Make sure Chat.messages(format='google') converts role assistant ->…

… model (#1622)

CHANGELOG.md

@@ -15,9 +15,11 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Bug fixes
 
-* An empty `ui.input_date()` value no longer crashes Shiny. (#1528)
+* A handful of fixes for `ui.Chat()`, including:
+  * A fix for use inside Shiny modules. (#1582)
+  * `.messages(format="google")` now returns the correct role. (#1622)
 
-* `ui.Chat()` now works as expected inside Shiny modules. (#1582)
+* An empty `ui.input_date()` value no longer crashes Shiny. (#1528)
 
 * Fixed bug where calling `.update_filter(None)` on a data frame renderer did not visually reset non-numeric column filters. (It did reset the column's filtering, just not the label). Now it resets filter's label. (#1557)
 

shiny/ui/_chat_provider_types.py

@@ -78,11 +78,15 @@ def as_google_message(message: ChatMessage) -> "GoogleMessage":
 
     import google.generativeai.types as gtypes  # pyright: ignore[reportMissingTypeStubs]
 
-    if message["role"] == "system":
+    role = message["role"]
+
+    if role == "system":
         raise ValueError(
             "Google requires a system prompt to be specified in the `GenerativeModel()` constructor."
         )
-    return gtypes.ContentDict(parts=[message["content"]], role=message["role"])
+    elif role == "assistant":
+        role = "model"
+    return gtypes.ContentDict(parts=[message["content"]], role=role)
 
 
 def as_langchain_message(message: ChatMessage) -> "LangChainMessage":