init

app.py

@@ -0,0 +1,60 @@
+import gradio as gr
+from rag import rbc_product
+from tool import rival_product
+
+with gr.Blocks() as demo:
+  with gr.Tab("RAG"):
+    gr.Markdown("""
+Marketing
+------------
+GraphRAG: Models customer-product relationship networks for next-best-action predictions
+
+DSPy: Optimizes cross-sell/upsell prompt variations through A/B testing
+
+Risk & Audit
+------------
+GraphRAG: Maps transactional relationships into dynamic knowledge graphs to detect multi-layered fraud patterns
+
+Tool Use: Integrates fraud detection APIs, anomaly scoring models, and regulatory compliance checkers
+
+DSPy: Optimizes fraud explanation prompts for regulatory reporting
+    """)
+    gr.Markdown("""
+    Retrieval: Public RBC Product Data
+    
+    Recommend: RBC Product 
+    """)
+    in_verbatim = gr.Textbox(label="Verbatim")
+    out_product = gr.Textbox(label="Product")
+
+
+    gr.Examples(
+      [
+        ["Low APR and great customer service. I would highly recommend if you’re looking for a great credit card company and looking to rebuild your credit. I have had my credit limit increased annually and the annual fee is very low."]
+      ],
+      [in_verbatim]
+    )
+    btn_recommend=gr.Button("Recommend")
+    btn_recommend.click(fn=rbc_product, inputs=in_verbatim, outputs=out_product)
+
+  with gr.Tab("Tool Use"):
+    gr.Markdown("""
+    Retrieval: Public Product Data using Tavily Search
+    
+    Recommend: Competition Product 
+    """)
+    in_verbatim = gr.Textbox(label="Verbatim")
+    out_product = gr.Textbox(label="Product")
+
+    gr.Examples(
+      [
+        ["Low APR and great customer service. I would highly recommend if you’re looking for a great credit card company and looking to rebuild your credit. I have had my credit limit increased annually and the annual fee is very low."]
+      ],
+      [in_verbatim]
+    )
+    btn_recommend=gr.Button("Recommend")
+    btn_recommend.click(fn=rival_product, inputs=in_verbatim, outputs=out_product)
+
+
+
+demo.launch(allowed_paths=["./xgb","./ts"])

product2.csv

@@ -0,0 +1,11 @@
+product,purpose,benefit,fee
+RBC Newcomer Advantage,Support newcomers settling in Canada,No credit history required for car/home loans,Varies by loan type
+RBC Advantage Banking for Students,Banking for full-time students,"Unlimited debits, free Interac e-Transfers",No monthly fee
+RBC Day to Day Banking,Basic banking needs,"12 free debits/month, unlimited transit transactions",$4/month
+RBC Digital Choice Business Account,Digital banking for small businesses,Unlimited electronic transactions,$6/month
+RBC Professionals' Package,Financial solutions for accredited professionals,Tailored advice and solutions,Varies by service
+RBC Newcomer Mortgage,Home financing for newcomers,No Canadian credit history required,Market rates
+RBC Student Credit Card,Credit building for students,Fee waiver for first year,$0 first year
+RBC Retirement Savings Account,Retirement planning,Tax-advantaged savings options,$25/year
+RBC Small Business Loan,Business financing,Flexible repayment terms,Prime+2%
+RBC Avion Credit Card,Travel rewards,Earn Avion points on purchases,$120 annual fee

rag.py

@@ -0,0 +1,42 @@
+import dspy
+
+import pandas as pd
+
+#lm = dspy.LM('ollama_chat/deepseek-r1', api_base='http://localhost:11434', api_key='')
+lm = dspy.LM('huggingface/Qwen/Qwen2.5-Coder-32B-Instruct')
+dspy.configure(lm=lm)
+
+df = pd.read_csv("product2.csv")
+df['content']=df['product']+"; "+df['purpose']+"; "+df['benefit']+"; "+df['fee']
+
+corpus = [row['content'] for i,row in df.iterrows()]
+
+embedder = dspy.Embedder('openai/text-embedding-3-small', dimensions=512)
+class RecommendProduct(dspy.Signature):
+  """
+  Recommend RBC financial product based on verbatim
+  """
+  context = dspy.InputField(desc="may contain relevant product information")
+  verbatim = dspy.InputField()
+  product = dspy.OutputField(desc="product name with benefit")
+
+class RAG(dspy.Module):
+  def __init__(self, num_passages=3):
+    super().__init__()
+    self.retrieve = dspy.retrievers.Embeddings(embedder=embedder, corpus=corpus, k=num_passages)
+    self.recommender = dspy.ChainOfThought(RecommendProduct)
+
+  def forward(self, verbatim):
+    context = self.retrieve(verbatim).passages
+    prediction = self.recommender(context=context, verbatim=verbatim)
+    return dspy.Prediction(context=context, product=prediction.product)
+
+
+customer="Low APR and great customer service. I would highly recommend if you’re looking for a great credit card company and looking to rebuild your credit. I have had my credit limit increased annually and the annual fee is very low."
+
+
+qa = RAG(num_passages=2)
+
+def rbc_product(customer:str):
+  response = qa(verbatim=f"Which RBC personal banking product best serve the follow customer needs, pain points: {customer}")
+  return response.product

requirements.txt

@@ -0,0 +1,12 @@
+markdownify
+requests
+duckduckgo_search
+pandas
+langchain
+langgraph
+litellm==1.63
+openai==1.61.0
+dspy
+llama-index
+faiss-cpu
+tavily-python

tool.py

@@ -0,0 +1,25 @@
+import os
+import dspy
+
+from dspy.predict.react import Tool
+from tavily import TavilyClient
+
+#lm = dspy.LM('ollama_chat/deepseek-r1', api_base='http://localhost:11434', api_key='')
+lm = dspy.LM('huggingface/Qwen/Qwen2.5-Coder-32B-Instruct')
+dspy.configure(lm=lm)
+
+search_client = TavilyClient(api_key=os.environ["T_TOKEN"])
+
+INST="""Recommend banking financial product based on verbatim"""
+def web_search(query: str) -> list[str]:
+    """Run a web search and return the personal banking product from the top 5 search results"""
+    response = search_client.search(query)
+    return [r["content"] for r in response["results"]]
+
+agent = dspy.ReAct("verbatim -> product", tools=[Tool(web_search)])
+
+customer="Low APR and great customer service. I would highly recommend if you’re looking for a great credit card company and looking to rebuild your credit. I have had my credit limit increased annually and the annual fee is very low."
+
+def rival_product(customer:str):
+    prediction = agent(verbatim=f"Which banking product best serve this customer needs, pain points: {customer}")
+    return prediction.product