This is the nginA API backend based on
- Python
- FastAPI/Pydantic
More about the origins of ngina: background story on LinkedIn
nginA offers a friendly self-service user interface to customers for
- Asynchronous AI applications (Long running workflows)
- Synchronous AI applications (Chatbots or Taskbots)
nginA is most useful in scenarios where you need to combine private data with AI applications and/or
want to give customers UI access for large amounts of AI generated content (PDFs, images, ...).
nginA is a wrapper around n8n to execute agentic workflows. It was written to enable self-hosted workflows which are compliant to EU regulations (e. g. hostend on EU servers).
All data is stored in a Supabase instance:
- nginA objects
- n8n schema tables
- Langchain memory
- ...
This setup simplifies backup/restore, because it automatically keeps the relations between n8n workflows and nginA workflow pointers in sync.
nginA is an API-centric product: though it at the core a automation of automations, it can be automated itself by using API calls.
When the API service powers up, it shows the API documentation page:
This API provides the following main building blocks:
Supabase provides the authentication subsystem which is extended by several nginA API methods.
To be able to distinguish between customers and you (the developer) we must use RBAC which adds some complexity to our setup (for security's sake).
The necessary steps are described here: nginA RBAC
Supabase provides the secure credential store via the PostgreSQL vault extension. nginA provides API methods for credential management - these are built upon the Supabase functionality
Recommended for trials and MVPs.
Subscribe to a Supabase managed service and create your PostgresInstance on an EU server.
Install the
Subscribe to render.com and pull the latest n8n Docker image. Set the environment variables to point to the Supabase instance:
Please enter "n8n" as schema name (instead of public).
Make sure you use the pooled connection settings from the Supabase connection list!
Clone the current public objects from replit.com and enter the secrets for both.
Recommended for production loads: Azure cloud.
Deploy Supabase using Docker to the Azure cloud. Attention: Docker compose installs several images!
Deploy n8n using the single Docker image to AKS.
Deploy the React frontend to a static website app.
Create a Docker image and deploy the Docker image to AKS.
The API features a streaming diagnostics endpoint to validate the current installation at any time (also in production): GET /api/v1/diagnostics
Just lean back and watch the tests scrolling by - for advanced automation this endpoint returns http status 200 if all tests passed and 500 if there were errors.