kawkav
Sign in
Kawkav / AI / Stacks

AI · one-click stacks

Production stacks, deployed in 60 seconds.

Pre-baked one-click templates — runtime, database, SSL and backups already wired. Drop an AI stack straight onto your own Kawkav compute: run your own LLM gateway, vector DB and RAG pipeline on infrastructure you control. The stack is free — you only pay for the server under it.

24
Ready stacks
5
Categories
~60s
To live
EU
GDPR by default
AI stacks

Run the whole AI toolchain on your own tenant.

LLM gateways, inference servers, vector databases and low-code pipelines — pre-wired to ship a private AI app fast on your own data.

🦙

Ollama + Open WebUI

Self-hosted LLM gateway. Run Llama, Mistral and Qwen on your tenant behind a polished chat UI.

ollamaopen-webuipostgresnginx
4 vCPU · 16 GB · GPU optional

vLLM Inference Server

High-throughput LLM serving. OpenAI-compatible API, continuous batching and paged attention.

vllmhuggingfaceprometheus
8 vCPU · 32 GB · GPU required
🔗

LangChain + ChromaDB

RAG-ready Python stack. Build AI apps with retrieval, agents and a bundled vector DB.

python 3.12langchainchromadbredis
2 vCPU · 8 GB
🌊

Flowise + Vector DB

Drag-and-drop LLM apps. Build chatbots, agents and RAG flows without writing code.

flowiseqdrantpostgres
2 vCPU · 8 GB
🤖

n8n + LLM Workflows

Visual automation with built-in LLM nodes. Connect 400+ apps to your own AI.

n8npostgresollama node
2 vCPU · 4 GB
🧭

Qdrant Vector DB

Production-grade vector search. HNSW indexing, payload filtering, gRPC + REST.

qdrantgrpcprometheus
2 vCPU · 8 GB
🎙

Whisper Transcription

Self-hosted transcription API. Multilingual, with batch and streaming endpoints.

faster-whisperfastapiredis
4 vCPU · 16 GB · GPU recommended
🎨

ComfyUI

Node-based image generation. Stable Diffusion, FLUX, ControlNet and custom workflows.

comfyuisd-xlflux.1cuda 12
8 vCPU · 24 GB · GPU required
Web & app

Full-stack frameworks, ready on NVMe.

The runtime, database and reverse proxy configured and tuned — point at a domain and ship.

Next.js + Postgres

Full-stack React with App Router, Postgres, Prisma and pgvector ready.

node 22postgres 16pgvectornginx
1 vCPU · 2 GB
🚀

FastAPI + Redis

Async Python API with Redis cache, queue and pub/sub, behind production gunicorn.

python 3.12fastapiredis 7postgres
1 vCPU · 2 GB
📊

Supabase Self-Host

Open-source Firebase. Postgres, auth, storage, realtime and edge functions.

postgres 15postgrestgotruerealtime
2 vCPU · 4 GB
🌐

WordPress · NVMe

Tuned WordPress on NVMe. PHP 8.3, OPcache, Redis object cache and auto-SSL.

wordpressphp 8.3mariadbredis
1 vCPU · 2 GB
📝

Strapi Headless

Headless CMS with REST + GraphQL APIs, role-based access and media library.

node 22strapipostgres
1 vCPU · 2 GB
👻

Ghost CMS

Headless publishing platform. Newsletters, memberships, SQLite or MySQL.

node 22ghostmysqlnginx
1 vCPU · 1 GB
DevTools & data

The plumbing your team runs on.

Git, CI/CD, container management, databases and object storage — self-hosted on owned metal.

📦

Docker + Portainer

Docker Engine with the Portainer UI. Manage containers, stacks and registries.

dockerportainerwatchtower
1 vCPU · 2 GB
🦊

GitLab CE

Self-hosted Git with CI/CD pipelines, container registry and issue tracking.

gitlab-cegitlab-runnerpostgres
4 vCPU · 8 GB

Coolify

Self-hosted Heroku alternative. Deploy any app from Git with zero config.

coolifydockertraefik
2 vCPU · 4 GB
🐘

PostgreSQL + pgAdmin

Tuned Postgres 16 with pgvector, pgAdmin UI and daily backups to object store.

postgres 16pgvectorpgadminpgbackrest
2 vCPU · 4 GB
🗄

MinIO Object Store

S3-compatible object storage with versioning, replication and lifecycle policies.

minioconsoletls
2 vCPU · 4 GB
🍵

Gitea + Drone

Lightweight Git hosting with CI runners — a lower footprint than GitLab.

giteadronepostgres
1 vCPU · 2 GB
Productivity

Your own cloud, your own data.

Team chat, files, wikis, analytics and secrets — self-hosted, GDPR-safe, under your control.

Nextcloud

Files, calendar, contacts and mail — your own cloud, your own data.

nextcloudpostgresrediscollabora
2 vCPU · 4 GB
💬

Mattermost

Self-hosted team chat. Channels, threads, file sharing and integrations.

mattermostpostgresnginx
2 vCPU · 4 GB
📚

Outline Wiki

Beautiful team knowledge base. Markdown, slash commands and RBAC.

outlinepostgresrediss3
1 vCPU · 2 GB
🔐

Vaultwarden

Self-hosted password manager. Bitwarden-compatible, lightweight Rust build.

vaultwardensqlitesmtp
1 vCPU · 1 GB
📈

Plausible Analytics

Privacy-friendly site analytics. GDPR-safe, no cookies, lightweight script.

plausiblepostgresclickhouse
1 vCPU · 2 GB
📧

Mailcow Mail Server

Full mail stack: SMTP, IMAP, webmail, antispam and DKIM/DMARC out of the box.

postfixdovecotsogorspamd
2 vCPU · 4 GB

Don't see what you need? Custom stack requests are welcome — we build, test and add it to the marketplace within a week. Request a stack →

Questions

The honest answers.

What does a stack cost?

The stack itself is free — you only pay for the Kawkav compute it runs on. Pick a VPS size at deploy and resize any time; the template lays everything else down for you.

What's actually included?

Runtime, database, reverse proxy, SSL and backups — pre-wired and tuned. Each card lists the exact components and a minimum server size so you know what you're getting.

Where does it run?

On your own Kawkav tenant, on owned metal in EU regions with GDPR by default. Your data and your models stay on infrastructure you control — no third-party endpoint.

Can I customise after deploy?

Fully. You get root on the server, so add packages, swap models, tune configs and script it however you like — the stack is a starting point, not a cage.

How do the AI stacks connect to my models?

Point Open WebUI, vLLM or a RAG pipeline at your own Sovereign Models endpoint, or run the weights locally on the box. Either way inference stays inside your tenancy.

Ready when you are

Pick a stack. Point at a domain. Ship.

Two dozen production stacks, one click each — the AI toolchain, web frameworks and the plumbing your team runs on, live in about a minute on your own compute.

Pairs well with