Example Graphs¶
The demo API registers the following graphs. They demonstrate LGOS features;
none is installed as a built-in model by the langgraph-openai-serve package.
Each registration also supplies a short GraphConfig.description used by the
demo model catalogs.
| Model | Demonstrates | Graph feature | Extra runtime requirement |
|---|---|---|---|
custom-input-output-context |
Request, output, and typed runtime-context adapters | None | None |
citation-events |
Structured OpenAI URL citations with portable Markdown content | None | None |
advanced-mcp-tools |
Async graph factories and a mock MCP-style tool | None | None |
complex-subgraphs |
Router-selected subgraphs and nested streamed output | None | None |
status-events |
Portable status updates for native client UI | client_events |
None |
custom-event-showcase |
Public progress and artifact events interleaved with text | client_events |
None |
interruptible-approval |
Checkpointed human approval represented as an OpenAI tool call | interrupts |
PostgreSQL |
simple-graph |
Streamed model output and discoverable runtime settings | None | Upstream chat model |
lgos-rag |
Agentic retrieval over the packaged demo corpus | None | Upstream chat and embedding models |
The demo API opens its PostgreSQL checkpointer during application startup, so PostgreSQL must be available even when you call a provider-free graph. Start it with the demo API instructions.
Start without provider credentials
Use custom-input-output-context, citation-events,
advanced-mcp-tools, complex-subgraphs, status-events,
custom-event-showcase, or interruptible-approval to explore the
transport without a real model API key.
Source Map¶
All graph code is owned by the independent demo/api project:
demo/api/src/lgos_demo_api/app.pyregisters graph names as OpenAI model names.demo/api/src/lgos_demo_api/graphs/simple.pypublishes safe runtime settings for conversation history and intended audience.demo/api/src/lgos_demo_api/graphs/lgos_rag.pyimplements agentic retrieval, relevance grading, bounded rewriting, and grounded streamed answers.demo/api/src/lgos_demo_api/corpus/contains the Markdown embedded in source installs, wheels, and API images.demo/api/src/lgos_demo_api/graphs/custom_io.pycontains input, output, and context adapters.demo/api/src/lgos_demo_api/graphs/advanced_mcp.pyconstructs an agent from an async factory and mock tool.demo/api/src/lgos_demo_api/graphs/complex_subgraphs.pyandgraphs/subgraphs/implement router-selected specialists.demo/api/src/lgos_demo_api/graphs/status_events.pyemits portable status updates.demo/api/src/lgos_demo_api/graphs/custom_events.pyemits explicitly public progress and artifact events.demo/api/src/lgos_demo_api/graphs/interruptible.pypauses and resumes a checkpointed approval flow.demo/api/src/lgos_demo_api/graphs/citations.pyemits citation events that LGOS maps to OpenAI annotations.
Continue with Run the Demo API for request examples or Chainlit and Open WebUI for UI behavior.