Merge pull request #524 from GoogleCloudPlatform/chatbot_scaffold

Add Streamlit Gemini chatbot scaffold

scaffolds/streamlit_on_cloudrun/llm_chatbot/Dockerfile

@@ -0,0 +1,13 @@
+FROM python:3.10.14
+
+WORKDIR /app
+
+COPY requirements.txt /app
+RUN pip install -r requirements.txt
+
+COPY assets /app/assets
+COPY app.py /app
+
+EXPOSE 8080
+
+CMD streamlit run --server.port 8080 --server.enableCORS false app.py

scaffolds/streamlit_on_cloudrun/llm_chatbot/README.md

@@ -0,0 +1,16 @@
+# Streamlit Application - Chat with Gemini
+
+This directory includes files to build and deploy an chatbot application on streamlit.
+
+For all the step by step instructions, refer to the `build_app.ipynb` which includes detailed explanations on each file.
+
+Or, if you want to quickly deploy and check the application, follow the steps below:
+
+1. Open `deploy.sh` and change the PROJECT_ID variable to your project id.
+
+2. Run `deploy.sh` and deploy the application to Cloud Run.
+```
+. deploy.sh
+```
+
+3. Follow the instruction printed at the end of the `deploy.sh` command, and open the application via Cloud Shell.

scaffolds/streamlit_on_cloudrun/llm_chatbot/app.py

@@ -0,0 +1,73 @@
+"""Streamlit Chatbot App"""
+
+import os
+import time
+
+import streamlit as st
+import vertexai
+from vertexai.generative_models import Content, GenerativeModel, Part
+
+st.set_page_config(page_title="Chat with Gemini", page_icon="♊")
+
+st.title("Chat with Gemini")
+
+st.markdown("Welcome to this simple web application to chat with Gemini")
+
+PROJECT_ID = os.environ.get("GCP_PROJECT")
+LOCATION = os.environ.get("GCP_REGION")
+
+vertexai.init(project=PROJECT_ID, location=LOCATION)
+
+if "gemini_model" not in st.session_state:
+    st.session_state["gemini_model"] = "gemini-1.5-flash-001"
+
+model = GenerativeModel(st.session_state["gemini_model"])
+
+# Initialize chat history
+if "messages" not in st.session_state:
+    st.session_state.messages = []
+
+# Display chat messages from history on app rerun
+for message in st.session_state.messages:
+    with st.chat_message(name=message["role"], avatar=message["avatar"]):
+        st.markdown(message["content"])
+
+
+def generate_response(input_text):
+    chat = model.start_chat(
+        history=[
+            Content(role=m["role"], parts=[Part.from_text(m["content"])])
+            for m in st.session_state.messages[:-1]
+        ]
+    )
+    return chat.send_message(input_text)
+
+
+def stream(text):
+    for word in text.split(" "):
+        yield word + " "
+        time.sleep(0.02)
+
+
+# React to user input
+if prompt := st.chat_input("Write a promt"):
+    # 1. Write the user message
+    with st.chat_message(name="user", avatar=None):
+        st.write(prompt)
+    # 2. Add user message to message history
+    st.session_state.messages.append(
+        {"role": "user", "content": prompt, "avatar": None}
+    )
+
+    # 3. Call Gemini and write the response
+    with st.chat_message(name="assistant", avatar="assets/gemini-icon.png"):
+        response = generate_response(prompt)
+        st.write_stream(stream(response.text))
+    # 4. Add Gemini response to message history
+    st.session_state.messages.append(
+        {
+            "role": "assistant",
+            "content": response.text,
+            "avatar": "assets/gemini-icon.png",
+        }
+    )