| Term | Meaning in these guides |
|---|---|
| Repository / repo | A folder of project files with a recorded history on GitHub. |
| Clone | Download a Git-managed copy that can receive later updates. A ZIP also works for following a guide. |
| Terminal | A window where you type commands. |
| Virtual environment | A separate set of Python packages for one project. |
| Dependency | A package the example needs to run. requirements.txt lists ours. |
| Model / weights | The learned numerical data used to turn inputs into predictions. Usually downloaded separately. |
| Inference | Running a trained model on a new input. |
| CPU / GPU | Computer processors. These examples use CPU inference; the optional separate LLM may have other requirements. |
| JSON | A text format for named values and lists; it is how we save inspectable results. |
| Tool call | A proposed function name and its arguments. Printing one does not execute it. |
| Confidence / probability | A model-reported score. A high value does not guarantee the answer is right. |
| Checkpoint revision / hash | An identifier used to tell exactly which model or file was used. |
| Loopback / localhost | A network address that points back to your own computer. |