Update app.py

app.py

@@ -62,16 +62,18 @@ def add_text_to_db(text):
     
     return f"Text added successfully with ID: {doc_id}"
 
-def search_similar_texts(query, k):
+def search_similar_texts(query, k, threshold):
     """
     Search for the top k similar texts in the vector database and rerank them.
+    Only return results with similarity scores above the threshold.
     
     Args:
         query (str): The search query.
         k (int): Number of results to return.
+        threshold (float): Minimum similarity score (0 to 1).
         
     Returns:
-        str: Formatted search results without similarity scores.
+        str: Formatted search results with similarity scores or "No such record".
     """
     if not query or not query.strip():
         return "Error: Query cannot be empty."
@@ -79,21 +81,29 @@ def search_similar_texts(query, k):
     if not isinstance(k, int) or k < 1:
         return "Error: k must be a positive integer."
     
+    if not isinstance(threshold, (int, float)) or threshold < 0 or threshold > 1:
+        return "Error: Threshold must be a number between 0 and 1."
+    
     # Retrieve and rerank
     retriever.search_kwargs["k"] = max(k * 2, 10)  # Retrieve 2k or at least 10
     compressor.top_n = k  # Rerank to top k
     docs = compression_retriever.get_relevant_documents(query)
     
     if not docs:
-        return "No results found."
+        return "No such record."
     
-    # Format results without similarity scores
+    # Filter results by threshold
     results = []
-    for i, doc in enumerate(docs[:k]):  # Ensure we return at most k
+    for i, doc in enumerate(docs[:k]):  # Ensure at most k results
         text = doc.metadata.get("text", "No text available")
-        results.append(f"Result {i+1}:\nText: {text}\n")
+        score = doc.metadata.get("score", 0.0)  # Reranker score
+        if score >= threshold:
+            results.append(f"Result {i+1}:\nText: {text}\nScore: {score:.4f}\n")
+    
+    if not results:
+        return "No such record."
     
-    return "\n".join(results) or "No results found."
+    return "\n".join(results)
 
 # Gradio interface
 with gr.Blocks() as demo:
@@ -110,6 +120,7 @@ with gr.Blocks() as demo:
             gr.Markdown("## Search Similar Texts")
             query_input = gr.Textbox(label="Enter search query")
             k_input = gr.Number(label="Number of results (k)", value=5, precision=0)
+            threshold_input = gr.Number(label="Similarity threshold (0 to 1)", value=0.5, minimum=0, maximum=1)
             search_button = gr.Button("Search")
             search_output = gr.Textbox(label="Search Results")
     
@@ -121,7 +132,7 @@ with gr.Blocks() as demo:
     )
     search_button.click(
         fn=search_similar_texts,
-        inputs=[query_input, k_input],
+        inputs=[query_input, k_input, threshold_input],
         outputs=search_output
     )