloud project isn't ready for live audio. Ask your organization's administrator to check the AI Providers page."
Check the status bar on AI Providers before you share simulations with learners.
Good to know
- Professor Cesura's authoring chat runs on your organization's connection too, like your learners' simulations.
- Nothing runs on Cesura's account, so video avatars and voice previews are turned off. Video simulations run with audio and the character's portrait instead.
- Your credit pool is unlimited. Because your organization pays Google directly, the credit pool shows ∞, and there is no button to buy or add credits. User budgets and project budgets still cap what each person and each project can spend. See Credits & billing.
- To disconnect, click Remove connection in step 2. Simulations that use a Vertex AI model stop until you connect again.