From 64d17d5599fc358a870d961c228426d2adab59cf Mon Sep 17 00:00:00 2001
From: Shane <shanehill@mail.com>
Date: Sun, 4 May 2025 10:27:26 +1000
Subject: [PATCH] try

---
 foodie_automator_rss.py |  7 ++++-
 foodie_utils.py         | 70 +++++++++++++++++++++++++++++------------
 2 files changed, 56 insertions(+), 21 deletions(-)

diff --git a/foodie_automator_rss.py b/foodie_automator_rss.py
index 05c5743..7d774ef 100644
--- a/foodie_automator_rss.py
+++ b/foodie_automator_rss.py
@@ -236,13 +236,18 @@ def curate_from_rss():
         print(f"Trying RSS Article: {title} from {source_name}")
         logging.info(f"Trying RSS Article: {title} from {source_name}")
 
-        image_query, relevance_keywords, skip = smart_image_and_filter(title, summary)
+        image_query, relevance_keywords, main_topic, skip = smart_image_and_filter(title, summary)
         if skip:
             print(f"Skipping filtered RSS article: {title}")
             logging.info(f"Skipping filtered RSS article: {title}")
             attempts += 1
             continue
 
+        # Pass main_topic to get_flickr_image for use in fallbacks
+        image_url, image_source, uploader, page_url = get_flickr_image(image_query, relevance_keywords, main_topic)
+        if not image_url:
+            image_url, image_source, uploader, page_url = get_image(image_query)
+
         # Fetch additional context via DDG
         ddg_context = fetch_duckduckgo_news_context(title)
         scoring_content = f"{title}\n\n{summary}\n\nContent: {content}\n\nAdditional Context: {ddg_context}"
diff --git a/foodie_utils.py b/foodie_utils.py
index 6143211..cc25fde 100644
--- a/foodie_utils.py
+++ b/foodie_utils.py
@@ -341,9 +341,10 @@ def smart_image_and_filter(title, summary):
         prompt = (
             "Analyze this article title and summary. Extract key entities (brands, locations, cuisines, or topics) "
             "for an image search about food industry trends or viral content. Prioritize specific terms if present, "
-            "otherwise focus on the main theme. "
+            "otherwise focus on the main theme. Also identify the main topic of the article (e.g., a specific food item or cuisine). "
             "Return 'SKIP' if the article is about home appliances, recipes, promotions, or contains 'homemade', else 'KEEP'. "
-            "Return as JSON with double quotes for all property names and string values (e.g., {\"image_query\": \"specific term\", \"relevance\": [\"keyword1\", \"keyword2\"], \"action\": \"KEEP\" or \"SKIP\"})."
+            "Return as JSON with double quotes for all property names and string values (e.g., "
+            "{\"image_query\": \"specific term\", \"relevance\": [\"keyword1\", \"keyword2\"], \"main_topic\": \"main food item\", \"action\": \"KEEP\" or \"SKIP\"})."
         )
         
         response = client.chat.completions.create(
@@ -357,39 +358,52 @@ def smart_image_and_filter(title, summary):
         raw_result = response.choices[0].message.content.strip()
         logging.info(f"Raw GPT smart image/filter response: '{raw_result}'")
         
-        # Remove ```json markers and fix single quotes in JSON structure
         cleaned_result = re.sub(r'```json\s*|\s*```', '', raw_result).strip()
-        # Replace single quotes with double quotes, but preserve single quotes within string values
         fixed_result = re.sub(r"(?<!\\)'(?=\s*[\w\s]*\])|(?<=\[|\{|\s)'|'(?=\s*[\]\},:])|(?<=\w)'(?=\s*:)", '"', cleaned_result)
         
         try:
             result = json.loads(fixed_result)
         except json.JSONDecodeError as e:
             logging.warning(f"JSON parsing failed: {e}, raw: '{fixed_result}'. Using fallback.")
-            return "food trends", ["cuisine", "dining"], False
+            # Fallback: Extract main topic using simple keyword matching
+            main_topic = extract_main_topic(title.lower() + " " + summary.lower())
+            return main_topic, [main_topic, "food"], False
         
         if not isinstance(result, dict) or "image_query" not in result or "relevance" not in result or "action" not in result:
             logging.warning(f"Invalid GPT response format: {result}, using fallback")
-            return "food trends", ["cuisine", "dining"], False
+            main_topic = extract_main_topic(title.lower() + " " + summary.lower())
+            return main_topic, [main_topic, "food"], False
         
         image_query = result["image_query"]
         relevance_keywords = result["relevance"]
+        main_topic = result.get("main_topic", extract_main_topic(title.lower() + " " + summary.lower()))
         skip_flag = result["action"] == "SKIP" or "homemade" in title.lower() or "homemade" in summary.lower()
         
-        logging.info(f"Smart image query: {image_query}, Relevance: {relevance_keywords}, Skip: {skip_flag}")
+        logging.info(f"Smart image query: {image_query}, Relevance: {relevance_keywords}, Main Topic: {main_topic}, Skip: {skip_flag}")
         
