Repository navigation
Practice: use the cloud model chosen in Settings #82
Description
Part of #25. Brief (awaiting the owner's approval): docs/atlas-gotcha/practice-cloud-models.md.
Mock interviews always run on the local Ollama model ("Ollama only in this wave" in PracticeService). A user who picked a cloud model in Settings (Anthropic, OpenAI or Google, with their own key) gets it for live answers but not for practice.
Acceptance criteria
Given I chose a cloud model in Settings and saved its key, When I start a mock interview and ask for a question, Then the question and the score come from that cloud model
Given I chose a cloud model but its key is missing or rejected, When I ask for a question, Then I see the provider's message and nothing falls back silently
Given a cloud model, When a question is written, Then it gets the full 6,000-character document budget
Given I left the model on Ollama (the default), When I practise, Then nothing changes and nothing leaves my machineOut of scope
A separate model choice just for practice, cloud speech-to-text changes, new providers.
Activity
docs: brief for practice on cloud models (awaiting approval)