Upload 2 files

app.py

@@ -0,0 +1,189 @@
+
+
+## Setup
+# Import the necessary Libraries
+import gradio as gr
+import pandas as pd
+import os
+import json
+import uuid
+import tiktoken
+import openai
+from dotenv import load_dotenv
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+from langchain_core.documents import Document
+from langchain_community.document_loaders import PyPDFDirectoryLoader
+from langchain_community.embeddings.sentence_transformer import (
+    SentenceTransformerEmbeddings
+)
+from langchain_community.vectorstores import Chroma
+
+from langchain_community.chat_models import ChatOpenAI
+
+from huggingface_hub import CommitScheduler
+from pathlib import Path
+
+
+
+# Create Client
+anyscale_api_key = userdata.get('anyscale_apiKey')
+
+client = OpenAI(base_url="https://api.endpoints.anyscale.com/v1",
+                api_key=anyscale_api_key
+)
+
+# Define the embedding model and the vectorstore
+model_name = "thenlper/gte-large"
+embedding_model = "thenlper/gte-large"
+embedding_model = SentenceTransformerEmbeddings(model_name='thenlper/gte-large')
+
+persisted_vectordb_location = '/content/drive/MyDrive/finsightsdb'
+collection_name = 'finsights_grey-10k'
+
+
+# Load the persisted vectorDB
+vectorstore_persisted = Chroma(
+    collection_name=collection_name,
+    persist_directory=persisted_vectordb_location,
+    embedding_function=embedding_model
+)
+
+# Prepare the logging functionality
+
+log_file = Path("logs/") / f"data_{uuid.uuid4()}.json"
+log_folder = log_file.parent
+
+scheduler = CommitScheduler(
+    repo_id="project3-logs",
+    repo_type="dataset",
+    folder_path=log_folder,
+    path_in_repo="data",
+    every=2
+)
+
+# Define the Q&A system message
+qna_system_message = """
+You are an assistant to a financial services firm. Your task is to determine the most effective platform to support the generation by the firm of advanced analytics and insights for investment management and financial planning.
+
+User input will include the necessary context for you to answer their questions. This context will begin with the token: ###Context.
+The context contains references to specific portions of documents relevant to the user's query, along with source links.
+The source for a context will begin with the token ###Source
+
+When crafting your response:
+1. Select only context relevant to answer the question.
+2. Include the source links in your response.
+3. User questions will begin with the token: ###Question.
+4. If the question is irrelevant to the firm's business respond with - "I am an AI assistant for Finsights Grey Inc. I can only help you with questions related to financial analytics."
+
+Please adhere to the following guidelines:
+- Your response should only be about the question asked and nothing else.
+- Answer only using the context provided.
+- Do not mention anything about the context in your final answer.
+- If the answer is not found in the context, it is very very important for you to respond with "I don't know. Please check the docs @ 'https://docs.finsights.io/'"
+- Always quote the source when you use the context. Cite the relevant source at the end of your response under the section - Source:
+- Do not make up sources. Use the links provided in the sources section of the context and nothing else. You are prohibited from providing other links/sources.
+
+Here is an example of how to structure your response:
+
+Answer:
+[Answer]
+
+Source:
+[Source]
+"""
+
+
+# Define the user message template
+qna_user_message_template = """
+###Context
+Here are some documents and their source links that are relevant to the question mentioned below.
+{context}
+
+###Question
+{question}
+"""
+
+# Define the predict function that runs when 'Submit' is clicked or when a API request is made
+def predict(user_input,company):
+
+    filter = "/dataset/"+company+"-10-k-2023.pdf"
+    relevant_document_chunks = vectorstore_persisted.similarity_search(user_input, k=5, filter={"source":filter})
+
+    # Create context_for_query
+    context_list = [d.page_content + "\n Page number: " + str(d.metadata['page']) + "\n ###Source: " + d.metadata['source'] + "\n\n " for d in relevant_document_chunks]
+    context_for_query = ". ".join(context_list)
+
+
+    # Create messages
+    prompt = [
+    {'role':'system', 'content': qna_system_message},
+    {'role': 'user', 'content': qna_user_message_template.format(
+         context=context_for_query,
+         question=user_input
+        )
+    }
+]
+
+    # Get response from the LLM
+    # Handle errors using try-except
+    # print the content of the response
+    try:
+      response = client.chat.completions.create(
+        model=model_name,
+        messages=prompt,
+        temperature=0
+     )
+
+      prediction = response.choices[0].message.content.strip()
+    except Exception as e:
+      prediction = f'Sorry, I encountered the following error: \n {e}'
+
+    print(prediction)
+
+  
+    # While the prediction is made, log both the inputs and outputs to a local log file
+    # While writing to the log file, ensure that the commit scheduler is locked to avoid parallel
+    # access
+
+    with scheduler.lock:
+        with log_file.open("a") as f:
+            f.write(json.dumps(
+                {
+                    'user_input': user_input,
+                    'retrieved_context': context_for_query,
+                    'model_response': prediction
+                }
+            ))
+            f.write("\n")
+
+    return prediction
+
+# Set-up the Gradio UI
+# Add text box and radio button to the interface
+# The radio button is used to select the company 10k report in which the context needs to be retrieved.
+# The text box is used to enter the question.
+# The submit button is used to run the predict function
+
+
+textbox = gr.Textbox()
+company = gr.Radio(choices=['aws', 'google', 'meta', 'msft', 'IBM'], label="Select a company:")
+   
+#predict = gr.Button("Submit")
+predict.click(predict, inputs=[textbox,company], outputs=[predict])
+
+
+
+
+# Create the interface
+# For the inputs parameter of Interface provide [textbox,company]
+# For the outputs parameter of Interface provide [predict]
+demo = gr.Interface(
+    fn=predict,
+    inputs=[textbox,company],
+    outputs=[predict],
+    title="AI-Powered Question Answering")
+
+# Run the interface
+
+demo.queue()
+demo.launch()

requirements.txt

@@ -0,0 +1,8 @@
+openai==1.23.2 
+tiktoken==0.6.0 
+langchain==0.1.1 
+langchain-community==0.0.13 
+chromadb==0.4.22 
+sentence-transformers==2.3.1 
+datasets
+pypdf