added context retrieval to UI

app.py

@@ -111,14 +111,16 @@ def predict(
             (history[i].strip(), history[i + 1].strip())
             for i in range(0, len(history) - 1, 2)
         ]
-        yield chat, history
+        yield chat, history, None, None, None
     
     # add context retrieval part here
     ta = retrieval.Retrieval()
     ta._load_pinecone_vectorstore()
     question = inputs
     top_context_list = ta.retrieve_contexts_from_pinecone(user_question=question, topk=NUM_ANSWERS_GENERATED)
-    print(top_context_list)
+    print(len(top_context_list))
+
+    yield chat, history, top_context_list[0], top_context_list[1], top_context_list[2]
 
 def reset_textbox():
     return gr.update(value="")
@@ -333,7 +335,7 @@ with gr.Blocks(
             chatbot,
             state,
         ],
-        [chatbot, state],
+        [chatbot, state, context1, context2, context3],
     )
     run_btn.click(
         predict,

requirements.txt

@@ -3,3 +3,5 @@ gradio==3.20.1
 pinecone-client
 sentence-transformers
 pandas
+langchain
+python-dotenv

retrieval.py

@@ -15,7 +15,6 @@ from dotenv import load_dotenv
 from PIL import Image
 from transformers import (AutoModelForSequenceClassification, AutoTokenizer, GPT2Tokenizer, OPTForCausalLM, T5ForConditionalGeneration)
 
-PINECONE_API_KEY="insert your pinecone api key here"
 
 class Retrieval:
     def __init__(self,
@@ -35,11 +34,11 @@ class Retrieval:
     
     def _load_pinecone_vectorstore(self,):
         model_name = "intfloat/e5-large"  # best text embedding model. 1024 dims.
-        pincecone_index = pinecone.Index("uiuc-chatbot")
+        
         embeddings = HuggingFaceEmbeddings(model_name=model_name)
         #pinecone.init(api_key=os.environ['PINECONE_API_KEY'], environment="us-west1-gcp")
         pinecone.init(api_key=PINECONE_API_KEY, environment="us-west1-gcp")
-
+        pincecone_index = pinecone.Index("uiuc-chatbot")
         print(pinecone.list_indexes())
 
         self.vectorstore = Pinecone(index=pincecone_index, embedding_function=embeddings.embed_query, text_key="text")