add streamlit app

app.py

@@ -0,0 +1,51 @@
+import os
+
+import streamlit as st
+from langchain.graphs import Neo4jGraph
+from langchain_openai.chat_models import ChatOpenAI
+from streamlit import session_state as ss
+
+from components import chat, display_chat_history, sidebar
+from src.ps_genai_agents.agents.graph import create_text2cypher_graph_agent
+
+
+def initialize_state() -> None:
+    """
+    Initialize the application state.
+    """
+
+    if "agent" not in ss:
+        ss["llm"] = ChatOpenAI(model="gpt-4o")
+        ss["graph"] = Neo4jGraph(
+            url=os.environ.get("IQS_NEO4J_URI"),
+            username=os.environ.get("IQS_NEO4J_USERNAME"),
+            password=os.environ.get("IQS_NEO4J_PASSWORD"),
+            enhanced_schema=True,
+            driver_config={"liveness_check_timeout": 0},
+        )
+        ss["agent"] = create_text2cypher_graph_agent(
+            chat_llm=ss["llm"], neo4j_graph=ss["graph"]
+        )
+        ss["messages"] = list()
+        ss["source"] = "IQS"
+
+
+def run_app():
+    """
+    Run the Streamlit application.
+    """
+
+    st.title("PS GenAI Retreat Workshop")
+    sidebar()
+    display_chat_history()
+    # Prompt for user input and save and display
+    if question := st.chat_input():
+        ss["current_question"] = question
+
+    if "current_question" in ss:
+        chat(ss.current_question)
+
+
+if __name__ == "__main__":
+    initialize_state()
+    run_app()

components/__init__.py

@@ -0,0 +1,4 @@
+"""This module contains the components used in the Streamlit app."""
+
+from .chat import chat, display_chat_history
+from .sidebar import sidebar

components/chat.py

@@ -0,0 +1,118 @@
+import io
+import zipfile
+from typing import Any, Dict, List
+from uuid import uuid4
+
+import pandas as pd
+import streamlit as st
+from neo4j.exceptions import SessionExpired
+from streamlit import session_state as ss
+
+
+def append_user_question(question: str) -> None:
+    ss.messages.append({"role": "user", "content": question})
+    st.chat_message("user").markdown(question)
+
+
+def append_llm_response(question: str) -> None:
+    with st.chat_message("assistant"):
+        message_placeholder = st.empty()
+        message_placeholder.status("thinking...")
+        print("question: ", question)
+        response: Any = ss.agent.invoke({"input": question, "chat_history": []})[
+            "agent_outcome"
+        ]
+
+        message_placeholder.markdown(response.answer)
+
+    show_response_information(response=response)
+
+    ss.messages.append({"role": "assistant", "content": response})
+
+
+def show_response_information(response: Any) -> None:
+    if (
+        hasattr(response, "cypher")
+        and response.cypher
+        and not response.answer.startswith("I can only generate queries based on")
+    ):
+        download_csv_button(cypher_result=response.cypher_result)
+
+        with st.expander("Cypher"):
+            if isinstance(response.cypher, str):
+                st.code(response.cypher, language="cypher")
+                st.json(response.cypher_result, expanded=False)
+            else:
+                [
+                    (
+                        st.write(response.sub_questions[i]),
+                        st.code(response.cypher[i], language="cypher"),
+                        st.json(response.cypher_result[i], expanded=False),
+                    )
+                    for i in range(len(response.cypher))
+                ]
+
+    if hasattr(response, "sources") and response.sources:
+        with st.expander("Vector Search"):
+            st.write("Source Node IDs")
+            st.write(response.sources)
+
+
+def chat(question: str):
+    try:
+        append_user_question(question=question)
+        append_llm_response(question=question)
+    except SessionExpired as e:
+        st.error("Neo4j Session expired. Please restart the application.")
+
+
+def display_chat_history() -> None:
+    for message in ss.messages:
+        print(message)
+        with st.chat_message(message["role"]):
+            if message["role"] == "user":
+                st.markdown(message["content"])
+            else:
+                st.markdown(message["content"].answer)
+        if not isinstance(message["content"], str):
+            show_response_information(response=message["content"])
+
+
+def prepare_csv(cypher_result: List[Dict[str, Any]]) -> str:
+    # if not cypher_result: return pd.DataFrame().to_csv().encode("utf-8")
+
+    index = [i for i in range(len(cypher_result[0].values()))]
+    return pd.DataFrame(data=cypher_result).to_csv(index=index).encode("utf-8")
+
+
+@st.experimental_fragment()
+def download_csv_button(cypher_result: List[Dict[str, Any]]) -> None:
+    try:
+        print("cypher result in button", cypher_result)
+        if len(cypher_result) > 0 and isinstance(cypher_result[0], list):
+            content = [prepare_csv(result) for result in cypher_result if result]
+            buf = io.BytesIO()
+            with zipfile.ZipFile(buf, "x") as zip:
+                for file_num, csv in enumerate(content):
+                    zip.writestr(f"cypher_result_part_{str(file_num+1)}.csv", csv)
+
+            st.download_button(
+                label="Download Cypher Results Tables as CSV",
+                data=buf.getvalue(),
+                file_name="cypher_results.zip",
+                mime="application/zip",
+                help="The cypher results .csv files in a .zip.",
+                key=str(uuid4()),
+            )
+        else:
+            csv = prepare_csv(cypher_result=cypher_result)
+            st.download_button(
+                label="Download Cypher Results Table as CSV",
+                data=csv,
+                file_name=f"cypher_results.csv",
+                mime="text/csv",
+                help="The cypher results .csv file.",
+                key=str(uuid4()),
+            )
+    except Exception as e:
+        print("Unable to generate Download Button for most recent question.")

