chore: added llm sample

usecases/llm/Dockerfile

@@ -0,0 +1,17 @@
+# Use the official Python 3.9 image as the base image
+FROM python:3.9
+
+# Set the working directory within the container
+WORKDIR /usr/src/app
+
+# Copy the requirements.txt file into the container
+COPY requirements.txt .
+
+# Install dependencies from requirements.txt
+RUN pip install --no-cache-dir -r requirements.txt
+
+# Copy your application files into the container
+COPY . .
+
+EXPOSE 8000
+CMD ["python", "-m", "chainlit", "run", "app.py", "--port" ,"8000"]

usecases/llm/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2023 Mrunmay Shelar
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

usecases/llm/README.md

@@ -0,0 +1,12 @@
+# Hyper-personalized chatbots using LLMs, Dozer and Vector Databases for Banking and Finance
+
+In this example, we will be using Dozer along with Chainlit, Langchain and Pinecone. 
+
+### Pre-requisites
+
+1. Pinecone account
+2. OpenAI API key
+3. Python 3.9/ Docker
+
+### Setup
+

usecases/llm/app.py

@@ -0,0 +1,211 @@
+import chainlit as cl
+import os
+from langchain.prompts import  PromptTemplate
+from langchain.chains import LLMChain
+from langchain.embeddings import HuggingFaceEmbeddings
+from langchain.schema import HumanMessage
+## For Vector DB
+import pinecone
+import uuid
+from langchain.vectorstores import Pinecone
+from langchain.chains.conversation.memory import ConversationBufferWindowMemory
+from langchain.chains import RetrievalQA
+from langchain.agents import Tool
+#from langchain import PromptTemplate, HuggingFaceHub, LLMChain
+from langchain.embeddings import HuggingFaceEmbeddings
+from langchain.agents import initialize_agent
+from langchain.embeddings.openai import OpenAIEmbeddings
+from langchain.chat_models import ChatOpenAI
+from langchain.chains import RetrievalQAWithSourcesChain
+from langchain.prompts.chat import ChatPromptTemplate,HumanMessagePromptTemplate,SystemMessagePromptTemplate
+from chainlit import user_session
+import os
+from pydozer.api import ApiClient
+from langchain.agents import ConversationalChatAgent, AgentExecutor
+
+
+### Dozer
+
+DOZER_CLOUD_HOST = "data.getdozer.io:443"
+
+def get_api_client(app_id=None, token=None):
+    return ApiClient("financial_profile", url=DOZER_CLOUD_HOST, app_id=app_id,
+                     secure=True, token=token) if app_id else ApiClient("financial_profile", url="localhost:80")
+
+
+customer_client = get_api_client(app_id=os.environ.get("DOZER_APP_ID"), token=os.environ.get("DOZER_TOKEN"))
+
+def getCustomerData(input):
+    data = customer_client.query({"$filter": {"id":input}})
+    rec = data.records[0].record
+    id = rec.values[0].int_value
+    name = rec.values[1].string_value
+    income = rec.values[2].int_value
+    age = rec.values[3].int_value
+    dependents = rec.values[4].int_value
+    address = rec.values[5].string_value
+    prob = rec.values[6].float_value
+    credit_amt = rec.values[7].int_value
+    repay_status = rec.values[8].float_value
+    util_ratio = rec.values[9].float_value
+
+
+    return [id,name,income,age,dependents,credit_amt,repay_status,util_ratio,address, prob]
+
+
+
+#user_env = user_session.get(".env")
+OPEN_API_KEY = os.environ.get("OPENAI_API_KEY")
+PINECONE_API_KEY = os.environ.get("PINECONE_API_KEY")
+PINECONE_ENV = os.environ.get("PINE_ENV")
+ # platform.openai.com
+model_name = 'text-embedding-ada-002'
+
+
+embed = OpenAIEmbeddings(
+    model=model_name,
+    openai_api_key=OPEN_API_KEY
+)
+
+# find ENV (cloud region) next to API key in console
+index_name = 'langchain-retrieval-agent'
+pinecone.init(
+    api_key=PINECONE_API_KEY,
+    environment=PINECONE_ENV
+)
+
+
+#embeddings = OpenAIEmbeddings()
+
+
+text_field = "Answer"
