Local AI stack (Docker, Ollama) that lets you build apps without Python

Name: Local AI stack (Docker, Ollama) that lets you build apps without Python
Availability: InStock
Author: aeberhart

by aeberhart·Mar 9, 2026·1 point·0 comments

Visit Project View on HN

AI Analysis

●MidShip It

Low-code AI platform wrapped in Docker, but Open WebUI already does local AI simpler.

Strengths

•JSONata and SQL for custom tools avoids Python dependency entirely
•MCP proxy integration enables web search, file access, and O365 tools
•Role-based access control for AI apps is genuinely useful for teams

Weaknesses

•PolyForm license isn't OSI open source, limits community adoption
•Zero stars and forks suggests this is commercial product marketing

Post Description

We built a local-first AI stack (Docker + Ollama) that lets you build LLM tools and workflows without writing Python. It supports:

* Multimodal chat * Retrieval Augmented Generation (incl. automated & scheduled doc import) * MCP tool support (web search, file access, O365) * Custom tools implemented using JSONata & SQL

Most local AI tools are chat UIs. We wanted something closer to a programmable AI platform that still runs locally.

* leverage a low code platform to insert programmatic hooks in the chat * provide custom tools written in SQL or JSONata * embed LLMs in workflows and custom UIs * give fine grained role based access control to AI apps (who can see what and which tools is the LLM allowed to use)

To summarize, it aims at being as flexible as custom Python code while being as easy to use as Open WebUI.