To scan with Giskard OSS, wrap your model so it accepts a DataFrame and returns one output per row, and give the Model an informative description plus complete feature_names. For RAG evaluation, generate the test set from your knowledge base with generate_testset instead of hand-writing questions.

Context: Official Giskard OSS repo (README): documents the wrap-your-model contract. Install with the llm extra, e.g. pip install giskard[llm], pinned to a major version. Your prediction function takes a pandas DataFrame and must return a list of outputs, one per row. Wrap it in giskard.Model with model_type text_generation, a name, an informative description, and feature_names. For RAG apps, generate_testset builds evaluation questions, reference answers, and context from your knowledge base via KnowledgeBase.

## Matched source
Source: Published skill
Original query: "Giskard OSS: prediction fn takes DataFrame, returns one output per row"
Key terms: dataframe, giskard, output, prediction, returns, takes
