Update app.py

app.py

@@ -1,33 +1,47 @@
-from transformers import RagTokenizer, RagRetriever, RagTokenForGeneration
-import openai
+from gpt_index import SimpleDirectoryReader, GPTListIndex, GPTSimpleVectorIndex, LLMPredictor, PromptHelper
 import os
+from langchain.chat_models import ChatOpenAI
+import gradio as gr
+from gpt_index import SimpleDirectoryReader, GPTSimpleVectorIndex, LLMPredictor, PromptHelper
+
+import sys
+
+llm = ChatOpenAI(temperature=0.7, model_name="gpt-3.5-turbo-instruct", engine="gpt-3.5-turbo-instruct", max_tokens=num_outputs)
+
+
 
-# Set your API key
 os.environ["OPENAI_API_KEY"] = os.getenv("OPENAI_API_KEY")
 
 
-# Initialize the RAG tokenizer and retriever
-tokenizer = RagTokenizer.from_pretrained("facebook/rag-token-nq")
-retriever = RagRetriever.from_pretrained("facebook/rag-token-nq", index_name="exact", use_dummy_dataset=True)
+def construct_index(directory_path):
+    max_input_size = 4096
+    num_outputs = 512
+    max_chunk_overlap = 20
+    chunk_size_limit = 600
+
+    prompt_helper = PromptHelper(max_input_size, num_outputs, max_chunk_overlap, chunk_size_limit=chunk_size_limit)
+
+    llm_predictor = LLMPredictor(llm=llm)
 
-# Initialize the RAG token for generation
-model = RagTokenForGeneration.from_pretrained("facebook/rag-token-nq", retriever=retriever)
+    documents = SimpleDirectoryReader(directory_path).load_data()
 
-# Define the context for the question
-context = "Python is a popular programming language used for various applications."
+    index = GPTSimpleVectorIndex(documents, llm_predictor=llm_predictor, prompt_helper=prompt_helper)
 
-# Ask a question
-question = "What is Python used for?"
+    index.save_to_disk('index.json')
 
-# Prepare input for the model
-input_dict = tokenizer(question, context, return_tensors="pt")
+    return index
+   
+def chatbot(input_text):
+    predetermined_text = "- please only answer using information found in the directory 'docs' and nothing else"  
+    input_text = input_text + predetermined_text
+    index = GPTSimpleVectorIndex.load_from_disk('index.json')
+    response = index.query(input_text, response_mode="compact")
+    return response.response
 
-# Generate an answer using OpenAI's GPT-3
-generated = openai.Completion.create(
-  model="text-davinci-003",
-  prompt=f"Question: {question} Context: {context}",
-  max_tokens=150
-)
+iface = gr.Interface(fn=chatbot,
+                     inputs=gr.components.Textbox(lines=7, label="What do you want to know about the project?"),
+                     outputs="text",
+                     title="Project Knowledge AI")
 
-# Print the generated answer
-print(generated.choices[0].text.strip())
+index = construct_index("docs")
+iface.launch()