Upload app.py

app.py

@@ -17,12 +17,16 @@ from numpy.linalg import norm
 
 import json
 import ast
+import requests
+
+
 
 import gradio as gr
 from datetime import datetime
 import time 
 import dataframe_image as dfi
 
+
 print("Packages loaded!")
 
 # write out functions 
@@ -55,129 +59,55 @@ def cosine_similarity_generator(master_exploded, embeddings, query, filename = t
   master_exploded_top = master_exploded[master_exploded['cos_sim']> 0.6]
   print("The number of results with cosine similarity > 0.6 are: ", len(master_exploded[master_exploded['cos_sim']> 0.6]))
   
-  master_exploded_top_k = master_exploded.sort_values(by=['cos_sim'], ascending=False).head(10)
+  top_k = master_exploded.sort_values(by=['cos_sim'], ascending=False).head(10)
   print(" The top k=10 results have a min cosine similarity of: ", master_exploded.sort_values(by=['cos_sim'], ascending=False).head(10)['cos_sim'].min())
   # print(master_exploded_top_k)
-  cosine_sum_by_name = master_exploded_top.groupby(["name", "query", "combined", "tokenized_sentences",]).agg({"cos_sim": ["sum"]}).reset_index()
+  cosine_sum_by_name = master_exploded_top.groupby(["id", "name", "query", "gitHubUrl", "combined" ]).agg({"cos_sim": ["sum"]}).reset_index()
   print("Taking sum of cosine similarities above 0.6 threshold...")
   cosine_sum_by_name.columns = cosine_sum_by_name.columns.map('_'.join)
   
   ranked_mentors = cosine_sum_by_name.reset_index().sort_values(by ="cos_sim_sum", ascending =False) 
-  path = "./Ranked_Results_Gradio/"
-  ranked_mentors_filename = path+'ranked_mentors_'+str(filename)+'.csv'
-  cos_sum_filename = path+'cos_sum_'+str(filename)+'.csv'
-  top_10_filename = path+'top_10_'+str(filename)+'.csv'
-  above_threshold_filename = path+"above_0.6_threshold_"+str(filename)+".csv"
+  # path = "./Ranked_Results_Gradio/"
+  # ranked_mentors_filename = path+'ranked_mentors_'+str(filename)+'.csv'
+  # cos_sum_filename = path+'cos_sum_'+str(filename)+'.csv'
+  # top_10_filename = path+'top_10_'+str(filename)+'.csv'
+  # above_threshold_filename = path+"above_0.6_threshold_"+str(filename)+".csv"
 
