Use ubuntu-22.04-large runner for build job

.github/workflows/build-deploy.yml

@@ -54,10 +54,21 @@ jobs:
           --health-timeout 5s
           --health-retries 5
 
-    runs-on: ubuntu-latest
-    steps:
-      - name: pulling git repo
-        uses: actions/checkout@v2
+    runs-on: ubuntu-22.04-large
+    steps:
+      - name: Reclaim disk space on runner
+        shell: bash
+        run: |
+          echo "==> Removing unused SDKs and cached archives to free space"
+          sudo rm -rf /usr/share/dotnet
+          sudo rm -rf /usr/local/lib/android
+          sudo rm -rf /opt/ghc
+          sudo rm -rf /usr/local/share/boost
+          sudo rm -rf /home/linuxbrew/.linuxbrew
+          docker system prune --all --force --volumes || true
+          df -h
+      - name: pulling git repo
+        uses: actions/checkout@v2
 
       - name: Create RSA private key file
         run: echo "${{ secrets.RSA_PRIVATE_KEY }}" > private_key.pem

.gitignore

@@ -14,3 +14,4 @@ public_key
 private_key
 env.
 venv
+test_app.db

app/analytics.py

@@ -1,31 +1,42 @@
-from sqlalchemy import func
-from datetime import (
-    datetime,
-    timedelta,
-    timezone,
-)  # Added timezone for correct UTC usage
-from . import models
-from transformers import pipeline, AutoTokenizer, AutoModelForSequenceClassification
-import torch
-from .config import settings
-import matplotlib.pyplot as plt
-import seaborn as sns
-import io
-import base64
-from .models import SearchStatistics, User
-from sqlalchemy.orm import Session
-
-# NOTE: Ensure that get_db() is defined in your project or import it accordingly.
-# from .database import get_db
-
-# Initialize the tokenizer and model for sentiment analysis
-tokenizer = AutoTokenizer.from_pretrained(
-    "distilbert-base-uncased-finetuned-sst-2-english"
-)
-model = AutoModelForSequenceClassification.from_pretrained(
-    "distilbert-base-uncased-finetuned-sst-2-english"
-)
-sentiment_pipeline = pipeline("sentiment-analysis", model=model, tokenizer=tokenizer)
+from sqlalchemy import func
+from datetime import (
+    datetime,
+    timedelta,
+    timezone,
+)  # Added timezone for correct UTC usage
+import logging
+from . import models
+from transformers import pipeline, AutoTokenizer, AutoModelForSequenceClassification
+from .config import settings
+import matplotlib.pyplot as plt
+import seaborn as sns
+import io
+import base64
+from .models import SearchStatistics, User
+from sqlalchemy.orm import Session
+from .utils import keyword_sentiment
+
+# NOTE: Ensure that get_db() is defined in your project or import it accordingly.
+# from .database import get_db
+
+logger = logging.getLogger(__name__)
+
+# Initialize the tokenizer and model for sentiment analysis with graceful fallback
+try:
+    tokenizer = AutoTokenizer.from_pretrained(
+        "distilbert-base-uncased-finetuned-sst-2-english"
+    )
+    model = AutoModelForSequenceClassification.from_pretrained(
+        "distilbert-base-uncased-finetuned-sst-2-english"
+    )
+    sentiment_pipeline = pipeline("sentiment-analysis", model=model, tokenizer=tokenizer)
+except Exception as exc:  # pragma: no cover - defensive fallback
+    logger.warning(
+        "Analytics sentiment pipeline unavailable, using keyword heuristic: %s",
+        exc,
+    )
+    tokenizer = model = None
+    sentiment_pipeline = None
 
 # ------------------------- Content Analysis Functions -------------------------
 
@@ -35,8 +46,17 @@ def analyze_sentiment(text):
     Analyze the sentiment of the given text using a pre-trained transformer model.
     Returns a dictionary with sentiment label and score.
     """
-    result = sentiment_pipeline(text)[0]
-    return {"sentiment": result["label"], "score": result["score"]}
+    if sentiment_pipeline is not None:
+        try:
+            result = sentiment_pipeline(text)[0]
+            return {"sentiment": result["label"], "score": float(result["score"])}
+        except Exception as exc:  # pragma: no cover - defensive fallback
+            logger.warning(
+                "Analytics sentiment pipeline failed, using keyword heuristic: %s",
+                exc,
+            )
+    label, score = keyword_sentiment(text)
+    return {"sentiment": label, "score": score}
 
 
 def suggest_improvements(text, sentiment):