+index = pinecone.Index('langchain-retrieval-agent')
+# switch back to normal index for langchain
+#index = pinecone.Index(index_name)
+
+vectorstore = Pinecone(
+    index, embed.embed_query, text_field
+)
+
+llm = ChatOpenAI(
+    openai_api_key = OPEN_API_KEY,
+    model_name='gpt-4',
+    temperature=0.0
+)
+
+conversational_memory = ConversationBufferWindowMemory(
+    memory_key='chat_history',
+    k=5,
+    return_messages=True
+)
+# retrieval qa chain
+chain = RetrievalQA.from_chain_type(
+    llm=llm,
+    chain_type="stuff",
+    retriever=vectorstore.as_retriever()
+)
+
+def setName(userName):
+    global name
+    name = userName
+
+def customerProfile(input):
+    data = getCustomerData(int(input))
+    id  = data[0]
+    name1 = data[1]
+    income = data[2]
+    age= data[3]
+    dependents= data[4]
+    address= data[5]
+    prob = data[6]
+    credit_amt= data[7]
+    repay_status= data[8]
+    util_ratio= data[9]
+
+
+    return ( f"ID = {id} ,name = {name1} ,age = {age}, income = {income}, dependents = {dependents},repay_status={repay_status}, credit utilisation ratio ={util_ratio} ,address={address}, available credit={credit_amt}, probability of default = {prob} ")
+
+
+tools = [
+    Tool(
+        name='Knowledge Base',
+        func=chain.run,
+        description=(
+            "Useful when you need general information about bank policies and bank offerings. "\
+            'use this tool when answering general knowledge queries to get '\
+            'more information about the topic'
+        )
+    ),
+    Tool(
+        name = 'Customer Data',
+        func = customerProfile,
+        description=(
+            "Useful when you need customer data to decide eligibility for a particular credit card. "
+            "Use to check the probability of default and available balance to use it for eligibility"
+        )
+    )
+]
+
+
+agent = initialize_agent(
+    agent='chat-conversational-react-description',
+    tools=tools,
+    llm=llm,
+    verbose=True,
+    max_iterations=5,
+    early_stopping_method='generate',
+    memory=conversational_memory
+)
+
+
+global res
+@cl.on_chat_start
+
+async def start():
+    intro = "Hi there, I am an assistant for Bank A. I am here to assist you with all your banking needs! Please enter your id:  "
+
+    res = await cl.AskUserMessage(content=intro,timeout=45,raise_on_timeout=True).send()
+
+    id = int(res['content'])
+    greeting = f"Hi {res['content']}. What brings you here?"
+    await cl.Message(content=greeting).send()
+    setName(res['content'])
+    # global credit 
+    # credit = getCredit(int(res['content']))
+    global customerinfo
+    customerinfo = customerProfile(int(res['content']) )
+    print(customerinfo)
+
+    agent = ConversationalChatAgent.from_llm_and_tools(
+        llm=llm,
+        tools=tools,
+        system_message = f"You have customer info of a customer which is as follows {customerinfo}. You have to use this information along with Knowledge base to decide if the customer is eligible for a credit card or not.",
+        verbose=True,
+        max_iterations=5,
+        early_stopping_method='generate',
+        memory=conversational_memory
+
+    )
+
+    global agent_chain
+    agent_chain = AgentExecutor.from_agent_and_tools(
+        agent = agent,
+        tools = tools, 
+        verbose = True,
+        max_iterations = 5,
+        early_stopping_method = 'generate',
+        memory = conversational_memory
+
+    )
+    cl.user_session.set("chain", agent_chain)
+
+
+
+@cl.on_message
+async def main(message: str):
+    agent_chain = cl.user_session.get("chain") 
+    response = await cl.make_async(agent_chain.run)(message.content)
+
+    await cl.Message(
+        content=response,
+    ).send()

