Update app.py

app.py

@@ -159,14 +159,22 @@ def phonetic_match(text, query, method='levenshtein_distance', apply_phonetic=Tr
         return jellyfish.levenshtein_distance(text_phonetic, query_phonetic)
     return 0
 
-def optimize_query(query, llm_model):
+#def optimize_query(query, llm_model):
+def optimize_query(query, llm_model, chunks, embedding_model, vector_store_type, search_type, top_k):
     llm = HuggingFacePipeline.from_model_id(
-        model_id="google/flan-t5-large",
+        model_id=llm_model,
         task="text2text-generation",
-        model_kwargs={"do_sample": True, "temperature": 0.7, "max_new_tokens": 512},
+        model_kwargs={"do_sample": True, "temperature": 0, "max_new_tokens": 64},
     )
+    
+    # Create a temporary vector store for query optimization
+    temp_vector_store = get_vector_store(vector_store_type, chunks, embedding_model)
+    
+    # Create a retriever with the temporary vector store
+    temp_retriever = get_retriever(temp_vector_store, search_type, {"k": top_k})
+    
     multi_query_retriever = MultiQueryRetriever.from_llm(
-        retriever=get_retriever(vector_store, search_type, search_kwargs),
+        retriever=temp_retriever,
         llm=llm
     )
     optimized_queries = multi_query_retriever.generate_queries(query)
@@ -453,6 +461,7 @@ def optimize_query(query, llm_model, chunks, embedding_model, vector_store_type,
     )
     optimized_queries = multi_query_retriever.generate_queries(query)
     return optimized_queries
+
     
 # New postprocessing function
 def rerank_results(results, query, reranker):
@@ -723,11 +732,21 @@ Text chunks:
 Provide your suggestions in a Python dictionary format."""
 
     # Use a HuggingFace model for text generation
-    llm = HuggingFacePipeline.from_model_id(
-        model_id="google/flan-t5-large",
-        task="text2text-generation",
-        model_kwargs={"do_sample": True, "temperature": 0.7, "max_new_tokens": 512},
+    model_id = "google/flan-t5-large"
+    tokenizer = AutoTokenizer.from_pretrained(model_id)
+    model = AutoModelForCausalLM.from_pretrained(model_id)
+    pipe = pipeline(
+        "text-generation", model=model, tokenizer=tokenizer, max_new_tokens=512
     )
+    llm = HuggingFacePipeline(pipeline=pipe)
+
+
+    
+    #llm = HuggingFacePipeline.from_model_id(
+    #    model_id="google/flan-t5-large",
+    #    task="text2text-generation",
+    #    model_kwargs={"do_sample": True, "temperature": 0.7, "max_new_tokens": 512},
+    #)
 
     # Generate suggestions 
     suggested_settings = llm.invoke(prompt)
@@ -1305,7 +1324,7 @@ if __name__ == "__main__":
     iface.launch(share=share)
 
 def run_automated_tests_and_analyze(*args):
-    file, query, model_types, model_names, split_strategies, chunk_sizes, overlap_sizes, \
+    file, query, auto_expected_result_input, model_types, model_names, split_strategies, chunk_sizes, overlap_sizes, \
     vector_store_types, search_types, top_k_values, optimize_vocab, use_query_optimization, use_reranking = args
 
     test_params = {
@@ -1327,7 +1346,7 @@ def run_automated_tests_and_analyze(*args):
         'use_reranking': [use_reranking]
     }
 
-    results_df, stats_df = automated_testing(file, query, test_params)
+    results_df, stats_df = automated_testing(file, query, test_params, auto_expected_result_input)
     recommendations = analyze_results(stats_df)
 
     return results_df, stats_df, recommendations