Update app.py

app.py

@@ -1,73 +1,39 @@
-import langchain as lc
-import langchain.llms as lc_llms
-import langchain.embeddings as lc_embed
-import langchain.document_loaders as lcdl
-import langchain.text_splitter as lcts
-import langchain.vectorstores as lcvs
-import langchain.chains as lcchains
-import langchain.hub as lchub
-
+import huggingface_hub as hf_hub 
+import gradio as gr 
 import os
-import gradio as gr
-
-hf_google_flan_t5_xxl = lc.HuggingFaceHub(
-    repo_id = 'google/flan-t5-xxl'
-)
-
-hf_gpt2_xl = lc.HuggingFaceHub(repo_id = 'gpt2-xl')
-
-sentence_embeddings = lc_embed.HuggingFaceInferenceAPIEmbeddings(
-    api_key = os.environ['HUGGINGFACEHUB_API_TOKEN'],
-    model_name = 'sentence-transformers/all-MiniLM-L6-v2'
-)
 
-text_loader = lcdl.PyPDFLoader('Story.pdf')
-
-documents = text_loader.load()
-
-text_splitter = lcts.RecursiveCharacterTextSplitter(
-    chunk_size = 1000,
-    chunk_overlap  = 200,
+client = hf_hub.InferenceClient(
+    model = 'google/flan-t5-xxl',
+    token = os.environ['HUGGINGFACEHUB_API_TOKEN']
 )
 
-splits = text_splitter.split_documents(documents)
-
-db = lcvs.Chroma.from_documents(splits, sentence_embeddings)
+def create_prompt(question, history):
+    template = '''You are a good chatbot that talks with human. 
 
-retriever = db.as_retriever(search_type="mmr")
+    {history}
 
-template = """
-You are an AI chatbot. Complete the conversation based on following context:
+    Human : {question}
+    Chatbot :'''
 
------------------------------
-Context : {context}
------------------------------
-Conversation History:
+    conv_template = '''
+    Human : {query}
+    Chatbot : {response}
+    '''
 
-{question}
-"""
-
-qa_prompt = lc.PromptTemplate.from_template(template) 
-
-qa_chain = lcchains.RetrievalQA.from_chain_type(
-    llm = hf_google_flan_t5_xxl,
-    retriever = retriever,
-    return_source_documents = True,
-    chain_type_kwargs={"prompt": qa_prompt}
-)
+    prev_chat = ''
 
-def rag_interface(ques, history):
-    prev_chat = " ".join([f"Human: {q} \nChatbot: {a} \n" for q, a in history])
-    current_ques = f'\nHuman: {ques} \nChatbot:'
+    for ques, ans in history:
+        prev_chat += conv_template.format(query = ques, response = ans)
 
-    query = prev_chat + current_ques
+    prompt = template.format(history = prev_chat, question = question)
 
-    result = qa_chain({
-        'query' : query
-    })
+    return prompt
 
-    return result['result']
+def chat_interface(question, history):
+    prompt = create_prompt(question, history)
+    response = client.text_generation(prompt)
+    return response
 
-demo = gr.ChatInterface(fn=rag_interface)
+chat_app = gr.ChatInterface(fn = chat_interface)
 
-demo.launch()
+chat_app.launch()