Create app.py

app.py

@@ -0,0 +1,149 @@
+import subprocess
+import sys
+import os
+import uuid
+import json
+from pathlib import Path
+
+from huggingface_hub import login, CommitScheduler
+import openai
+import gradio as gr
+# Install dependencies if not already installed
+def install_packages():
+    packages = ["openai", "langchain_community", "sentence-transformers", "chromadb", "huggingface_hub", "python-dotenv"]
+    for package in packages:
+        subprocess.check_call([sys.executable, "-m", "pip", "install", package])
+
+install_packages()
+from dotenv import load_dotenv
+from langchain_community.embeddings.sentence_transformer import SentenceTransformerEmbeddings
+from langchain_community.vectorstores import Chroma
+# Load environment variables from .env file
+load_dotenv()
+
+# Get API tokens from environment variables
+openai.api_key = os.getenv("OPENAI_API_KEY")  # Ensure OPENAI_API_KEY is in your .env file
+hf_token = os.getenv("hf_token")
+
+if not hf_token:
+    raise ValueError("Hugging Face token is missing. Please set 'hf_token' as an environment variable.")
+
+# Log in to Hugging Face
+login(hf_token)
+
+print("Logged in to Hugging Face successfully.")
+
+# Set up embeddings and vector store
+embeddings = SentenceTransformerEmbeddings(model_name="thenlper/gte-large")
+collection_name = 'report-10k-2024'
+
+vectorstore_persisted = Chroma(
+    collection_name=collection_name,
+    persist_directory='./report_10kdb',
+    embedding_function=embeddings
+)
+
+# Set up the retriever
+retriever = vectorstore_persisted.as_retriever(
+    search_type='similarity',
+    search_kwargs={'k': 5}
+)
+
+# Define Q&A system messages
+qna_system_message = """
+You are an AI assistant for Finsights Grey Inc., helping automate extraction, summarization, and analysis of 10-K reports.
+Your responses should be based solely on the context provided.
+If an answer is not found in the context, respond with "I don't know."
+"""
+
+qna_user_message_template = """
+###Context
+Here are some documents that are relevant to the question.
+{context}
+###Question
+{question}
+"""
+
+# Define the predict function
+def predict(user_input, company):
+    filter = f"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 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)}
+    ]
+
+    try:
+        # Get response from the LLM
+        response = openai.Completion.create(
+            model='gpt-3.5-turbo',
+            messages=prompt,
+            temperature=0
+        )
+        prediction = response['choices'][0]['message']['content']
+
+    except Exception as e:
+        prediction = f"Error: {str(e)}"
+
+    # Log inputs and outputs to a local log file
+    log_file = Path("logs/") / f"data_{uuid.uuid4()}.json"
+    log_file.parent.mkdir(parents=True, exist_ok=True)  # Create log directory if it doesn't exist
+
+    scheduler = CommitScheduler(
+        repo_id="RAGREPORTS-log",
+        repo_type="dataset",
+        folder_path=log_file.parent,
+        path_in_repo="data",
+        every=2
+    )
+
+    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
+
+def get_predict(question, company):
+    company_map = {
+        "AWS": "aws",
+        "IBM": "IBM",
+        "Google": "Google",
+        "Meta": "meta",
+        "Microsoft": "msft"
+    }
+    selected_company = company_map.get(company)
+    if not selected_company:
+        return "Invalid company selected"
+
+    return predict(question, selected_company)
+
+# Set up the Gradio UI
+with gr.Blocks(theme="gradio/seafoam@>=0.0.1,<0.1.0") as demo:
+    with gr.Row():
+        company = gr.Radio(["AWS", "IBM", "Google", "Meta", "Microsoft"], label="Select a company")
+        question = gr.Textbox(label="Enter your question")
+
+    submit = gr.Button("Submit")
+    output = gr.Textbox(label="Output")
+
+    submit.click(
+        fn=get_predict,
+        inputs=[question, company],
+        outputs=output
+    )
+
+demo.queue()
+demo.launch()