A collection of hands-on demos for OGX, covering everything from basic chat completions to RAG, agents, observability, and end-to-end applications. All demos run locally in your terminal.
uv syncThese demos use Ollama as the default model backend, but OGX supports many others (vLLM, Together, Fireworks, OpenAI, etc.). See the OGX documentation for the full list of supported providers.
# Install Ollama from https://ollama.com/download, then:
ollama pull llama3.2:3bStart the server in a separate terminal:
uvx --from 'ogx[starter]' ogx run starterThe server listens on localhost:8321 by default.
# Example: basic client setup
uv run python -m demos.01_foundations.01_client_setup localhost 8321All demos live under the demos/ directory and can be run directly from the terminal.
| Directory | Topic | Description |
|---|---|---|
01_foundations |
Foundations | Client setup, chat completions, vector DBs, tool registration, MCP |
02_responses_basics |
Responses API | Higher-level response generation with tools and structured outputs |
03_rag |
RAG | Retrieval-augmented generation with vector stores and search |
04_agents |
Agents | Conversational agents with chat, tools, RAG, and multi-agent coordination |
05_observability |
Observability | OpenTelemetry, Jaeger, Prometheus, and Grafana integration |
06_openai_compatibility |
OpenAI Compatibility | Using OpenAI-compatible endpoints with OGX |
07-end-to-end-apps |
End-to-End Apps | Full applications (e.g. Knowledge Assistant) |
Each demo directory has its own README with detailed instructions and learning objectives. Start with 01_foundations if you are new to OGX.
If you want to deploy OGX on Kubernetes instead of running locally, see deployment/kubernetes/ for manifests and instructions covering vLLM, the OGX Kubernetes operator, and MCP servers on a Kind cluster.