Merge branch 'main' into fix/pydantic-v2-swagger-ui

Author: mmphego

src/google/adk/models/lite_llm.py

@@ -81,10 +81,19 @@
 }
 
 _SUPPORTED_FILE_CONTENT_MIME_TYPES = set(
-    ["application/pdf", "application/json", "text/plain"]
+    ["application/pdf", "application/json"]
 )
 
 
+def _decode_inline_text_data(raw_bytes: bytes) -> str:
+  """Decodes inline file bytes that represent textual content."""
+  try:
+    return raw_bytes.decode("utf-8")
+  except UnicodeDecodeError:
+    logger.debug("Falling back to latin-1 decoding for inline file bytes.")
+    return raw_bytes.decode("latin-1", errors="replace")
+
+
 class ChatCompletionFileUrlObject(TypedDict, total=False):
   file_data: str
   file_id: str
@@ -371,6 +380,15 @@ def _get_content(
         and part.inline_data.data
         and part.inline_data.mime_type
     ):
+      if part.inline_data.mime_type.startswith("text/"):
+        decoded_text = _decode_inline_text_data(part.inline_data.data)
+        if len(parts) == 1:
+          return decoded_text
+        content_objects.append({
+            "type": "text",
+            "text": decoded_text,
+        })
+        continue
       base64_string = base64.b64encode(part.inline_data.data).decode("utf-8")
       data_uri = f"data:{part.inline_data.mime_type};base64,{base64_string}"
       # LiteLLM providers extract the MIME type from the data URI; avoid
@@ -397,7 +415,10 @@ def _get_content(
             "file": {"file_data": data_uri},
         })
       else:
-        raise ValueError("LiteLlm(BaseLlm) does not support this content part.")
+        raise ValueError(
+            "LiteLlm(BaseLlm) does not support content part with MIME type "
+            f"{part.inline_data.mime_type}."
+        )
     elif part.file_data and part.file_data.file_uri:
       file_object: ChatCompletionFileUrlObject = {
           "file_id": part.file_data.file_uri,

tests/unittests/models/test_litellm.py

@@ -109,12 +109,6 @@
         "data:application/json;base64,eyJoZWxsbyI6IndvcmxkIn0=",
         id="json",
     ),
-    pytest.param(
-        b"hello world",
-        "text/plain",
-        "data:text/plain;base64,aGVsbG8gd29ybGQ=",
-        id="txt",
-    ),
 ]
 
 STREAMING_MODEL_RESPONSE = [
@@ -1477,6 +1471,38 @@ def test_get_content_text():
   assert content == "Test text"
 
 
+def test_get_content_text_inline_data_single_part():
+  parts = [
+      types.Part.from_bytes(
+          data="Inline text".encode("utf-8"), mime_type="text/plain"
+      )
+  ]
+  content = _get_content(parts)
+  assert content == "Inline text"
+
+
+def test_get_content_text_inline_data_multiple_parts():
+  parts = [
+      types.Part.from_bytes(
+          data="First part".encode("utf-8"), mime_type="text/plain"
+      ),
+      types.Part.from_text(text="Second part"),
+  ]
+  content = _get_content(parts)
+  assert content[0]["type"] == "text"
+  assert content[0]["text"] == "First part"
+  assert content[1]["type"] == "text"
+  assert content[1]["text"] == "Second part"
+
+
+def test_get_content_text_inline_data_fallback_decoding():
+  parts = [
+      types.Part.from_bytes(data=b"\xff", mime_type="text/plain"),
+  ]
+  content = _get_content(parts)
+  assert content == "ÿ"
+
+
 def test_get_content_image():
   parts = [
       types.Part.from_bytes(data=b"test_image_data", mime_type="image/png")