Appwrite — Open-Source Backend for AI Apps

What it is

Appwrite is an open-source backend-as-a-service platform that provides authentication, databases, file storage, serverless functions, and messaging APIs. It runs as a set of Docker containers and exposes REST and GraphQL APIs. SDKs are available for web, mobile, and server-side languages.

It targets developers building AI applications who need a complete backend without the complexity of managing separate services for auth, storage, and database. It serves as a self-hostable alternative to Firebase.

How it saves time or tokens

Appwrite eliminates backend infrastructure setup. Instead of configuring a database, auth provider, file storage, and serverless runtime separately, you run one Docker Compose command. For AI apps, this means you can focus on model integration and user experience while Appwrite handles user management, data persistence, and file uploads. The serverless functions let you run AI inference as backend operations triggered by database events or HTTP requests.

How to use

1. Start Appwrite with Docker:

docker run -it --rm \
  --volume /var/run/docker.sock:/var/run/docker.sock \
  --volume "$(pwd)"/appwrite:/usr/src/code/appwrite:rw \
  --entrypoint="install" \
  appwrite/appwrite:latest

1. Access the console at http://localhost and create a project.

1. Use the SDK in your app:

import { Client, Databases } from 'appwrite';

const client = new Client()
    .setEndpoint('http://localhost/v1')
    .setProject('your-project-id');

const databases = new Databases(client);

// Store AI conversation
await databases.createDocument(
    'ai-db', 'conversations',
    'unique()',
    { prompt: 'Explain Docker', response: 'Docker is...', model: 'gpt-4o' }
);

Example

from appwrite.client import Client
from appwrite.services.databases import Databases

client = Client()
client.set_endpoint('http://localhost/v1')
client.set_project('your-project-id')
client.set_key('your-api-key')

db = Databases(client)

# Create a collection for AI model outputs
db.create_collection(
    database_id='ai-db',
    collection_id='predictions',
    name='Model Predictions'
)

# Store a prediction result
db.create_document(
    database_id='ai-db',
    collection_id='predictions',
    document_id='unique()',
    data={
        'input': 'Sample input text',
        'output': 'Model prediction result',
        'confidence': 0.95,
        'model_version': 'v2.1'
    }
)

Related on TokRepo

• AI tools for databases -- Backend and database tools for AI
• AI tools for self-hosted -- Self-hostable infrastructure

Common pitfalls

• Appwrite runs multiple Docker containers. Ensure your server has sufficient resources (at least 2GB RAM recommended). Development laptops should allocate enough Docker memory.
• The serverless functions runtime has cold-start latency. For real-time AI inference, consider running your model server separately and using Appwrite for data management.
• Database queries have size limits. For large-scale analytics on AI outputs, export data to a dedicated analytics database rather than querying Appwrite directly.