Gemini models (3.7, 3.8) execution stalls after 25s+ – Outputting static reports instead of executing tool/code actions

Gemini API
, ,
1

Summary of Issue

Across multiple Gemini models (3.7 Flash, 3.8), the execution pipeline stalls. The model runs for 25–30+ seconds, exhausts tokens/quota, but only prints static text reports and summaries instead of executing the actual tool calls, file edits, and actions.

Problem Details

Instead of completing downstream automated execution, the system pauses at descriptive reports. Re-running triggers the same loop: prolonged processing time followed by text summaries with no actual execution.

Exact Failed Operations & Artifact Logs

  1. View file: /app/src/main/java/com/example/ui/screens/MainApp.kt (lines 30-110)
  2. View file: /app/src/main/java/com/example/ui/screens/home/DashboardScreen.kt (lines 220-280)
  3. Create file: /app/src/main/java/com/example/data/google/GoogleAuthStateManager.kt
  4. Edit file: /app/src/main/java/com/example/ui/screens/MainApp.kt (fullscreen state handling)
  5. Edit file: /app/src/main/java/com/example/ui/screens/home/DashboardScreen.kt (visibility toggles)
  6. View & Edit: /app/src/main/java/com/example/ui/screens/home/GoogleBusinessDashboardScreen.kt (lines 170-240, 580-710, 1220-1399)
  7. View & Edit: /app/src/main/java/com/example/data/google/GoogleDriveSyncManager.kt (lines 1-50)

What is Needed

  1. Why are models across versions (3.7, 3.8) defaulting to report generation instead of executing tool calls?
  2. What specific parameter or schema configuration prevents this fallback to text reports?
  3. How to fix this pipeline stall so operations execute completely?
2

Hello @gkeg.in ,

Are you executing your code edit through the Gemini API with custom function calling, or through the AI Studio Build interface or some other coding agent? If using the API, please share your tools, tool_config and request loop.