Setting up Retrieval-Augmented Generation (RAG) infrastructure is a pain. Even if you know what you’re doing, provisioning vector databases, configuring retrieval pipelines, handling model integration, and optimizing everything for performance takes weeks. So, we decided to develop an AI agent to automate the entire process.
The agent takes a text prompt describing your setup requirements and builds the full RAG pipeline automatically. It provisions the right vector database, sets up retrieval logic, connects models (OpenAI, open-source LLMs, fine-tuned ones), and deploys everything — whether in the cloud or on-prem. It’s designed to support major frameworks and databases out of the box, with extensibility for custom integrations.
The goal is to eliminate DevOps overhead and make AI infrastructure deployable in minutes instead of months. This should help smaller teams and startups focus on their models and applications rather than infrastructure.
That’s what we thought. Now the question to the community is if it is relevant and required? Please let me know if such a tool would be useful for you.
Also, while we’re in early development we are looking for alpha testers — especially teams experimenting with LLMs, RAG pipelines, and vector databases. If you’re interested in trying it out (and breaking things), please comment (or register at donkit.ai). Your feedback will directly shape how this evolves.