AI stack¶
Local AI for your organisation: a GPU model server, one gateway every client uses, and a chat front end — no prompt leaves your site. How the parts fit: DESIGN.md. On the website: AI Stack.
Planned: n8n — AI-empowered workflow automation, through the LiteLLM gateway (a placeholder directory, no module yet).
| Module | What it is | Status |
|---|---|---|
| vllm-amd | vLLM AMD ROCm inference — Ryzen AI MAX+ 395 (tappaas2) | Development |
| litellm | LiteLLM AI gateway — one OpenAI-compatible API, with per-user keys and usage tracking, in front of local (vLLM) and remote models. | Development |
| openwebui | Open WebUI chat interface — private AI chat for your users, with conversation history and RAG, through the LiteLLM gateway. | Development |
Install them as INSTALL.md says.
Architecture view¶
How modules are grouped into stacks: ADR-004, Where a module's directory is. The module lists here and in INSTALL.md are generated by src/foundation/tappaas-cicd/scripts/gen-stack-readme.py from src/module-catalog.json.