feat: add spreadsheet mime types to load_artifact_tool

pyproject.toml

@@ -55,6 +55,10 @@ dependencies = [
   "opentelemetry-exporter-otlp-proto-http>=1.36.0",
   "opentelemetry-resourcedetector-gcp>=1.9.0a0, <2.0.0",
   "opentelemetry-sdk>=1.36.0, <1.40.0",
+  "openpyxl>=3.1.0",                                         # For spreadsheet parsing
+  "pandas>=2.2.3",                                           # For spreadsheet parsing
+  "odfpy>=1.4.1",                                            # For spreadsheet parsing
+  "xlrd>=2.0.1",                                             # For spreadsheet parsing
   "pyarrow>=14.0.0",
   "pydantic>=2.0, <3.0.0",                                   # For data validation/models
   "python-dateutil>=2.9.0.post0, <3.0.0",                    # For Vertext AI Session Service
@@ -63,6 +67,7 @@ dependencies = [
   "sqlalchemy-spanner>=1.14.0",                              # Spanner database session service
   "sqlalchemy>=2.0, <3.0.0",                                 # SQL database ORM
   "starlette>=0.49.1, <1.0.0",                               # For FastAPI CLI
+  "tabulate>=0.9.0",                                         # For spreadsheet parsing
   "tenacity>=9.0.0, <10.0.0",                                # For Retry management
   "typing-extensions>=4.5, <5",
   "tzlocal>=5.3, <6.0",                                      # Time zone utilities

src/google/adk/tools/load_artifacts_tool.py

@@ -28,6 +28,9 @@
 from ..features import is_feature_enabled
 from .base_tool import BaseTool
 
+import io
+import pandas as pd
+
 # MIME types Gemini accepts for inline data in requests.
 _GEMINI_SUPPORTED_INLINE_MIME_PREFIXES = (
     'image/',
@@ -40,6 +43,11 @@
     'application/json',
     'application/xml',
 })
+_SPREADSHEET_MIME_TYPES = frozenset({
+    'application/vnd.openxmlformats-officedocument.spreadsheetml.sheet', # .xlsx
+    'application/vnd.ms-excel', # .xls
+    'application/vnd.oasis.opendocument.spreadsheet', # .ods
+})
 
 if TYPE_CHECKING:
   from ..models.llm_request import LlmRequest
@@ -76,6 +84,35 @@ def _maybe_base64_to_bytes(data: str) -> bytes | None:
       return None
 
 
+def _parse_spreadsheet(data: bytes, mime_type: str) -> str:
+  """Parses a spreadsheet into a markdown representation."""
+  try:
+    # Load the spreadsheet
+    xl = pd.ExcelFile(io.BytesIO(data))
+
+    output = []
+
+    # Process each sheet
+    for sheet_name in xl.sheet_names:
+      df = xl.parse(sheet_name)
+      if df.empty:
+        continue
+
+      # Convert to markdown table
+      markdown_table = df.to_markdown(index=False, numalign="left", stralign="left")
+
+      output.append(f"### Sheet: {sheet_name}\n\n{markdown_table}")
+
+    if not output:
+      return "[Empty Spreadsheet]"
+
+    return "\n\n".join(output)
+
+  except Exception as e:
+    logger.warning(f"Failed to parse spreadsheet: {e}")
+    return f"[Error parsing spreadsheet: {e}]"
+
+
 def _as_safe_part_for_llm(
     artifact: types.Part, artifact_name: str
 ) -> types.Part:
@@ -111,6 +148,10 @@ def _as_safe_part_for_llm(
     except UnicodeDecodeError:
       return types.Part.from_text(text=data.decode('utf-8', errors='replace'))
 
+  if mime_type in _SPREADSHEET_MIME_TYPES:
+    text_content = _parse_spreadsheet(data, mime_type)
+    return types.Part.from_text(text=text_content)
+
   size_kb = len(data) / 1024
   return types.Part.from_text(
       text=(

tests/unittests/tools/test_load_artifacts_tool.py

@@ -180,3 +180,65 @@ def test_get_declaration_with_json_schema_feature_enabled():
           },
       },
   }
+
+
+@mark.asyncio
+async def test_load_artifacts_parses_spreadsheet():
+  """Spreadsheet artifacts are parsed into markdown."""
+  artifact_name = 'test.xlsx'
+
+  import pandas as pd
+  import io
+
+  df = pd.DataFrame({'col1': [1, 2], 'col2': ['a', 'b']})
+  output = io.BytesIO()
+  # Use openpyxl as engine since it is in deps
+  with pd.ExcelWriter(output, engine='openpyxl') as writer:
+    df.to_excel(writer, sheet_name='Sheet1', index=False)
+  xlsx_bytes = output.getvalue()
+
+  artifact = types.Part(
+      inline_data=types.Blob(
+          data=xlsx_bytes, 
+          mime_type='application/vnd.openxmlformats-officedocument.spreadsheetml.sheet'
+      )
+  )
+
+  tool_context = _StubToolContext({artifact_name: artifact})
+  llm_request = LlmRequest(
+      contents=[
+          types.Content(
+              role='user',
+              parts=[
+                  types.Part(
+                      function_response=types.FunctionResponse(
+                          name='load_artifacts',
+                          response={'artifact_names': [artifact_name]},
+                      )
+                  )
+              ],
+          )
+      ]
+  )
+
+  await load_artifacts_tool.process_llm_request(
+      tool_context=tool_context, llm_request=llm_request
+  )
+
+  artifact_part = llm_request.contents[-1].parts[1]
+  assert artifact_part.inline_data is None
+
+  # Check for Markdown table content
+  # We expect something like:
+  # ### Sheet: Sheet1
+  #
+  # | col1 | col2 |
+  # | :--- | :--- |
+  # | 1    | a    |
+  # | 2    | b    |
+
+  assert "Sheet1" in artifact_part.text
+  assert "col1" in artifact_part.text
+  assert "col2" in artifact_part.text
+  assert "1" in artifact_part.text
+  assert "a" in artifact_part.text