VegaDūtaPlatform · Open source
Open source, and yours to run
VegaDūta is open source under the Apache-2.0 license. You can use the hosted service at vegaduta.ai, or run the platform yourself with the Docker Compose stack in the repository's deploy directory.
The same code runs in both places. Self-hosting means your conversations, documents, and credentials stay on infrastructure you control.
By the VegaDūta team · Last updated
In short
- VegaDūta is open source under the Apache-2.0 license.
- The supported way to self-host VegaDūta is the Docker Compose stack in the repository's deploy directory.
- Kubernetes charts in the repository are drafts and are not a supported install path yet.
What the Compose stack runs
The stack brings up the platform's services together, including the core API, the web console, the knowledge sidecar that handles document extraction and embeddings, Postgres, Redis, and Keycloak for sign-in. Configuration is by environment file.
Models: your keys or your hardware
A self-hosted install uses the same 14 bring-your-own-key providers as the hosted service. One of them is Ollama, which points the platform at models running on your own hardware, so an install can work without sending prompts to a hosted model API.
What is not supported yet
Docker Compose is the install path we support. The repository also contains Helm and Terraform directories; those are drafts and should not be treated as a supported Kubernetes deployment. An optional Temporal-based workflow engine is in the code and is off by default.
Frequently asked questions
Is VegaDūta open source?
Yes. The platform is licensed under Apache-2.0.
How do I self-host VegaDūta?
With Docker Compose, using the stack in the repository's deploy directory. It runs the core API, web console, knowledge sidecar, Postgres, Redis, and Keycloak, along with supporting services.
Can I run VegaDūta on Kubernetes?
Not as a supported path today. Helm and Terraform files exist in the repository as drafts. Docker Compose is the supported way to self-host.
Can a self-hosted install use local models?
Yes. The Ollama connector points the platform at a model server on your own hardware, alongside the other bring-your-own-key providers.
Sources
See it working in two minutes
The sandbox provisions a real tenant — describe an agent in one sentence and test it, no account, no card. Or browse ~90 industry workflow recipes to see what teams build.
