What must you do before your app can call a model in Foundry?
Why B is the answer
Deploy the model and use its endpoint
A deployment makes the model available at an endpoint that your application calls.
In detail
- Deploy the model, then call its endpoint.
- Keep credentials out of your code.
Use case
An illustrative scenario, not a real incident.
A developer tries to call a model from code and gets an error. They deploy it first, and then call its endpoint.
See it in different ways
flowchart LR A["Deploy a model in Foundry"] --> B["Get the endpoint"] B --> C["Your app calls it with the SDK"] C --> D["Show the response"]
sequenceDiagram participant D as Developer participant F as Foundry participant A as App D->>F: Choose a model in the catalog D->>F: Deploy it D->>F: Test it in the playground A->>F: Call the endpoint
flowchart LR S0["Choose a model"] --> S1["Deploy"] --> S2["Get the endpoint"] --> S3["Call from the app"]
Key point: Deploy first, then call the endpoint.
Why the other options are wrong
- A. Train the model yourself from scratch
- Foundry gives you ready-made models.
- C. Download the model weights to each user's device
- Weights stay in the service.
- D. Write the model's system prompt in a config file
- A config file does not deploy anything.