components/questions.py

@@ -0,0 +1,19 @@
+from typing import List, Literal
+
+iqs_questions = [
+    "How many vehicles are there?",
+    "Summarize the responses under fcd10 for honda pilot. What is the men to women proportion for these responses and what is the problem for fcd10?",
+    "What are the top 5 most severe problems for women aged 30-34 for all Acura models?",
+    "Please summarize the verbatims for 2023 RDX for question 010 Trunk/TG Touch-Free Sensor DTU and create categories for the problems. As an output, I want the summary, corresponding categories and their verbatims",
+    "What are the top 5 problems about seats for each age buckets for men over the age of 53?",
+    "What are the top 5 problems about seats for each age buckets over the age of 53? Summarize the responses for each bucket",
+    "What is the customer with the most reported problems? Can you list the problems, summarize them and include the problem id's as well as the customer gender and age range.",
+    "Summarize and compare the sentiment for responses related to noise and sound in Honda Accord and Honda Pilot for women and return the number of responses considered.",
+    "What color is the sky?",
+]
+
+patient_journey_questions = []
+
+
+def get_demo_questions(source: Literal["IQS", "Patient Journey"]) -> List[str]:
+    return iqs_questions if source == "IQS" else patient_journey_questions

components/sidebar.py

@@ -0,0 +1,52 @@
+import os
+
+from langchain.graphs import Neo4jGraph
+from streamlit import session_state as ss
+from streamlit import sidebar as sb
+
+from src.ps_genai_agents.agents.graph import create_text2cypher_graph_agent
+
+from .questions import get_demo_questions
+
+
+def sidebar() -> None:
+    """
+    The Streamlit app side bar.
+    """
+
+    source = sb.radio("Data Source", options=["IQS", "Patient Journey"])
+    if source != ss["source"]:
+        ss["source"] = source
+        if source == "IQS":
+            ss["graph"] = Neo4jGraph(
+                url=os.environ.get("IQS_NEO4J_URI"),
+                username=os.environ.get("IQS_NEO4J_USERNAME"),
+                password=os.environ.get("IQS_NEO4J_PASSWORD"),
+                enhanced_schema=True,
+                driver_config={"liveness_check_timeout": 0},
+            )
+        else:
+            ss["graph"] = Neo4jGraph(
+                url=os.environ.get("PJ_NEO4J_URI"),
+                username=os.environ.get("PJ_NEO4J_USERNAME"),
+                password=os.environ.get("PJ_NEO4J_PASSWORD"),
+                enhanced_schema=True,
+                driver_config={"liveness_check_timeout": 0},
+            )
+        ss["agent"] = create_text2cypher_graph_agent(
+            chat_llm=ss["llm"], neo4j_graph=ss["graph"]
+        )
+
+    demo_questions = get_demo_questions(source=ss["source"])
+
+    sb.title("Demo Questions")
+    with sb.expander("Demo Questions"):
+        for ex in demo_questions:
+            if sb.button(label=ex, key=ex):
+                ss.current_question = ex
+
+    sb.divider()
+    if len(ss.messages) > 0:
+        if sb.button("Reset Chat", type="primary"):
+            ss.messages = list()
+            del ss.current_question