Zep Memory Chatbot with Graph Visualization (Streamlit)

A Streamlit-based chatbot that integrates Zep for long-term conversational memory and visualizes the underlying knowledge graph using st_link_analysis.

This project demonstrates:

• Persistent chat memory using Zep
• Context-aware responses from stored conversation history
• Exporting Zep graph data into JSON
• Interactive graph visualization inside a Streamlit app

---

Features

• 💬 Chatbot with long-term memory (Zep)
• 🧠 Context retrieval for improved LLM responses
• 🕸️ Zep graph (nodes & edges) export
• 📊 Graph visualization using st_link_analysis
• 🖥️ Single Streamlit application

---

Project Structure

.
├── streamlit_app.py         # Main Streamlit app (chat + graph visualization)
├── zep_graph.py             # Optional graph export/debug utility
├── zep_memory_bot.py        # Optional CLI-based chatbot
├── requirements.txt
├── docker-compose.yml       # Optional (remove secrets before committing)
├── data/                    # Optional local data (not required)
├── zep_thread_id.txt        # Runtime file (do not commit)
└── .env                     # Secrets file (do not commit)

---

Prerequisites

• Python 3.10 or higher
• Zep account (Zep Cloud API key or Zep OSS)
• OpenAI API key (or compatible LLM provider)

---

Installation

1. Clone the repository

git clone <your-github-repo-url>
cd <repo-folder>

2. Create and activate a virtual environment

python -m venv .venv

# Windows
.venv\Scripts\activate

# macOS / Linux
source .venv/bin/activate

3. Install dependencies

pip install -r requirements.txt

---

Environment Configuration

Create a .env file in the project root:

OPENAI_API_KEY=your_openai_api_key
ZEP_API_KEY=your_zep_api_key
OPENAI_MODEL=gpt-4o-mini
ZEP_USER_ID=vk_user

⚠️ Never commit .env or API keys to GitHub.

---

Running the Application

streamlit run streamlit_app.py

Once running:

• Start chatting with the bot
• Memory is stored and retrieved from Zep
• Use the UI control to refresh and visualize the Zep knowledge graph

---

Graph Data Format

The Zep graph is converted into the format required by st_link_analysis:

{
  "nodes": [
    { "data": { "id": "node_1", "label": "User" } }
  ],
  "edges": [
    {
      "data": {
        "id": "edge_1",
        "source": "node_1",
        "target": "node_2",
        "label": "related_to"
      }
    }
  ]
}

---

GitHub Safety Notes

• Do not commit .env
• Do not commit zep_thread_id.txt
• Remove API keys from docker-compose.yml before pushing
• Use environment variables for all secrets

---

License

Add a license (MIT recommended) if you plan to make this repository public or reusable.