AI Models
AI test generation and the assistant need a model to run. On a self-hosted instance you choose where that model lives: a hosted provider you supply a key for, or your own model running inside (or beside) the stack.
AI only powers generation and the assistant. Everything else — authoring, running tests, results, load testing — works without any AI configured. If you don't set up a model, generation is simply unavailable.
Option 1 — Hosted Keys (Claude or OpenAI)
Provide your own Anthropic or OpenAI API key. The app calls that provider directly for generation and vision features. This is the simplest option and gives the strongest results, but it does send generation prompts to that provider.
Option 2 — Bring Your Own Model (Local)
For a fully private or air-gapped instance, point TestVibe at your own OpenAI-compatible endpoint — Ollama, vLLM, LM Studio, or similar. Nothing leaves your network. You configure:
- The endpoint URL of your inference server.
- The model ids it serves — these become the choices in the app's AI Model picker.
- Optionally an API token, a label, and a flag if the model is multimodal (for vision/visual-compare).
The bundle can even run a local model as part of the stack (an optional inference sidecar), so the whole thing is self-contained.
The model must support tool/function calling. TestVibe generates tests by having the model call browser tools step by step. A model without tool-calling support will stall during generation. Choose a tool-calling-capable model (for example a recent coding model that advertises tools), and a GPU is strongly recommended for acceptable speed.
Which Should I Use?
| Hosted keys | Local model | |
|---|---|---|
| Setup effort | Lowest — just a key | Higher — run an inference server |
| Privacy | Prompts go to the provider | Stays entirely in your network |
| Quality | Strongest | Depends on the model you run |
| Air-gapped | No | Yes |
| Hardware | None | GPU strongly recommended |
Pick hosted keys for the easiest path and best quality; pick a local model when prompts must never leave your network.