Comprehensive Bug Report: Google Workspace Extension Drive Integration Failure & Quota Depletion
Component: Gemini Web UI / Google Workspace Extensions (Google Drive API Integration)
Title: Non-Existent Write Functions (drive:create_file) Combined with One-Way Directory Access Cause Workflow Failures and Unwarranted Quota Depletion
Status/Priority: P2 / S2 (Severe impact to professional/legal workflows / Quota waste)
Description
A critical architectural and synchronization failure exists within the Gemini Workspace Extension for Google Drive. The environment exposes directory-discovery tools (drive:search_files), creating a valid operational assumption that the model has full read/write access to user storage. However, when the model attempts to execute file creation or saving tasks, the backend throws a hard failure: The following function call(s) are invalid and do not exist: drive:create_file with an INVALID_ARGUMENT status.
Furthermore, when this extension call fails, the model is trapped in a systemic blind spot: it cannot silently write files to disk, nor can it read back previously generated project documents from specialized folders unless manually re-uploaded. This forces conversational error loops, manual text-dump workarounds, and extensive re-processing of context, which rapidly exhausts the user’s rolling 5-hour compute quota on infrastructure errors.
Steps to Reproduce
-
Establish a structured project workspace in Google Drive containing specific parent and child directories (e.g., a root folder and subfolders for evidence and logs).
-
Instruct the Gemini Advanced/Pro model to ingest complex, multi-source textual data (e.g., multi-part legal email logs spanning hundreds of records) and compile a master document.
-
Apply strict operational directives requiring the model to autonomously generate standalone files and save them directly into designated Drive subfolders (
Master Rev OandBuild chats) without manual user intervention. -
Observe the model successfully execute directory lookups via
drive:search_files, confirming folder existence and IDs. -
Trigger the file creation step, resulting in the backend rejection:
The following function call(s) are invalid and do not exist: drive:create_file(INVALID_ARGUMENT). -
Observe the cascading failure: the model drops its automated workflow state, reverts to generating massive raw text blocks in the chat UI for manual user copying (directly violating procedural directives), and burns compute tokens attempting to self-correct around a non-existent API.
Expected Result
-
Parity in Drive Tools: If the workspace extension supports directory searching (
drive:search_files), it must natively support a corresponding, fully implemented file creation and writing function (drive:create_fileor equivalent secure storage hook) so automated document management workflows can function as advertised. -
Graceful Degradation: If file writing is intentionally restricted in the web sandbox, the tool schema should omit write functions entirely rather than exposing a ghost API that traps the model in an execution loop.
-
Quota Protection: System errors resulting from
INVALID_ARGUMENTtool schema mismatches should be intercepted at the gateway and excluded from the user’s rolling compute limit.
Actual Result
The user experiences a complete breakdown of automation. The model is penalized with quota depletion for attempting to use extension tools that are fundamentally broken or missing on the backend. Users engaged in intensive professional tasks (such as drafting legal briefs or managing codebases) are forced to perform tedious manual file management, defeating the entire value proposition of the Workspace Extension integration.
Requested Fix
-
Full API Implementation: Deploy and properly register the
drive:create_file(and companion update/write endpoints) within the active Gemini Workspace extension tool schema so models can securely write directly to designated user folders. -
Read/Write Parity Audit: Audit all Workspace extension schemas to ensure that directory listing and file creation capabilities are symmetrically supported.
-
Metering Exemption: Implement automatic token refunds or metering exemptions for inference cycles consumed during unhandled extension errors (
INVALID_ARGUMENT/ missing function declarations).