What signatures and input examples do when logging a model.
From Ultra Transcenders DBX-GENAI by Tony Rough (coming December 2026)
A signature is more than documentation, because Unity Catalog enforces its presence. New model versions in Unity Catalog must have a model signature. There are several ways to get one, and the interface you author against largely decides which.
| Route | How the signature appears | Notes |
|---|---|---|
ResponsesAgent interface |
MLflow infers a valid signature automatically | The interface the Model Serving logging page recommends; compatible with Databricks product features |
input_example on log_model |
Inferred from the example and its predicted output | MLflow 2.5.0 and above; the example must be a valid input to the agent |
| Manual definition | You define it at logging time | Most control, most effort |
Missing signatures have specific costs, stated on the Unity Catalog model page for versions that lack one:
The endpoint’s OpenAPI schema (in Public Preview) also needs a logged signature. For agents, product features assume one of a set of supported signatures, so a hand-rolled schema can leave an agent incompatible with them.
Common trap: Believing an input example is only sample data for the UI - with MLflow 2.5.0 and above, the example is what
log_modeluses to infer the signature that Unity Catalog requires. Omit both and registration of the new version fails its signature requirement.
This note is one section of Ultra Transcenders DBX-GENAI: Databricks Generative AI Engineer Associate, an independent study guide that explains every topic the exam covers by technology, with comparison tables, diagrams and the common traps, plus a glossary linked to the Databricks documentation.
Due on Amazon in December 2026, in Kindle and paperback editions.
About the book · DBX-GENAI terms in the glossary · All DBX-GENAI study notes
Which chunking method suits which kind of document.
What decides between the two endpoint types for an index.
What a reranker adds to retrieval, and what it costs.
How to choose a framework for a generative AI application.
Which workloads suit ai_query(), and how to run them.
The built-in judges that need expected answers, and those that don't.
How the evaluation and monitoring phases of a generative AI application differ.