+        specific_single_words = ["kimchi", "sushi", "pizza", "taco", "burger"]
         if not image_query:
             logging.warning(f"Image query is empty, using fallback")
-            return "food trends", ["cuisine", "dining"], skip_flag
-        # Allow single-word queries if they are specific (e.g., food items)
-        specific_single_words = ["kimchi", "sushi", "pizza", "taco", "burger"]  # Add more as needed
+            return main_topic, [main_topic, "food"], skip_flag
         if len(image_query.split()) < 2 and image_query.lower() not in specific_single_words:
             logging.warning(f"Image query '{image_query}' too vague, using fallback")
-        return "food trends", ["cuisine", "dining"], skip_flag
+            return main_topic, [main_topic, "food"], skip_flag
+        
+        return image_query, relevance_keywords, main_topic, skip_flag
     
     except Exception as e:
         logging.error(f"Smart image/filter failed: {e}, using fallback")
-        return "food trends", ["cuisine", "dining"], False
+        main_topic = extract_main_topic(title.lower() + " " + summary.lower())
+        return main_topic, [main_topic, "food"], False
+
+def extract_main_topic(text):
+    # Common food-related keywords (expand as needed)
+    food_keywords = ["kimchi", "sushi", "pizza", "taco", "burger", "ramen", "curry", "pasta", "salad", "soup"]
+    for keyword in food_keywords:
+        if keyword in text:
+            return keyword
+    # Fallback to a generic term if no specific food item is found
+    return "food trends"
 
 def upload_image_to_wp(image_url, post_title, wp_base_url, wp_username, wp_password, image_source="Pixabay", uploader=None, pixabay_url=None):
     try:
@@ -934,15 +948,25 @@ def process_photo(photo, search_query):
             logging.warning(f"Medium size not available for photo {photo.id}: {e}")
             return None
     
-    if not img_url or img_url in used_images:
-        logging.info(f"Image URL invalid or already used for photo {photo.id}: {img_url}")
+    if not img_url:
+        logging.info(f"Image URL invalid for photo {photo.id}")
+        return None
+    
+    # Check if the image is highly relevant to the query
+    query_keywords = set(search_query.lower().split())
+    photo_keywords = set(tags + title.split())
+    is_relevant = bool(query_keywords & photo_keywords)  # Check if any query keyword is in tags or title
+    
+    # Allow reuse of highly relevant images
+    if img_url in used_images and not is_relevant:
+        logging.info(f"Image already used and not highly relevant for photo {photo.id}: {img_url}")
         return None
     
     uploader = photo.owner.username
     page_url = f"https://www.flickr.com/photos/{photo.owner.nsid}/{photo.id}"
     
     used_images.add(img_url)
-    save_used_images()  # This will now save in the correct format
+    save_used_images()
     
     flickr_data = {
         "title": search_query,
@@ -1041,14 +1065,13 @@ def classify_keywords(keywords):
         logging.warning(f"Keyword classification failed: {e}. Defaulting to all specific.")
         return {kw: "specific" for kw in keywords}
 
-def get_flickr_image(search_query, relevance_keywords):
+def get_flickr_image(search_query, relevance_keywords, main_topic):
     global last_flickr_request_time, flickr_request_count
     
     reset_flickr_request_count()
     flickr_request_count += 1
     logging.info(f"Flickr request count: {flickr_request_count}/3600")
     
-    # Enforce a minimum delay of 10 seconds between Flickr requests
     current_time = time.time()
     time_since_last_request = current_time - last_flickr_request_time
     if time_since_last_request < 10:
@@ -1081,7 +1104,6 @@ def get_flickr_image(search_query, relevance_keywords):
         classifications = classify_keywords(keywords)
         logging.info(f"Keyword classifications: {classifications}")
         
-        # Prioritize specific keywords
         specific_keywords = [kw for kw, classification in classifications.items() if classification == "specific"]
         if specific_keywords:
             for keyword in specific_keywords:
@@ -1092,9 +1114,17 @@ def get_flickr_image(search_query, relevance_keywords):
                     if result:
                         return result
 
-    # Step 4: Final fallback using relevance keywords
+    # Step 4: Fallback using main topic
+    logging.info(f"No results found. Falling back to main topic: '{main_topic}'")
+    photos = search_flickr(main_topic)
+    for photo in photos:
+        result = process_photo(photo, main_topic)
+        if result:
+            return result
+
+    # Step 5: Final fallback using relevance keywords
     fallback_query = " ".join(relevance_keywords) if isinstance(relevance_keywords, list) else relevance_keywords
-    logging.info(f"No results found. Falling back to generic query: '{fallback_query}'")
+    logging.info(f"No results with main topic. Falling back to relevance keywords: '{fallback_query}'")
     photos = search_flickr(fallback_query)
     for photo in photos:
         result = process_photo(photo, search_query)