Soliplex Docker Compose Template

A Docker Compose template that assembles a full Soliplex stack — a self-hosted RAG / AI system — from a backend, a Flutter web frontend, a document-ingest pipeline, and Postgres. This repository is configuration and Dockerfiles, not application source: the backend is installed from the published soliplex package and the frontend from a release build.

There are two ways to use it: run this repository directly as a reference stack to try Soliplex now, or generate your own tailored stack from it (for a real deployment). Both are below.

Try it

You need Docker (with the Compose plugin), uv, and an Ollama server the containers can reach.

git clone https://github.com/soliplex/soliplex-template.git
cd soliplex-template
uv run scripts/generate_secrets.py    # writes .secrets/*.gen (gitignored)
echo 'OLLAMA_BASE_URL=http://your-ollama-host:11434' > .env
docker compose up                     # first run builds images; takes a few minutes

OLLAMA_BASE_URL must point at an Ollama server that serves the models named in backend/environment/installation.yaml.

Once it is up, open the web frontend at http://localhost:9000. The terminal client (TUI) is bundled in the backend image, so you can run it against the stack without installing anything:

docker compose exec backend soliplex-tui --url http://localhost:8000

Generate your own stack

For a real deployment, scaffold a tailored project — your own project name, host ports, model choices, auth mode, and an installable Python package wired into the backend — instead of running this repository as-is. Use the bundled soliplex-template Agent Skill (ask an agent to generate a stack), or run the generator script directly:

uv run skills/soliplex-template/scripts/generate_soliplex_project.py \
    --out ../my-stack --interactive

See the "Generate a project" guide in the documentation for the full parameter list and the skill workflow.

What's in the stack

• nginx — serves the web frontend and reverse-proxies the API.
• backend — the Soliplex server (soliplex-cli serve).
• haiku-ingester — the document pipeline that feeds the RAG vector store.
• docling-serve — converts uploaded documents for ingestion.
• postgres — thread persistence, authorization, and the ingester job queue.

Two more services are optional when you generate a project (and both run in this reference stack): a web TUI (the terminal client served as a web app, proxied at /tui/) and a local Gitea (proxied at /gitea/).

Documentation

Full documentation — prerequisites, exposed ports, architecture, secrets, the RAG pipeline, configuration, and generating a customized project — lives at https://soliplex.github.io/soliplex-template/ (sources under docs/, built with Zensical).

Preview or build it locally:

uv run zensical serve     # preview at http://localhost:8000
uv run zensical build     # static site under site/