Spaces:
Running
Running
| title: 🧠 MemMachine Playground – AI Memory for LLMs & Agents | |
| emoji: 🧠 | |
| colorFrom: green | |
| colorTo: gray | |
| sdk: docker | |
| pinned: true | |
| short_description: Official MemMachine playground for AI, agent memory for LLMs | |
| app_port: 7860 | |
| hf_oauth: true | |
| hf_oauth_scopes: | |
| hf_oauth_authorized_org: | |
| - Memverge | |
| tags: | |
| - ai-memory | |
| - persistent-memory | |
| - agent-memory | |
| - multi-llm | |
| - llm-playground | |
| - ai-agents | |
| - memory | |
| - chatbot | |
| - open-source | |
| - agents | |
| license: apache-2.0 | |
| # 🧠 MemMachine Playground | |
| **This is the official interactive playground for MemMachine (memmachine.ai) — the universal memory layer for AI agents** | |
| A powerful playground for experimenting with various Large Language Models (LLMs) enhanced by MemMachine's persistent memory system. Compare conversations with and without memory to see how AI memory transforms your interactions. | |
| ## ✨ Features | |
| - **Multiple LLM Providers**: OpenAI, Anthropic (Claude), AWS Bedrock and more | |
| - **Persistent Memory**: AI remembers your conversations across sessions | |
| - **Profile Memory**: Builds a personalized profile of you over time | |
| - **Episodic Memory**: Remembers specific conversations and context | |
| - **Secure Authentication**: Token-based authentication to protect your memories | |
| - **Session Management**: Create and manage multiple conversation sessions | |
| ## 🚀 Getting Started | |
| ### Step 1: Create a Hugging Face Read Token | |
| To authenticate and access your personalized memories, you'll need a Hugging Face access token: | |
| 1. **Go to Hugging Face Settings** | |
| - Visit: https://huggingface.co/settings/tokens | |
| - Or click your profile → Settings → Access Tokens | |
| 2. **Create a New Token** | |
| - Click **"New token"** button | |
| - **Name**: Enter a descriptive name (e.g., "MemMachine Playground") | |
| - **Role**: Select **"Read"** (this is sufficient for authentication) | |
| - Click **"Generate token"** | |
| 3. **Copy Your Token** | |
| - ⚠️ **Important**: Copy the token immediately (it starts with `hf_`) | |
| - You won't be able to see it again after closing the page | |
| - Store it securely | |
| ### Step 2: Authenticate in the App | |
| 1. **Open the Playground** | |
| - The app will prompt you to authenticate on first use | |
| 2. **Enter Your Token** | |
| - Paste your Hugging Face Read token in the authentication field | |
| - Click **"Authenticate"** | |
| - Your username will be automatically detected and locked | |
| 3. **Start Chatting** | |
| - Your memories are now personalized to your account | |
| - All conversations are stored securely under your username | |
| ## 💡 How to Use | |
| ### Basic Usage | |
| 1. **Choose a Model** | |
| - Select from the dropdown in the sidebar | |
| - Models include GPT-4, Claude, Bedrock, and more | |
| 2. **Start a Conversation** | |
| - Type your message in the chat input | |
| - The AI will respond using your memory context | |
| 3. **Compare with Memory** | |
| - Toggle "Compare with Control persona" to see responses with/without memory | |
| - See how memory enhances the AI's understanding | |
| ## 🔍 Keywords (for discoverability) | |
| **Core concepts** | |
| - AI memory | |
| - Persistent memory for LLMs | |
| - Agent memory | |
| - Memory for AI agents | |
| **Use cases** | |
| - AI agents with memory | |
| - Long-term memory for LLMs | |
| **Product & platform** | |
| - Multi-LLM playground | |
| - MemMachine AI memory | |
| - LLM memory | |
| ### Advanced Features | |
| - **Multiple Sessions**: Create different conversation sessions for different topics | |
| - **Persona Selection**: Test with different user personas (Charlie, Jing, Charles, Control) | |
| - **Profile Management**: View and delete your AI profile when needed | |
| - **Memory Search**: The AI searches your past conversations for relevant context | |
| ## 🔒 Security & Privacy | |
| - **Token Authentication**: Your Hugging Face token is used only to identify your username | |
| - **Session-Only Storage**: Tokens are stored only in your browser session | |
| - **Isolated Memories**: Each user's memories are completely isolated | |
| - **No Token Sharing**: Your token is never shared or stored permanently | |
| ## 🏗️ Architecture | |
| - **Frontend**: Streamlit (runs in this Hugging Face Space) | |
| - **Backend**: MemMachine server running on EC2 | |
| - **Memory Storage**: Neo4j (graph) + Postgres (vector search) | |
| - **Authentication**: Hugging Face token-based authentication | |
| ## 📝 Notes | |
| - **Token Expiration**: Read tokens don't expire automatically - they remain valid until you revoke them | |
| - **Memory Persistence**: Your memories are tied to your username, not your token. You can create new tokens without losing memories | |
| - **Rate Limiting**: The backend implements rate limiting (50 requests/minute) to prevent abuse | |
| ## 🆘 Troubleshooting | |
| **"Invalid token" error?** | |
| - Make sure you copied the entire token (including `hf_` prefix) | |
| - Verify the token has "Read" permissions | |
| - Try creating a new token if the issue persists | |
| **Can't see my memories?** | |
| - Ensure you're authenticated with the same username | |
| - Memories are stored per username, not per token | |
| **Need help?** | |
| - Check that your token has Read permissions | |
| - Verify you're using a valid Hugging Face account | |
| --- | |
| **Powered by [MemMachine](https://github.com/memverge/memmachine)** 🧠 |