usecases/llm/chainlit.md

@@ -0,0 +1,31 @@
+# Welcome to GoldenGate Bank ! 🏦💵
+
+Hi there, Customer ! 👋 We're excited to have you on board. GoldenGate Bank has been serving customers since 1970 . With a commitment to excellence, we have grown into a reputable financial institution, providing a wide range of services to meet your banking needs.
+
+# Our Services 🎄🎄🎁
+
+At 'GoldenGate Bank', we offer a comprehensive range of banking services to cater to your financial requirements. Our services include:
+
+**Savings Accounts**: Choose from a variety of savings accounts with competitive interest rates and personalized features.   
+**Current Accounts**: Our current accounts are designed to support businesses and individuals with seamless transaction capabilities.  
+**Fixed Deposits**: Benefit from attractive interest rates and flexible tenure options with our fixed deposit accounts.
+
+# Credit Cards 💳💳
+
+Experience the convenience and benefits of GoldenGate Bridge credit cards, tailored to suit your lifestyle and spending habits. We offer the following credit cards:
+
+1. **Rewards Card** 📦🎁
+Earn reward points on every purchase.       
+Eligibility: Minimum monthly income of $X,XX and a credit score of [Credit Score Requirement].
+
+2. **Travel Card** ✈️🧳  
+Enjoy exclusive travel perks and offers.  
+Eligibility: Minimum monthly income of $X,XX and a credit score of [Credit Score Requirement].
+
+3. **Cashback Card** 💵💵  
+Get cashback on eligible purchases.  
+Eligibility: Minimum monthly income of $X,XX and a credit score of [Credit Score Requirement].
+
+4. **Premium Card** 💰💲  
+Access premium benefits, including airport lounge access and concierge services.  
+Eligibility: Minimum monthly income of $X,XXX and a credit score of [Credit Score Requirement].

usecases/llm/docker-compose.yml

@@ -0,0 +1,13 @@
+version: '4.25' 
+
+services:
+  app:
+    build:
+      context: .
+      dockerfile: Dockerfile
+    volumes:
+      - .:/usr/src/app
+    env_file:
+      - .env
+    ports:
+      - '8000:8000'

usecases/llm/dozer-config.yaml

@@ -0,0 +1,63 @@
+app_name: llm-sample
+version: 1
+connections:
+  - name: onnx
+    config : !Postgres
+      user: "{{PG_USER}}"
+      password: "{{PG_PASSWORD}}"
+      host: "{{PG_HOST}}"
+      port: 5432
+      database: "{{PG_DB}}"
+  - name: s3data
+    config: !S3Storage
+      details:
+        access_key_id: "{{AWS_ACCESS_KEY}}"	
+        secret_access_key: "{{AWS_SECRET_KEY}}"
+        region: "{{AWS_REGION_S3}}"
+        bucket_name: "{{AWS_BUCKET_NAME}}"
+      tables:
+        - !Table
+          name: customers
+          config: !CSV
+            path: customer_data/
+            extension: .csv
+
+sources:
+  - name: credit
+    table_name: cred2
+    connection: onnx
+
+  - name: data
+    table_name: customers
+    connection: s3data
+
+sql: | 
+
+
+  WITH template AS 
+
+  (SELECT c.ID as ID, c.Name as name,c.Income as total_income, c.Age as age,c.Dependents as dependents, c.address as address  FROM data c ), 
+
+  res1 AS (  SELECT torch_jit( LIMIT_BAL, SEX, EDUCATION, MARRIAGE, AGE, PAY_AMT6,BILL_AMT6,PAY_6,PAY_AMT5,BILL_AMT5,PAY_5,PAY_AMT4,BILL_AMT4,PAY_4,
+                 PAY_AMT3,BILL_AMT3,PAY_3,
+                 PAY_AMT2,BILL_AMT2,PAY_2,
+                 PAY_AMT1,BILL_AMT1,PAY_0) as result, ID as id , LIMIT_BAL, 
+                 (PAY_0 + PAY_2 + PAY_3 + PAY_4 + PAY_5 + PAY_6)/5 as avg_repayment_status,
+                 (BILL_AMT1+BILL_AMT2+BILL_AMT3+BILL_AMT4+BILL_AMT5+BILL_AMT6)/(LIMIT_BAL * 6) as utilization_ratio
+                 from credit )
+  SELECT ID as id, name, total_income, age, dependents, address, result, LIMIT_BAL as avail_credit, avg_repayment_status, utilization_ratio
+  INTO customer_profile
+  FROM template t
+  JOIN res1 r
+  ON t.ID = r.id;
+
+endpoints:
+  - name: financial_profile
+    path: /customers 
+    table_name: customer_profile  
+
+
+udfs:
+  - name: torch_jit
+    config: !Onnx
+      path: ./sequential_31.onnx