Merge pull request #1103 from i-dot-ai/feature/streamlit_app

Feature/streamlit app

.vscode/redbox.code-workspace

@@ -5,6 +5,13 @@
 			"path": "../django_app"
 		},
 		{
+<<<<<<< Updated upstream
+=======
+			"name": "streamlit_app",
+			"path": "../streamlit_app"
+		},
+		{
+>>>>>>> Stashed changes
 			"name": "Redbox Core",
 			"path": "../redbox-core"
 		},

streamlit_app/app.py

@@ -0,0 +1,60 @@
+# Set up
+import streamlit as st
+from redbox.app import Redbox
+from redbox.models.chain import RedboxQuery, RedboxState, AISettings
+from uuid import uuid4
+from redbox.models.settings import Settings
+from dotenv import load_dotenv
+import asyncio
+
+# Load Redbox environment variables
+load_dotenv("../.env")
+
+
+# Function defining how the Redbox streamlit app will work
+def run_streamlit():
+    # Create two tabs within the interface, one labeled "chat", the other labeled "settings"
+    chat_tab, settings_tab = st.tabs(["chat", "settings"])
+    # Create a list that will store the chat history, if one does not already exist
+    if "messages" not in st.session_state:
+        st.session_state.messages = []
+    # Create a list that will store the AI settings, if one does not already exist
+    if "ai_settings" not in st.session_state:
+        st.session_state.ai_settings = AISettings()
+    # Create an instance of Redbox, if one does not already exist
+    if "redbox" not in st.session_state:
+        st.session_state.redbox = Redbox(env=Settings(), debug=True)
+
+    with chat_tab:
+        # Iterate over the messages list, display each message in the chat window with the role and corresponding text
+        for message in st.session_state.messages:
+            with st.chat_message(message["role"]):
+                st.markdown(message["text"])
+        # Present an input field where the user can type a message to Redbox
+        if prompt := st.chat_input("How can I help you today?"):
+            # New messages should be added to the chat history under the "user" role
+            st.session_state.messages.append({"role": "user", "text": prompt})
+            # Construct a Redbox object with required information
+            state = RedboxState(
+                request=RedboxQuery(
+                    question=prompt,
+                    s3_keys=[],
+                    user_uuid=uuid4(),
+                    chat_history=[],
+                    ai_settings=st.session_state.ai_settings,
+                ),
+            )
+            # Run an asynchronous function from the Redbox instance to get the AI's response
+            response: RedboxState = asyncio.run(st.session_state.redbox.run(state))
+            llm_answer = response["text"]
+            # route = response["route_name"] - to be added back in if / when routing is added
+            # Display the AI's response in the chat under the "ai" role and append it to the chat history
+            with st.chat_message("ai"):
+                st.write(llm_answer)
+            st.session_state.messages.append({"role": "ai", "text": llm_answer})
+
+
+run_streamlit()
+
+# getting the chat history to work, showing the routes, and sources, and getting it to look like Redbox.
+# Go back to James re-files, data and streaming