Update app.py

app.py

@@ -1,4 +1,6 @@
 import os
+import uuid
+import json
 
 import gradio as gr
 
@@ -7,7 +9,14 @@ from openai import OpenAI
 from langchain_community.embeddings.sentence_transformer import SentenceTransformerEmbeddings
 from langchain_community.vectorstores import Chroma
 
-client = OpenAI(api_key=os.environ['OPENAI_API_KEY'])
+from huggingface_hub import CommitScheduler
+from pathlib import Path
+
+
+client = OpenAI(
+    base_url="https://api.endpoints.anyscale.com/v1",
+    api_key=os.environ['ANYSCALE_API_KEY']
+)
 
 embedding_model = SentenceTransformerEmbeddings(model_name='thenlper/gte-small')
 
@@ -24,6 +33,19 @@ retriever = vectorstore_persisted.as_retriever(
     search_kwargs={'k': 5}
 )
 
+# Prepare the logging functionality
+
+log_file = Path("logs/") / f"data_{uuid.uuid4()}.json"
+log_folder = log_file.parent
+
+scheduler = CommitScheduler(
+    repo_id="document-qna-chroma-anyscale-logs",
+    repo_type="dataset",
+    folder_path=log_folder,
+    path_in_repo="data",
+    every=2
+)
+
 qna_system_message = """
 You are an assistant to a financial services firm who answers user queries on annual reports.
 Users will ask questions delimited by triple backticks, that is, ```.
@@ -43,9 +65,10 @@ Here are some documents that are relevant to the question.
 ```
 """
 
+# Define the predict function that runs when 'Submit' is clicked or when a API request is made
 def predict(user_input):
 
-    relevant_document_chunks = retriever.get_relevant_documents(user_input)
+    relevant_document_chunks = retriever.invoke(user_input)
     context_list = [d.page_content for d in relevant_document_chunks]
     context_for_query = ".".join(context_list)
     
@@ -60,7 +83,7 @@ def predict(user_input):
 
     try:
         response = client.chat.completions.create(
-            model="gpt-3.5-turbo",
+            model='mlabonne/NeuralHermes-2.5-Mistral-7B',
             messages=prompt,
             temperature=0
         )
@@ -69,12 +92,28 @@ def predict(user_input):
 
     except Exception as e:
         prediction = e
-        
+
+    # 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
 
 
 textbox = gr.Textbox(placeholder="Enter your query here", lines=6)
 
+# Create the interface
 demo = gr.Interface(
     inputs=textbox, fn=predict, outputs="text",
     title="AMA on Tesla 10-K statements",
@@ -83,11 +122,11 @@ demo = gr.Interface(
     examples=[["What was the total revenue of the company in 2022?", "$ 81.46 Billion"],
               ["Summarize the Management Discussion and Analysis section of the 2021 report in 50 words.", ""],
               ["What was the company's debt level in 2020?", ""],
-              ["Identify 5 key risks identified in the 2019 10k report? Respond with bullet point summaries.", ""]
+              ["Identify 5 key risks identified in the 2019 10k report? Respond with bullet point summaries.", ""],
+              ["What is the view of the management on the future of electric vehicle batteries?",""]
              ],
     concurrency_limit=16
 )
 
-
 demo.queue()
 demo.launch(auth=("demouser", os.getenv('PASSWD')))