update flask rou

data/get_best_alpha.py

@@ -0,0 +1,80 @@
+import requests
+import json
+from collections import defaultdict
+
+# Load questions from the JSON file
+with open('data/questions_example.json', 'r', encoding='utf-8') as file:
+    questions = json.load(file)['questions']
+
+# Load ground truth data
+with open('data/ground_truths_example.json', 'r', encoding='utf-8') as f:
+    ground_truths = json.load(f)["ground_truths"]
+
+# Dictionary to hold the best alpha and accuracy
+best_alpha = 0.0
+best_accuracy = 0
+
+# Loop through alpha values from 0.0 to 1.0
+for alpha in [round(x * 0.1, 1) for x in range(11)]:
+    output_data = {"answers": []}  # Reset output format with "answers" array
+
+    url = "http://127.0.0.1:5000/api/chat"
+
+    # Send each question to the API with the current alpha
+    for question in questions:
+        # Add the alpha key to the question payload
+        question_with_alpha = {**question, "alpha": alpha}
+
+        # Send POST request
+        response = requests.post(url, json=question_with_alpha)
+
+        if response.status_code == 200:
+            response_json = response.json()
+            qid = question.get("qid")
+            retrieve = response_json.get("retrieve")
+
+            # Append formatted result to the answers array
+            output_data["answers"].append({
+                "qid": qid,
+                "retrieve": retrieve
+            })
+        else:
+            print(f"請求失敗，狀態碼: {response.status_code}，Alpha 值: {alpha}")
+
+    # Save predictions for the current alpha
+    pred_file = f'data/pred_retrieve_alpha_{alpha}.json'
+    with open(pred_file, 'w', encoding='utf-8') as output_file:
+        json.dump(output_data, output_file, ensure_ascii=False, indent=4)
+
+    # Load predictions for comparison
+    pred_dict = {item["qid"]: item["retrieve"] for item in output_data["answers"]}
+
+    # Initialize counters and data structures for accuracy calculation
+    correct_count = 0
+    category_counts = defaultdict(lambda: {"correct": 0, "total": 0})
+
+    # Compare predictions to ground truth
+    for ground in ground_truths:
+        qid = ground["qid"]
+        category = ground["category"]
+        correct_retrieve = ground["retrieve"]
+        predicted_retrieve = pred_dict.get(qid)
+
+        if predicted_retrieve == correct_retrieve:
+            correct_count += 1
+            category_counts[category]["correct"] += 1
+
+        category_counts[category]["total"] += 1
+
+    # Calculate accuracy for the current alpha
+    accuracy = correct_count / len(ground_truths)
+    print("Corrrect count: ", correct_count)
+    print(f"Alpha: {alpha}, 正確率: {accuracy:.2%}")
+
+    # Track the best alpha and accuracy
+    if accuracy > best_accuracy:
+        best_alpha = alpha
+        best_accuracy = accuracy
+
+# Output the best alpha and accuracy
+print(f"最佳 Alpha 值: {best_alpha}, 準確率: {best_accuracy:.2%}")

src/flask_app.py

@@ -68,21 +68,24 @@ def post(self):
         source = request.json.get('source')
         question = request.json.get('query')
         category = request.json.get('category')
+        # alpha = request.json.get('alpha')
 
         # {
         # "qid": 1,
         # "source": [442, 115, 440, 196, 431, 392, 14, 51],
         # "query": "匯款銀行及中間行所收取之相關費用由誰負擔?",
         # "category": "insurance"
         # },
+
+        alpha = 0.5
 
         if not question:
             response = jsonify({'qid': '1', 'retrieve': '1'})
             response.status_code = 200
             return response
         else:
             try:
-                response = search_do(question, category, source)
+                response = search_do(question, category, source, alpha)
                 response = {
                     'qid': qid,
                     'retrieve': int(response)

src/utils/weaviate_op.py

@@ -4,6 +4,7 @@
 from langchain.embeddings import OpenAIEmbeddings
 
 import utils.config_log as config_log
+# import config_log as config_log
 
 config, logger, CONFIG_PATH = config_log.setup_config_and_logging()
 config.read(CONFIG_PATH)
@@ -72,16 +73,16 @@ def hybrid_search(self, query, source, num, alpha):
         return results
 
 
-def search_do(question, category, source):
+def search_do(question, category, source, alpha):
     if category == "finance":
-        vdb_named = "Finance"
+        vdb_named = "Financedev"
     elif category == "insurance":
-        vdb_named = "Insurance"
+        vdb_named = "Insurancedev"
     else:
-        vdb_named = "Faq"
+        vdb_named = "Faqdev"
 
     searcher = WeaviateSemanticSearch(vdb_named)
-    results = searcher.hybrid_search(question, source, 1, alpha=0.5)
+    results = searcher.hybrid_search(question, source, 1, alpha=alpha)
 
     result_li = []
     for _, result in enumerate(results, 1):
@@ -92,7 +93,7 @@ def search_do(question, category, source):
 
 
 if __name__ == '__main__':
-    vdb = "Faq"
+    vdb = "Insurancedev"
     client = WeaviateSemanticSearch(vdb)
 
     # 統計筆數