workflows-py/examples

LlamaAgents Examples

A collection of runnable examples showing how to build, serve, and deploy agent workflows with llama-index-workflows and llama-agents-*.

New to the project? Start at the top of the list and work down — each step builds on the previous.

Start here

1. feature_walkthrough.ipynb — The single best place to begin. A guided tour of workflows, steps, events, context, branching, loops, and streaming, all in one notebook.
2. agent.ipynb — Build a simple agent as a workflow. Covers tool calling and the agent loop pattern.
3. document_processing.ipynb — A realistic document pipeline: parsing, extraction, and orchestration.

Serving workflows as an API

4. server/ — Wrap a workflow as a REST API with WorkflowServer, standalone or mounted inside an existing FastAPI app.
5. client/ — Call a running workflow server from Python with WorkflowClient, including streaming and human-in-the-loop.

Durability and scale

6. durable_workflows.ipynb — Save and resume workflow runs using pluggable storage.
7. dbos/ — Production-grade durability with DBOS: crash recovery, multi-replica servers, and idle release.

Deployment

8. docker/ — Containerize a workflow server with Docker.
9. k8s-otel/ — Deploy to Kubernetes with OpenTelemetry, Tilt, and a kind cluster.

Observability and evaluation

10. observability/ — Trace workflows with Arize Phoenix, Langfuse, and the built-in context logger.
11. eval_driven_prompt_refinement.ipynb — Iterate on prompts using evaluation-driven feedback loops.

Advanced patterns

• streaming_internal_events.ipynb — Stream intermediate events from nested workflow steps.
• state_management_with_vector_databases.ipynb — Persist workflow state in a vector database.
• document_agents/ — A finance triage agent built with document workflows.
• visualization/ — Visualize workflow graphs, including resource nodes.

---

For more on the library, see the llama-index-workflows package README and the project root README.