diff --git a/05_src/assignment_chat/__init__.py b/05_src/assignment_chat/__init__.py
new file mode 100644
index 00000000..e69de29b
diff --git a/05_src/assignment_chat/app.py b/05_src/assignment_chat/app.py
new file mode 100644
index 00000000..d16163e9
--- /dev/null
+++ b/05_src/assignment_chat/app.py
@@ -0,0 +1,36 @@
+import os
+import gradio as gr
+from dotenv import load_dotenv
+from langchain.chat_models import init_chat_model
+from langchain_core.messages import HumanMessage, AIMessage, SystemMessage
+
+
+load_dotenv(r'C:\Users\pca_9\Desktop\deploying-ai\05_src\.secrets')
+
+
+gateway_key = os.getenv('API_GATEWAY_KEY')
+
+llm = init_chat_model(
+    "gpt-4o-mini", 
+    model_provider="openai",
+    base_url='https://k7uffyg03f.execute-api.us-east-1.amazonaws.com/prod/openai/v1',
+    default_headers={"x-api-key": gateway_key},
+    api_key="any_value"
+)
+
+def simple_chat(message, history):
+    # Guardrails: No Cats, Dogs, Horoscopes, or Taylor Swift
+    system_prompt = SystemMessage(content="You are a witty assistant. You refuse to talk about Taylor Swift, pets, or zodiac signs.")
+    langchain_messages = [system_prompt]
+    
+    for msg in history:
+        role = HumanMessage if msg['role'] == 'user' else AIMessage
+        langchain_messages.append(role(content=msg['content']))
+    
+    langchain_messages.append(HumanMessage(content=message))
+    response = llm.invoke(langchain_messages)
+    return response.content
+
+if __name__ == "__main__":
+    demo = gr.ChatInterface(fn=simple_chat, type="messages")
+    demo.launch()
\ No newline at end of file
diff --git a/05_src/assignment_chat/readme.md b/05_src/assignment_chat/readme.md
new file mode 100644
index 00000000..90b4ee51
--- /dev/null
+++ b/05_src/assignment_chat/readme.md
@@ -0,0 +1,15 @@
+# My AI Chatbot - Assignment 1
+
+## What is this?
+This is a web-based chat app I built using **Gradio** and **LangChain**. It’s hooked up to a GPT-4o-mini model running through an Amazon API Gateway. I’ve tuned the assistant to be witty and helpful, but it has some "no-go" zones to keep the conversation on track.
+
+## What it can do
+* **Natural Conversations:** It handles back-and-forth chatting and actually remembers what we talked about earlier in the session.
+* **Strict Guardrails:** I’ve programmed the bot to dodge specific topics. It will politely (or wittily) refuse to talk about Pets, Zodiac signs, or Taylor Swift.
+* **Secure Connection:** It connects to a cloud-based API Gateway using custom headers to keep the API key safe.
+
+## Behind the Scenes (My Decisions)
+1.  **Finding the Secrets:** Pathing can be a headache in Python, so I used an absolute path to the `.secrets` file. This ensures the app finds the API key no matter which folder you launch it from.
+2.  **Mapping Messages:** Gradio and LangChain speak different "languages" when it comes to chat history. I wrote a loop to translate Gradio’s history into `HumanMessage` and `AIMessage` objects so the bot doesn't lose its train of thought.
+3.  **Keeping it Standard:** I stuck strictly to the libraries provided in the course (LangChain, Gradio, and Dotenv). No extra "fluff" or outside packages were used, making it easy to run on any standard setup.
+4.  **Prompt-Based Control:** Instead of writing a bunch of complex "if/else" code to block topics, I used a `SystemMessage`. This sets the bot's "personality" and rules right at the source, making the rejections feel more natural and less like a computer error.
\ No newline at end of file