Create app.py

app.py

@@ -0,0 +1,87 @@
+
+import gradio as gr
+import pandas as pd
+
+from langchain.embeddings.openai import OpenAIEmbeddings
+from langchain.text_splitter import CharacterTextSplitter
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+from langchain.vectorstores import Chroma
+from langchain.document_loaders import TextLoader
+from langchain.chat_models import ChatOpenAI
+from langchain import PromptTemplate
+from langchain.chains import LLMChain
+from langchain.chains.qa_with_sources import load_qa_with_sources_chain
+from langchain.llms import OpenAI
+from langchain.vectorstores import FAISS
+
+
+
+
+db = pd.read_pickle('index.pkl')
+
+#-----------------------------------------------------------------------------
+def get_response_from_query(db, query, k=3):
+
+    docs = db.similarity_search(query, k=k)
+
+    docs_page_content = " ".join([d.page_content for d in docs])
+
+    # llm = BardLLM()
+    llm = ChatOpenAI(model_name="gpt-3.5-turbo-16k",temperature=0)
+
+    prompt = PromptTemplate(
+        input_variables=["question", "docs"],
+        template="""
+        A bot that is open to discussions about different cultural, philosophical and political exchanges. I will use do different analysis to the articles provided to me. Stay truthful and if you weren't provided any resources give your oppinion only.
+        Answer the following question: {question}
+        By searching the following articles: {docs}
+
+        Only use the factual information from the documents. Make sure to mention key phrases from the articles.
+
+        If you feel like you don't have enough information to answer the question, say "I don't know".
+
+        """,
+    )
+
+    chain = LLMChain(llm=llm, prompt=prompt)
+    # chain = RetrievalQAWithSourcesChain.from_chain_type(llm=llm, prompt=prompt,
+    #                                                     chain_type="stuff", retriever=db.as_retriever(), return_source_documents=True)
+
+    response = chain.run(question=query, docs=docs_page_content,return_source_documents=True)
+    r_text = str(response)
+
+    ##evaluation part
+
+    prompt_eval = PromptTemplate(
+        input_variables=["answer", "docs"],
+        template="""
+       You job is to evaluate if the response to a given context is faithful.
+
+        for the following: {answer}
+        By searching the following article: {docs}
+
+       Give a reason why they are similar or not, start with a Yes or a No.
+        """,
+    )
+
+    chain_part_2 = LLMChain(llm=llm, prompt=prompt_eval)
+
+
+    evals = chain_part_2.run(answer=r_text, docs=docs_page_content)
+
+    return response,docs,evals
+
+def greet(query):
+
+    answer,sources,evals = get_response_from_query(db,query,2)
+    return answer,sources,evals
+
+demo = gr.Interface(fn=greet, title="cicero-semantic-search", inputs="text",
+                    outputs=[gr.components.Textbox(lines=3, label="Response"),
+                             gr.components.Textbox(lines=3, label="Source"),
+                             gr.components.Textbox(lines=3, label="Evaluation")])
+
+demo.launch(share=True)
+
+
+