Streamlit Chat Bot

Introduction

This project is lightweight Streamlit application that wraps a LangChain-powered chat interface. Users can converse with an OpenAI model in the browser, while conversation history is preserved with Streamlit session state. Responses are streamed token-by-token for a responsive feel.

Features

• Streaming chat responses using langchain-openai and StrOutputParser for incremental output.
• Persistent conversation history stored in st.session_state so earlier turns stay visible during a session.
• Chat-style UI built with st.chat_message and st.chat_input components.
• Environment-driven configuration that loads secrets (like OPENAI_API_KEY) from a local .env file via python-dotenv.

Project structure

.
├── main.py                # CLI entry that prints a greeting (not used by Streamlit)
├── src/
│   └── main.py            # Streamlit app with chat logic
├── pyproject.toml         # Project metadata and dependencies
└── uv.lock                # Resolved dependency lockfile for uv/pip

Or see the full source code on GitHub.

Requirements

• Python 3.9+
• An OpenAI API key with access to chat models
• The dependencies listed in pyproject.toml (Streamlit, LangChain, python-dotenv, etc.)

Setup

1. Install dependencies (use your preferred installer):

   # Using pip
   pip install -e .
   
   # Or using uv
   uv pip install -e .

2. Create a .env file in the project root with your API key:

   echo "OPENAI_API_KEY=sk-..." > .env

3. Verify environment variables are loaded (the app calls load_dotenv() at startup). Your key must be available as OPENAI_API_KEY before the first chat request.

Running the app

Start the Streamlit UI pointed at the application module:

streamlit run src/main.py

The app sets a page title, icon, and wide layout. Once running, open the provided local URL in your browser to begin chatting.

How it works

• On launch, src/main.py loads environment variables and initializes st.session_state["chat_history"] to hold a list of LangChain HumanMessage and AIMessage objects. 【F:src/main.py†L11-L69】
• The helper get_chat_response builds a simple prompt template that includes prior turns and the latest user input, then streams the model response via ChatOpenAI configured with your API key. 【F:src/main.py†L27-L44】
• Previous messages are rendered with st.chat_message, preserving user/assistant roles. User input is collected with st.chat_input, appended to history, and the streamed assistant response is written to the page and stored. 【F:src/main.py†L47-L69】

Tips for use

• Keep your .env file out of version control to avoid leaking keys.
• Because responses stream, network interruptions may leave partial output; resubmit if needed.
• To reset the conversation, clear the chat_history in the sidebar (if using Streamlit's session state inspector) or restart the app.

Development notes

• main.py at the repository root is only a placeholder CLI that prints a greeting; the Streamlit experience lives in src/main.py.
• No automated tests are included. When contributing features, consider adding unit or integration tests around prompt construction and session behavior.
• Follow Streamlit's guidance for secrets management in production deployments (e.g., environment variables or Streamlit Community Cloud secrets).