Update README.md

README.md

@@ -1,10 +1,76 @@
 ---
-title: N8n Suite
-emoji: 🌍
-colorFrom: yellow
-colorTo: yellow
+title: n8n Self-Hosted AI Starter Kit
+emoji: 🚀
+colorFrom: green
+colorTo: blue
 sdk: docker
-pinned: false
+app_port: 5678
+pinned: true
+# Add hardware requests if needed, e.g. for Ollama with larger models
+# hardware: cpu-upgrade # or gpu-small / gpu-a10g-small etc.
+# For GPU, Ollama should pick it up if drivers are compatible.
+# This Dockerfile is built for CPU by default for wider compatibility.
+persistent_storage:
+  # Request persistent storage. Adjust size as needed for models, db, etc.
+  # n8n data, postgres db, qdrant vectors, ollama models
+  # Example: 10Gi for models and data
+  data:
+    type: "disk"
+    size: 10 # GiB
+    mount_path: "/data" # This is where it will be mounted in the container
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# n8n Self-Hosted AI Starter Kit on Hugging Face Spaces
+
+This Space deploys the [n8n Self-Hosted AI Starter Kit](https://github.com/n8n-io/self-hosted-ai-starter-kit) in a single container environment. It includes:
+
+*   **n8n**: Low-code automation platform.
+*   **Ollama**: Run local LLMs.
+*   **Qdrant**: Vector database.
+*   **PostgreSQL**: Database for n8n.
+
+## Setup
+
+1.  **Secrets**:
+    Before this Space can run correctly, you **MUST** configure the following secrets in your Space settings:
+    *   `N8N_ENCRYPTION_KEY`: A long, random string for encrypting sensitive n8n data.
+        *   Generate one with: `openssl rand -hex 32`
+    *   `DB_POSTGRESDB_PASSWORD`: Password for the PostgreSQL `n8n` user and database.
+        *   Choose a strong password.
+
+2.  **Wait for Build**: The first time, the Docker image will build, and services will start. This can take a few minutes. The `start.sh` script will also pull a default Ollama model (`llama3.1:8b` or similar small model), which can take time.
+
+3.  **Access n8n**: Once the Space is running, click the "Open app" button or use the direct URL to access n8n on port 5678.
+
+## How to Use
+
+1.  **n8n First-Time Setup**: When you first access n8n, you'll be prompted to create an owner account.
+2.  **Local Services**:
+    *   **Ollama**: Running on `http://localhost:11434` (internally within the Space). The default model (`llama3.1:8b` or similar) should be available. You can use the "Ollama" node in n8n and set its "Base URL" to `http://localhost:11434`.
+    *   **Qdrant**: Running on `http://localhost:6333` (internally). Use the "Qdrant" node in n8n and set its "URL" to `http://localhost:6333`.
+    *   **PostgreSQL**: n8n is already configured to use the internal PostgreSQL instance.
+3.  **Persistent Storage**:
+    *   n8n workflows and credentials are stored in `/data/.n8n`.
+    *   PostgreSQL data is stored in `/data/postgres`.
+    *   Qdrant vector data is stored in `/data/qdrant_storage`.
+    *   Ollama models are stored in `/data/.ollama`.
+    This data will persist across Space restarts if you've configured persistent storage for `/data`.
+
+4.  **Accessing Local Files in n8n**:
+    The original starter kit mentions a `/data/shared` directory for n8n to access local files. This directory is available in this Space. When using n8n nodes like "Read/Write Files from Disk", use paths like `/data/shared/yourfile.txt`.
+
+## Important Notes & Limitations
+
+*   **Single Container**: All services (n8n, Postgres, Qdrant, Ollama) run in a single Docker container. This is less robust than a multi-container `docker-compose` setup but necessary for this Hugging Face Space deployment model.
+*   **Resource Usage**: Running all these services, especially Ollama with LLMs, can be resource-intensive. Ensure your Space has adequate CPU, RAM, and storage. Consider upgrading hardware if performance is slow.
+*   **Ollama Model Management**:
+    *   A default model is pulled on startup.
+    *   To pull other models, you would typically exec into the container (possible via `hf cli` or other tools, but not directly via the Space UI) and run `ollama pull <modelname>`. Alternatively, some n8n Ollama nodes might trigger a download if the model isn't present (behavior may vary).
+*   **Initial Startup Time**: The first startup might be slow due to database initialization and model download. Check the logs for progress.
+*   **WEBHOOK_URL / N8N_EDITOR_BASE_URL**: These are automatically configured by the `start.sh` script to use the Hugging Face Space URL.
+
+## Troubleshooting
+
+*   Check the Space logs for any errors during startup or operation.
+*   Ensure secrets are correctly set.
+*   If n8n is inaccessible, one of the backend services might have failed to start.