Fix delete_revisions import with fallback cache cleanup

- Handle missing delete_revisions function gracefully
- Add manual cache directory deletion as fallback
- Improve error handling for cache cleanup

Dockerfile

@@ -15,7 +15,11 @@ WORKDIR /app
 COPY requirements.txt .
 
 # Install Python dependencies
-RUN pip install --no-cache-dir -r requirements.txt
+# Note: flash-attn requires CUDA for compilation, so skip it for CPU-only Cloud Run deployments
+# The app gracefully handles missing flash-attn (see app.py FLASH_ATTN_AVAILABLE check)
+RUN sed '/^flash-attn/d' requirements.txt > requirements-no-flash.txt && \
+    pip install --no-cache-dir -r requirements-no-flash.txt && \
+    echo "Installed dependencies (skipped flash-attn - requires CUDA)"
 
 # Copy application code
 COPY . .

MANUAL_DEPLOY.md

@@ -0,0 +1,54 @@
+# Manual Deployment Steps (Cloud Console)
+
+Since there are permission propagation issues, here's how to deploy manually via Cloud Console:
+
+## Step 1: Enable APIs via Cloud Console
+
+1. **Cloud Build API**: https://console.cloud.google.com/apis/library/cloudbuild.googleapis.com?project=spherical-gate-477614-q7
+   - Click "Enable"
+
+2. **Cloud Run API**: https://console.cloud.google.com/apis/library/run.googleapis.com?project=spherical-gate-477614-q7
+   - Click "Enable"
+
+3. **Container Registry API**: https://console.cloud.google.com/apis/library/containerregistry.googleapis.com?project=spherical-gate-477614-q7
+   - Click "Enable"
+
+## Step 2: Create Cloud Build Trigger (Optional)
+
+Or use Cloud Shell:
+
+```bash
+# In Cloud Shell
+cd Milestone-6/router-agent/zero-gpu-space
+gcloud builds submit --tag gcr.io/spherical-gate-477614-q7/router-agent:latest .
+```
+
+## Step 3: Deploy to Cloud Run
+
+After image is built:
+
+```bash
+gcloud run deploy router-agent \
+    --image gcr.io/spherical-gate-477614-q7/router-agent:latest \
+    --platform managed \
+    --region us-central1 \
+    --allow-unauthenticated \
+    --port 7860 \
+    --memory 8Gi \
+    --cpu 4 \
+    --timeout 3600 \
+    --set-env-vars "GRADIO_SERVER_NAME=0.0.0.0,GRADIO_SERVER_PORT=7860"
+```
+
+## Alternative: Use Cloud Shell
+
+Cloud Shell has proper permissions. Open:
+https://shell.cloud.google.com/?project=spherical-gate-477614-q7
+
+Then run:
+```bash
+git clone https://github.com/ALikesToCode/CourseGPT-Pro-DSAI-Lab-Group-6.git
+cd CourseGPT-Pro-DSAI-Lab-Group-6/Milestone-6/router-agent/zero-gpu-space
+./deploy-cloud-build.sh
+```
+

