RAD Platform includes a complete self-hosted generative-AI stack you can
deploy into your own Google Cloud project — model serving, chat UIs, agent
builders, RAG pipelines, vector databases, gateways, and automation. This
page groups those modules so you can assemble a working LLM stack
layer by layer.
Every app below follows the same pattern as the rest of the platform: a
lab (guided deploy → verify → operate → tear down) and a module
reference (configuration guide), on Cloud Run, GKE Autopilot, or both.
Model serving
| App | What it does | Lab | Module |
|---|
| Ollama | Run open-weight LLMs (Llama, Mistral, Gemma) behind an API | Lab | Module |
| LiteLLM | One OpenAI-compatible gateway in front of many model providers | Lab | Module |
Chat and assistant UIs
| App | What it does | Lab | Module |
|---|
| Open WebUI | Self-hosted chat UI for local and remote models | Lab | Module |
| LibreChat | Multi-provider chat with agents and search | Lab | Module |
| AnythingLLM | Document chat workspaces over your own data | Lab | Module |
Agent and workflow builders
| App | What it does | Lab | Module |
|---|
| Flowise | Visual builder for LLM flows and agents | Lab | Module |
| Dify | LLM app platform: assistants, workflows, knowledge bases | Lab | Module |
| OpenClaw | Multi-tenant gateway for isolated, persistent AI assistants | Lab | Module |
| n8n AI | AI-augmented automation workflows | Lab | Module |
RAG pipelines and data
| App | What it does | Lab | Module |
|---|
| RAGFlow | End-to-end retrieval-augmented generation over documents | Lab | Module |
| Qdrant | Vector database for embeddings | Lab | Module |
| Chroma | Lightweight embedding store for RAG prototypes | Lab | Module |
| Crawl4AI | LLM-friendly web crawling for ingestion pipelines | Lab | Module |
| SearXNG | Private metasearch, a common RAG/agent search backend | Lab | Module |
Suggested stack order
- Serve a model — deploy Ollama, verify the API.
- Put a UI on it — connect Open WebUI.
- Add retrieval — stand up Qdrant and ingest with
Crawl4AI or RAGFlow.
- Build an agent — wire it together in Flowise
or Dify, fronted by LiteLLM.
Deploying and operating these services exercises the same skills covered in
the certification study paths —
GKE Autopilot and Cloud Run operations, networking, IAM, and observability —
on infrastructure you assembled yourself.