   # save 3 files: Ranked mentors, top 10 matches baed on cosine similarity sum, and then all of the results per run. 
   # ranked_mentors.head(10).to_csv(ranked_mentors_filename)
   # cosine_sum_by_name.sort_values(by ="cos_sim_sum", ascending =False).head(10).to_csv(top_10_filename)
   # cosine_sum_by_name.to_csv(cos_sum_filename)
   # master_exploded_top.sort_values(by ="cos_sim", ascending =False).to_csv(above_threshold_filename)
-  return master_exploded_top, master_exploded_top_k, cosine_sum_by_name
-
-
-
-# def recosys(query):
-#     # main function 
-#     master_exploded = load_pickle()
-#     print("Pickle loaded!")
-#     embeddings, query = sentence_embedding_generator(query)
-#     print("Query embeddings generated...")
-#     # change query to date time 
-#     ranked_mentors, top_10, cosine_sum_by_name = cosine_similarity_generator(master_exploded,  embeddings, query, time.strftime("%Y%m%d-%H%M%S"))
-#     print("Ranked mentors printed! Check your directory.")
-#     # to_return_df = gr.Dataframe(
-#     #    headers = ["name_", "combined_", "cos_sim_sum"], 
-#     #     datatype=["str", "str", "number"],
-#     #     row_count=10,
-#     #     col_count=(3, "fixed"),
-#     #     )
-#     return top_10
-
-# def recosys_embedding(query): 
-#    master_exploded = load_pickle()
-#    print("Pickle loading....")
-#    embeddings, query = sentence_embedding_generator(query)
-#    return embeddings, master_exploded
-
-# def embeddings_to_ranking(embeddings, master_exploded, query): 
-#    ranked_mentors, top_10, cosine_sum_by_name = cosine_similarity_generator(master_exploded,  embeddings, query, time.strftime("%Y%m%d-%H%M%S"))
-#    return top_10
-
-# def flip_text(x):
-#     return x[::-1]
-
-master_exploded = load_pickle()
-
-demo = gr.Blocks() 
-
-with demo: 
-   gr.Markdown(
-      """
-      # Mentor Semantic Search Recommender System 
-      What kind of mentor are you looking for? 
-      """
-      )
-   input = gr.Textbox(placeholder="someone who has experience working in big tech")
-
-   def generate_results(input):
-     embeddings, query = sentence_embedding_generator(str(input))
-     ranked_mentors, top_10, cosine_sum_by_name = cosine_similarity_generator(master_exploded,  
-                                                                              embeddings, 
-                                                                              query, 
-                                                                              time.strftime("%Y%m%d-%H%M%S"))
-     print(top_10.columns)
-     return(top_10['name'].to_list())
-   
-demo = gr.Interface(fn=generate_results, inputs="text", outputs="text")
-demo.launch()   
-# demo.launch(share=True)   
-#    embeddings_btn = gr.Button("Generate query embeddings")
-   
-#    def embedding_gen(input): 
-#      embeddings, query = sentence_embedding_generator(str(input))
-#      return embeddings
-     
-   
-#    embeddings_btn.click(fn=embedding_gen, inputs=input, outputs="number")
-#    ranked_mentors, top_10, cosine_sum_by_name = cosine_similarity_generator(master_exploded,  
-#                                                                               embeddings, 
-#                                                                               query, 
-#                                                                               time.strftime("%Y%m%d-%H%M%S"))
-#    def top_5_mentors(top_10):
-#      return top_10.head(5)
-
-# #    def scorer(embeddings, query):
-# #      ranked_mentors, top_10, cosine_sum_by_name = cosine_similarity_generator(master_exploded,  
-# #                                                                               embeddings, 
-# #                                                                               query, 
-# #                                                                               time.strftime("%Y%m%d-%H%M%S"))
-# #      return top_10
-   
-#    submit_btn = gr.Button("Search for mentors")
-#    submit_btn.click(fn=top_5_mentors, inputs=top_10, outputs=gr.Dataframe())
-# #    dfi.export(df_styled,"mytable.png")
-# #    submit_btn.click(fn=recosys, inputs=input, outputs=gr.Dataframe(), api_name="RecoSys")
-   
-#    input.change(fn=recosys, inputs = input, outputs = output)
-
-
-# demo = gr.Interface(
-#     recosys,
-#     inputs="text",
-#     "dataframe")
-#     outputs=gr.Dataframe(
-#        headers = ["name_", "combined_", "cos_sim_sum"], 
-#         datatype=["str", "str", "number"],
-#         row_count=10,
-#         col_count=(3, "fixed"),
-#         )
-    
-#     # "dataframe",
-#     # description="What kind fo mentoir are you looking for?",
-# )
-# demo = gr.Interface(fn=recosys, inputs="text", outputs="dataframe")
-
-
-
+  return master_exploded_top, top_k, cosine_sum_by_name
+
+def dataframe_output(cosine_sum_by_name): 
+  # return master_exploded_top_k
+  json_df = cosine_sum_by_name.to_json(orient="columns")
+  return json_df
+
+
+def generate_results(input): 
+  master_exploded = load_pickle()
+  embeddings, query = sentence_embedding_generator(str(input))
+  ranked_mentors, top_10, cosine_sum_by_name = cosine_similarity_generator(master_exploded,  
+                                                                            embeddings, 
+                                                                            query, 
+                                                                            time.strftime("%Y%m%d-%H%M%S"))
+  print(cosine_sum_by_name.columns)
+  df_output = pd.read_json(dataframe_output(cosine_sum_by_name))
+  print(df_output)
+  # df_output = dataframe_output(cosine_sum_by_name)
+  print("JSON created...")
+  subset = df_output.head(10)  # Select the first 10 rows
+
+  return subset
+
+iface = gr.Interface(
+    fn=generate_results,
+    inputs=gr.inputs.Textbox(label="What kind of mentor are you looking for?"),
+    outputs=gr.outputs.Dataframe(type="pandas"),
+    title="SharpestMinds Mentor Recommender Semantic Search App",
+    description="Converts a string query into an embedding, and then compares the aggregate cosine similarity by mentor.",
+)
+
+iface.launch()