Agentic AI Docker Stack

Agentic AI Docker Stack

Create a ready-to-run Docker Compose deployment of an Agentic AI Stack with Open WebUI, Ollama, NanoClaw. It will be used in a training with approximately 10 participants.

The deployment should require only .env configuration and a docker compose file, so that all I have to do is put in the configurations and run docker compose.

It is important for me that the freelancer takes care all the necessary testing and delivers a working solution.

My Server: 32 GB RAM, CPU-only, configurable with 1–16 CPUs, approximately 10 users. I will do the deployment on the server, your task only is to provide the project folder with .env file and docker-compose.yml.

Components:

Open WebUI

Ollama + Qwen3 14B, model configurable via .env, automatic initial download, persistent volume.
Open WebUI, multi-user, admin-managed accounts, private user chats/workspaces, publicly exposed port.
No reverse proxy/TLS configuration; Nginx Proxy Manager is handled separately.

NanoClaw with:
WhatsApp shared group
Microsoft Teams shared channel
Signal skill installed, no Signal group
Browser + web search/fetch
Shell/code execution and file access inside isolated containers
persistent memory
scheduler
MCP support
dashboard

All NanoClaw participants should use the same shared agent on WhatsApp/Teams. Open WebUI users have separate private workspaces.

Use Docker volumes for all persistent state. Ollama and internal services must not be publicly exposed.

LiteLLM: Decide whether it should be included. Prefer it if it provides a clean abstraction for switching between the local model and future external providers such as OpenAI.

The configuration should include an OpenAI API key, and it should easily be possible for me to switch between the local model and an OpenAI model.

If you disagree with any of the design decisions, feel free to suggest improvements.

Deliver a single project directory containing docker-compose.yml, .env.example, required configuration/scripts and a concise README.