Update app.py

app.py

@@ -2,11 +2,12 @@ import os
 import sys
 
 # --- HUGGING FACE CHROMADB FIX ---
-# This is critical for deployment on HF Spaces to prevent SQLite errors
+# Forces the use of pysqlite3-binary to avoid DB crashes
 try:
     __import__('pysqlite3')
     sys.modules['sqlite3'] = sys.modules.pop('pysqlite3')
 except ImportError:
+    # If this fails, we continue, but it might crash later if the system sqlite is old
     pass
 # ---------------------------------
 
@@ -15,10 +16,12 @@ from langchain_openai import ChatOpenAI, OpenAIEmbeddings
 from langchain_community.document_loaders import PyPDFLoader
 from langchain_text_splitters import RecursiveCharacterTextSplitter
 from langchain_chroma import Chroma
+
+# UPDATED IMPORT: This is the stable path for RetrievalQA
 from langchain.chains import RetrievalQA
 from langchain.prompts import PromptTemplate
 
-# Global variables to store state (sufficient for demo purposes)
+# Global variables
 vectorstore = None
 qa_chain = None
 
@@ -39,21 +42,21 @@ def process_pdf(file_path, api_key):
         text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
         splits = text_splitter.split_documents(docs)
         
-        # 3. Embed & Store (Chroma)
+        # 3. Embed & Store
+        # We explicitly use the lighter embedding model to save cost/time
         embeddings = OpenAIEmbeddings()
         vectorstore = Chroma.from_documents(documents=splits, embedding=embeddings)
         
-        # 4. Create Retrieval Chain with "Auditor" Persona
+        # 4. Create Chain
         llm = ChatOpenAI(model_name="gpt-4o", temperature=0)
         
-        # Custom Prompt to enforce "Audit" style behavior
         audit_template = """You are Veritas, an AI Compliance Auditor. 
         Use the following pieces of context to answer the question at the end. 
         
         RULES:
-        1. If the answer is in the text, state it clearly and reference the context.
+        1. If the answer is in the text, state it clearly.
         2. If the answer is NOT in the text, you must explicitly state: "FAIL: Information not found in source document."
-        3. Do not hallucinate or guess. 
+        3. Do not hallucinate.
         
         Context: {context}
         
@@ -70,10 +73,10 @@ def process_pdf(file_path, api_key):
             chain_type_kwargs={"prompt": QA_CHAIN_PROMPT}
         )
         
-        return "✅ Document Processed Successfully. The Veritas Auditor is ready."
+        return "✅ Document Processed. Veritas is ready."
         
     except Exception as e:
-        return f"❌ Error processing document: {str(e)}"
+        return f"❌ Error: {str(e)}"
 
 def audit_query(query):
     global qa_chain
@@ -86,47 +89,24 @@ def audit_query(query):
     except Exception as e:
         return f"Error: {str(e)}"
 
-# --- GRADIO INTERFACE ---
-with gr.Blocks(theme=gr.themes.Soft(primary_hue="slate")) as demo:
-    
-    gr.Markdown(
-        """
-        # 🛡️ Veritas: AI Compliance Auditor
-        ### Automated RAG Hallucination Detection for Financial Documentation
-        """
-    )
+# --- INTERFACE ---
+with gr.Blocks(theme=gr.themes.Soft()) as demo:
+    gr.Markdown("# 🛡️ Veritas: AI Compliance Auditor")
     
     with gr.Row():
-        with gr.Column(scale=1):
-            api_input = gr.Textbox(
-                label="OpenAI API Key", 
-                type="password", 
-                placeholder="sk-..."
-            )
-            file_input = gr.File(
-                label="Upload Financial Report (PDF)", 
-                file_types=[".pdf"]
-            )
-            upload_btn = gr.Button("Initialize Auditor", variant="primary")
-            status_output = gr.Textbox(label="System Status", interactive=False)
+        with gr.Column():
+            api_input = gr.Textbox(label="OpenAI API Key", type="password")
+            file_input = gr.File(label="Upload PDF", file_types=[".pdf"])
+            upload_btn = gr.Button("Initialize", variant="primary")
+            status = gr.Textbox(label="Status", interactive=False)
             
-        with gr.Column(scale=2):
+        with gr.Column():
             query_input = gr.Textbox(label="Audit Query")
-            audit_btn = gr.Button("Run Audit Check")
-            response_output = gr.Textbox(label="Auditor Verdict", lines=10)
+            audit_btn = gr.Button("Run Audit")
+            output = gr.Textbox(label="Verdict", lines=10)
 
-    # Button Actions
-    upload_btn.click(
-        process_pdf, 
-        inputs=[file_input, api_input], 
-        outputs=status_output
-    )
-    
-    audit_btn.click(
-        audit_query, 
-        inputs=query_input, 
-        outputs=response_output
-    )
+    upload_btn.click(process_pdf, inputs=[file_input, api_input], outputs=status)
+    audit_btn.click(audit_query, inputs=query_input, outputs=output)
 
 if __name__ == "__main__":
     demo.launch()