Update app.py

app.py

@@ -15,7 +15,7 @@ from sklearn.preprocessing import normalize
 from rank_bm25 import BM25Okapi
 from gensim.models import Word2Vec
 from typing import List, Optional, Tuple
-import tempfile
+import gradio as gr
 
 
 logger = logging.getLogger(__name__)
@@ -152,7 +152,7 @@ class MistralRAGChatbot:
             return np.array(response.data[0].embedding)
         except Exception as e:
             logging.error(f"Error fetching embedding: {e}")
-            return np.zeros((1024,))  # Assuming embedding size of 384
+            return np.zeros((1024,))  
 
     def advanced_fusion_retrieval(self, user_query: str, docs: List[dict]) -> List[dict]:
         query_embedding = self.create_embeddings([user_query])[0]
@@ -183,7 +183,7 @@ class MistralRAGChatbot:
         return [{'text': self.texts[i], 'method': 'advanced_fusion', 'score': combined_scores[i], 'index': i} for i in sorted_indices[:5]]
 
     def create_embeddings(self, text_list: List[str]) -> np.ndarray:
-        expected_dim = 1024  # The dimension expected by the Annoy index
+        expected_dim = 1024 
         embeddings = []
         for text in text_list:
             word_vectors = [self.word2vec_model.wv[token] for token in text.split() if token in self.word2vec_model.wv]
@@ -195,7 +195,6 @@ class MistralRAGChatbot:
             embeddings.append(avg_embedding)
         return np.array(embeddings, dtype=np.float32)
 
-
     async def generate_response_with_rag(
         self,
         user_query: str,
@@ -225,7 +224,6 @@ class MistralRAGChatbot:
             response = "An error occurred while generating the response."
         return response, [doc['text'] for doc in reranked_docs[:5]], reranked_docs[:5]
 
-
     def retrieve_documents(
         self,
         user_query: str,
@@ -248,7 +246,7 @@ class MistralRAGChatbot:
     def retrieve_with_annoy(self, user_query: str, query_embedding: np.ndarray, top_k: int) -> Tuple[List[int], List[float]]:
         n_results = min(top_k, len(self.texts))
         indices, distances = self.annoy_index.get_nns_by_vector(query_embedding, n_results, include_distances=True)
-        scores = [1.0 - (dist / max(distances)) for dist in distances]  # Normalize distances to a [0, 1] score
+        scores = [1.0 - (dist / max(distances)) for dist in distances] 
         logging.debug(f"Annoy retrieval returned {len(indices)} documents.")
         return indices, scores
 
@@ -315,9 +313,9 @@ class MistralRAGChatbot:
         return reranked_docs
 
     def reciprocal_rank_fusion(self, user_query: str, docs: List[dict]) -> List[dict]:
-        k = 60  # fusion parameter
+        k = 60  
         method_ranks = {}
-        fused_scores = {}  # Initialize fused_scores here
+        fused_scores = {} 
         for doc in docs:
             method = doc['method']
             if method not in method_ranks:
@@ -328,9 +326,9 @@ class MistralRAGChatbot:
             idx = doc['index']
             if idx not in fused_scores:
                 fused_scores[idx] = sum(1 / (k + rank) for method_rank in method_ranks.values() for i, rank in method_rank.items() if i == idx)
-        reranked_docs = sorted(docs, key=lambda x: fused_scores.get(x['index'], 0), reverse=True)  # Use get() to handle missing keys
+        reranked_docs = sorted(docs, key=lambda x: fused_scores.get(x['index'], 0), reverse=True)
         for doc in reranked_docs:
-            doc['rrf_score'] = fused_scores.get(doc['index'], 0)  # Use get() to handle missing keys
+            doc['rrf_score'] = fused_scores.get(doc['index'], 0) 
         return reranked_docs
 
     def weighted_score_fusion(self, user_query: str, docs: List[dict]) -> List[dict]:
@@ -405,17 +403,9 @@ def create_vector_db_and_annoy_index(pdf_path, vector_db_path, annoy_index_path)
     print("Vector database and Annoy index creation completed.")
 
 
-
-import gradio as gr
-
 def chatbot_interface(file, user_query, response_style, selected_retrieval_methods, selected_reranking_methods, chunk_size, overlap):
     vector_db_path = "vector_db.pkl"
     annoy_index_path = "vector_index.ann"
-
-    
-    
-    
-
     
     store_embeddings_in_vector_db(file.name, 'vector_db.pkl', 'vector_index.ann', chunk_size, overlap)
 
@@ -451,7 +441,7 @@ iface = gr.Interface(
         gr.File(label="Upload a PDF"), 
         gr.Textbox(lines=5, label="User Query"),
         gr.Dropdown(["Detailed", "Concise", "Creative", "Technical"], label="Response Style"),
-        gr.Dropdown(["annoy", "tfidf", "bm25", "euclidean", "jaccard"], label="Retrieval Methods", multiselect=True), # This line is changed
+        gr.Dropdown(["annoy", "tfidf", "bm25", "euclidean", "jaccard"], label="Retrieval Methods", multiselect=True),
         gr.Dropdown(["advanced_fusion", "reciprocal_rank_fusion", "weighted_score_fusion", "semantic_similarity"], label="Reranking Methods"),
         gr.Slider(minimum=1024, maximum=2048, step=128, value=2048, label="Chunk Size"),
         gr.Slider(minimum=100, maximum=300, step=100, value=200, label="Overlap")