QUICK_DEPLOY.md

@@ -1,104 +0,0 @@
-# Quick Deploy to Google Cloud Run
-
-## Current Status
-
-**Issue**: Permission denied on project `light-quest-475608-k7`
-
-## Quick Fix Options
-
-### Option 1: Fix Permissions (Recommended)
-
-1. **Grant yourself permissions** on the existing project:
-   - Visit: https://console.cloud.google.com/iam-admin/iam/project?project=light-quest-475608-k7
-   - Add role: `Editor` or `Owner` to your account (`jameswilsonlearnsrocode@gmail.com`)
-   - Wait 2-3 minutes for propagation
-
-2. **Enable APIs**:
-   ```bash
-   gcloud services enable cloudbuild.googleapis.com run.googleapis.com containerregistry.googleapis.com --project=light-quest-475608-k7
-   ```
-
-3. **Deploy**:
-   ```bash
-   cd Milestone-6/router-agent/zero-gpu-space
-   ./deploy-cloud-build.sh
-   ```
-
-### Option 2: Create New Project
-
-```bash
-# Run setup helper
-./setup-gcp-permissions.sh
-
-# Or manually:
-gcloud projects create router-agent-deploy --name="Router Agent Deployment"
-gcloud config set project router-agent-deploy
-gcloud services enable cloudbuild.googleapis.com run.googleapis.com containerregistry.googleapis.com
-
-# Then deploy
-./deploy-cloud-build.sh
-```
-
-### Option 3: Use Different Account/Project
-
-If you have access to another project:
-
-```bash
-export GCP_PROJECT_ID="your-other-project-id"
-gcloud config set project ${GCP_PROJECT_ID}
-./deploy-cloud-build.sh
-```
-
-## Manual Deployment Steps
-
-If scripts don't work, deploy manually:
-
-```bash
-# 1. Set project
-gcloud config set project YOUR_PROJECT_ID
-
-# 2. Enable APIs (if not already enabled)
-gcloud services enable cloudbuild.googleapis.com run.googleapis.com containerregistry.googleapis.com
-
-# 3. Submit build
-cd Milestone-6/router-agent/zero-gpu-space
-gcloud builds submit --config=cloudbuild.yaml .
-
-# 4. Or build and deploy separately
-gcloud builds submit --tag gcr.io/YOUR_PROJECT_ID/router-agent:latest .
-gcloud run deploy router-agent \
-    --image gcr.io/YOUR_PROJECT_ID/router-agent:latest \
-    --platform managed \
-    --region us-central1 \
-    --allow-unauthenticated \
-    --port 7860 \
-    --memory 8Gi \
-    --cpu 4
-```
-
-## Required Permissions
-
-Your account needs these roles:
-- `roles/cloudbuild.builds.editor` - To build images
-- `roles/run.admin` - To deploy to Cloud Run
-- `roles/serviceusage.serviceUsageConsumer` - To enable APIs
-- `roles/storage.admin` - To push to Container Registry
-
-Or use `roles/editor` or `roles/owner` for full access.
-
-## Troubleshooting
-
-**Permission Denied**: 
-- Check IAM: https://console.cloud.google.com/iam-admin/iam
-- Ensure billing is enabled
-- Wait 2-3 minutes after granting permissions
-
-**API Not Enabled**:
-- Enable manually: https://console.cloud.google.com/apis/library
-- Or use: `gcloud services enable <api-name>`
-
-**Build Fails**:
-- Check logs: `gcloud builds list` then `gcloud builds log BUILD_ID`
-- Ensure Dockerfile is correct
-- Check requirements.txt for compatibility
-

test_space_simple.sh

@@ -0,0 +1,48 @@
+#!/bin/bash
+# Simple test script for Hugging Face Space (no Python dependencies)
+
+SPACE_URL="https://alovestocode-zerogpu-llm-inference.hf.space"
+
+echo "=========================================="
+echo "Testing Hugging Face Space: ZeroGPU"
+echo "=========================================="
+echo ""
+
+echo "1. Checking Space accessibility..."
+HTTP_CODE=$(curl -s -o /dev/null -w "%{http_code}" "$SPACE_URL")
+if [ "$HTTP_CODE" == "200" ]; then
+    echo "   ✅ Space is accessible (HTTP $HTTP_CODE)"
+else
+    echo "   ❌ Space returned HTTP $HTTP_CODE"
+    exit 1
+fi
+
+echo ""
+echo "2. Checking API info endpoint..."
+API_INFO=$(curl -s "$SPACE_URL/api/info" 2>&1)
+if echo "$API_INFO" | grep -q "api_name\|endpoints"; then
+    echo "   ✅ API info endpoint accessible"
+    echo "$API_INFO" | python3 -m json.tool 2>/dev/null | head -20 || echo "   (Raw response received)"
+else
+    echo "   ⚠️  API info response: $API_INFO"
+fi
+
+echo ""
+echo "3. Checking Space status..."
+STATUS=$(curl -s "https://huggingface.co/api/spaces/Alovestocode/ZeroGPU-LLM-Inference" 2>&1 | python3 -c "import sys, json; data=json.load(sys.stdin); print(data.get('runtime', {}).get('stage', 'UNKNOWN'))" 2>/dev/null || echo "UNKNOWN")
+echo "   Space Status: $STATUS"
+
+echo ""
+echo "=========================================="
+if [ "$HTTP_CODE" == "200" ]; then
+    echo "✅ Basic connectivity test PASSED"
+    echo ""
+    echo "Space URL: $SPACE_URL"
+    echo "To test the API with Python, install gradio_client:"
+    echo "  pip install gradio_client"
+    echo "  python test_api_gradio_client.py"
+else
+    echo "❌ Basic connectivity test FAILED"
+fi
+echo "=========================================="
+