feat: add google_drive/upload_from_file action for token-efficient file uploads

Add a new CrewAIPlatformFileUploadTool that reads files directly from disk
and uploads to Google Drive via the platform API, bypassing the LLM context
window entirely. This solves two problems:

1. File content no longer consumes LLM tokens
2. Binary/large files no longer risk exceeding the 128k context limit

The new action accepts a file_path parameter instead of content. It handles:
- Auto-detection of MIME type from file extension
- Optional custom file name (defaults to local filename)
- File size validation (50 MB limit for simple uploads)
- Base64 encoding of file content before sending to API

The existing google_drive/upload_file action is unchanged — full backwards
compatibility with the 10k+ existing executions.

Changes:
- New tool: CrewAIPlatformFileUploadTool
- Builder auto-injects the tool when apps include google_drive or
  google_drive/upload_from_file
- 14 unit tests covering upload, error handling, MIME detection, SSL
- Updated docs with new action reference and usage examples

docs/en/enterprise/integrations/google_drive.mdx

@@ -86,6 +86,22 @@ CREWAI_PLATFORM_INTEGRATION_TOKEN=your_enterprise_token
 
   </Accordion>
 
+  <Accordion title="google_drive/upload_from_file">
+    **Description:** Upload a file from a local path to Google Drive. The file is read directly from disk — its content never passes through the LLM context window, making this efficient for large or binary files.
+
+    <Note>
+      Unlike `google_drive/upload_file` which requires passing file content as a parameter (consuming LLM tokens), this action takes a file path and reads the file directly. Use this for binary files, large documents, or any case where you want to avoid token overhead.
+    </Note>
+
+    **Parameters:**
+    - `file_path` (string, required): Path to the local file to upload.
+    - `name` (string, optional): Name for the file in Google Drive (defaults to the local filename).
+    - `mime_type` (string, optional): MIME type of the file (auto-detected from file extension if not provided).
+    - `parent_folder_id` (string, optional): ID of the parent folder where the file should be created.
+    - `description` (string, optional): Description of the file.
+
+  </Accordion>
+
   <Accordion title="google_drive/download_file">
     **Description:** Download a file from Google Drive.
 
@@ -205,6 +221,42 @@ crew = Crew(
 crew.kickoff()
 ```
 
+### Uploading Files from Disk (Token-Efficient)
+
+```python
+from crewai import Agent, Task, Crew
+
+# Use upload_from_file to avoid passing file content through the LLM context
+upload_agent = Agent(
+    role="File Uploader",
+    goal="Upload local files to Google Drive efficiently",
+    backstory="An AI assistant that uploads files without wasting context tokens.",
+    apps=[
+        'google_drive/upload_from_file',
+        'google_drive/list_files'
+    ]
+)
+
+# The agent only needs to specify the file path — the file is read directly
+# from disk and never passes through the LLM context window
+upload_task = Task(
+    description="Upload the file at /data/reports/quarterly_report.pdf to Google Drive in the Reports folder",
+    agent=upload_agent,
+    expected_output="File uploaded successfully with file ID"
+)
+
+crew = Crew(
+    agents=[upload_agent],
+    tasks=[upload_task]
+)
+
+crew.kickoff()
+```
+
+<Tip>
+  Use `google_drive/upload_from_file` instead of `google_drive/upload_file` when uploading existing files from disk. This is especially important for binary files (PDFs, images, etc.) or large files that would otherwise consume significant LLM context tokens or exceed context limits.
+</Tip>
+
 ### Advanced File Management
 
 ```python