multilingual

pages/Semantic_Search.py

@@ -285,12 +285,7 @@ if(search_all_type==True):
     'NeuralSparse Search',
     ]
 
-
-
-
-
-    
-    
+   
 def generate_images(tab,inp_):
         #write_top_bar()
         seed = random.randint(1, 10)

semantic_search/all_search_execute.py

@@ -68,7 +68,7 @@ def handler(input_,session_id):
     print("*********")
     print(input_)
     search_types = input_["searchType"]
-    
+    print(type(search_types))
     if("NormType" not in input_.keys()):
         norm_type = "min_max"
     else:
@@ -111,37 +111,70 @@ def handler(input_,session_id):
             weights.append(weight)
             
       
-        
-    ######## Updating hybrid Search pipeline #######   
-    print("Updating Search pipeline with new weights")        
-    s_pipeline_payload = {"version": 1234}
-    s_pipeline_payload["phase_results_processors"] = [
-                {
-                    "normalization-processor": {
-                    "normalization": {
-                        "technique": norm_type
-                    },
-                    "combination": {
-                        "technique": combine_type,
-                        "parameters": {
-                        "weights": weights
+    if(num_queries>1):
+        ######## Updating hybrid Search pipeline #######   
+        print("Updating Search pipeline with new weights")        
+        s_pipeline_payload = {"version": 1234}
+        s_pipeline_payload["phase_results_processors"] = [
+                    {
+                        "normalization-processor": {
+                        "normalization": {
+                            "technique": norm_type
+                        },
+                        "combination": {
+                            "technique": combine_type,
+                            "parameters": {
+                            "weights": weights
+                            }
+                        }
                         }
                     }
-                    }
-                }
-                ]
+                    ]
             
-
+        hybrid_search_processor = s_pipeline_payload["phase_results_processors"]
         
-    opensearch_search_pipeline = (requests.get(host+'_search/pipeline/hybrid_search_pipeline', auth=awsauth,headers=headers)).text
-    if(opensearch_search_pipeline!='{}'):
-        path = "_search/pipeline/hybrid_search_pipeline" 
-        url = host + path
-        r = requests.put(url, auth=awsauth, json=s_pipeline_payload, headers=headers)
-        print("Hybrid Search Pipeline updated: "+str(r.status_code))
-        ######## Combining hybrid+rerank pipeline ####### 
-        opensearch_rerank_pipeline = (requests.get(host+'_search/pipeline/rerank_pipeline', auth=awsauth,headers=headers)).text
-    ######## start of Applying LLM filters ####### 
+        opensearch_search_pipeline = (requests.get(host+'_search/pipeline/hybrid_search_pipeline', auth=awsauth,headers=headers)).text
+        if(opensearch_search_pipeline!='{}'):
+            path = "_search/pipeline/hybrid_search_pipeline" 
+            url = host + path
+            r = requests.put(url, auth=awsauth, json=s_pipeline_payload, headers=headers)
+            print("Hybrid Search Pipeline updated: "+str(r.status_code))
+            ######## Combining hybrid+rerank pipeline ####### 
+            path = "_search/pipeline/hybrid_rerank_pipeline" 
+            url = host + path
+            s_pipeline_payload['response_processors'] = [
+                                                                    {
+                                                                        "rerank": {
+                                                                        "ml_opensearch": {
+                                                                            "model_id": "deBS3pYB5VHEj-qVuPHT"
+                                                                        },
+                                                                        "context": {
+                                                                            "document_fields": [
+                                                                            "product_description"
+                                                                            ]
+                                                                        }
+                                                                        }
+                                                                    }
+                                                                    ]
+            r = requests.put(url, auth=awsauth, json=s_pipeline_payload, headers=headers)
+            print("Hybrid Rerank Search Pipeline updated: "+str(r.status_code))
+        
+            ######## Updating opensearch_translation_pipeline Search pipeline #######   
+            opensearch_translation_pipeline = (requests.get(host+'_search/pipeline/ml_inference_for_vector_search_and_language_translation', auth=awsauth,headers=headers)).text
+            path = "_search/pipeline/ml_inference_for_vector_search_and_language_translation" 
+            url = host + path
+            opensearch_translation_pipeline["phase_results_processors"] = hybrid_search_processor
+            r = requests.put(url, auth=awsauth, json=opensearch_translation_pipeline, headers=headers)
+            print("translation hybrid Search Pipeline updated: "+str(r.status_code))
+
+            ######## Updating opensearch_translation_pipeline_with_rerank Search pipeline #######
+            opensearch_translation_pipeline_with_rerank = (requests.get(host+'_search/pipeline/ml_inference_for_vector_search_and_language_translation_with_rerank', auth=awsauth,headers=headers)).text
+            path = "_search/pipeline/ml_inference_for_vector_search_and_language_translation_with_rerank" 
+            url = host + path
+            opensearch_translation_pipeline_with_rerank["phase_results_processors"] = hybrid_search_processor
+            r = requests.put(url, auth=awsauth, json=opensearch_translation_pipeline_with_rerank, headers=headers)
+            print("translation hybrid rerank Search Pipeline updated: "+str(r.status_code))
+        ######## start of Applying LLM filters ####### 
     if(st.session_state.input_rewritten_query!=""):
             filter_ = {"filter": {
                  "bool": {
@@ -456,17 +489,30 @@ def handler(input_,session_id):
     
     else:
         if( st.session_state.input_hybridType == "OpenSearch Hybrid Query"):
-            url_ = url + "?search_pipeline=hybrid_search_pipeline" 
-            
-            if(st.session_state.re_ranker == 'true' and st.session_state.input_reranker == 'Cohere Rerank'):
+            if(st.session_state.input_multilingual):
+                if(st.session_state.re_ranker == 'true' and st.session_state.input_reranker == 'Cohere Rerank'):
+                    path = "demostore-search-index-reindex-new/_search?search_pipeline=ml_inference_for_vector_search_and_language_translation_with_rerank" 
+                    url = host + path
+                    hybrid_payload["ext"] = {"rerank": {
+                                                "query_context": {
+                                                    "query_text": query
+                                                }
+                                                }}
+                else:
+                    path = "demostore-search-index-reindex-new/_search?search_pipeline=ml_inference_for_vector_search_and_language_translation" 
+                    url = host + path
+            else:
+                url_ = url + "?search_pipeline=hybrid_search_pipeline" 
                 
-                url_ = url + "?search_pipeline=hybrid_rerank_pipeline" 
-            
-                hybrid_payload["ext"] = {"rerank": {
-                                          "query_context": {
-                                             "query_text": query
-                                          }
-                                        }}
+                if(st.session_state.re_ranker == 'true' and st.session_state.input_reranker == 'Cohere Rerank'):
+                    
+                    url_ = url + "?search_pipeline=hybrid_rerank_pipeline" 
+                
+                    hybrid_payload["ext"] = {"rerank": {
+                                            "query_context": {
+                                                "query_text": query
+                                            }
+                                            }}
             r = requests.get(url_, auth=awsauth, json=hybrid_payload, headers=headers)
             response_ = json.loads(r.text)
             docs = response_['hits']['hits']