Examples¶

Every operator, decorator, and integration in this provider has a runnable Dag under example_dags. This page groups them by scenario. Guides embed the same Dags inline where a step-by-step walkthrough exists; the rest are linked directly to source below.

New to the provider? Start with Common AI Operators to pick an operator, then come back here for a worked example of the pattern you need.

Single-prompt tasks¶

Guide	What it shows
LLMOperator	`@task.llm` for text, structured output, classification, and dynamic task mapping over LLM results (example_llm.py, example_llm_classification.py, example_llm_analysis_pipeline.py).
LLMBranchOperator	`@task.llm_branch` picking which downstream task runs.
LLMFileAnalysisOperator & @task.llm_file_analysis	`@task.llm_file_analysis` reasoning over files, images, and PDFs.
LLMSchemaCompareOperator	`@task.llm_schema_compare` comparing two schemas with an LLM.
LLMSQLQueryOperator	`@task.llm_sql` generating SQL from a natural-language question.

Agents & tools¶

Guide	What it shows
AgentOperator & @task.agent	`AgentOperator` / `@task.agent` multi-turn tool use, durable execution, and pydantic-ai capabilities (example_agent.py, example_agent_durable.py, example_agent_capabilities.py).
Toolsets	Loading `SKILL.md` Agent Skills (example_agent_skills.py), allow-listed AWS API access via `AWSToolset` (example_aws_toolset.py), and exposing an Airflow toolset to a LangChain agent, the reverse bridge (example_langchain_toolset_bridge.py).
MCP Server Connection	Connecting an agent to an MCP server through an Airflow connection.
Human-in-the-Loop (HITL) Review For Agentic Operators	Adding a human-in-the-loop review gate to agent output.
example_langchain_tool_agent.py	A LangChain ReAct agent that decides its own tool calls, composed with `LLMOperator` for report formatting and AIP-90 HITL review.

Retrieval & document processing¶

Guide	What it shows
DocumentLoaderOperator	Parsing PDF, DOCX, CSV, and JSON into `list[dict]` for embedding.
PydanticAIHook	Calling `PydanticAIHook` and a pydantic-ai `Agent` directly.
LangChainHook	`LangChainHook` chat-only, embedding-only, and combined patterns.
LlamaIndexHook	`LlamaIndexHook` plus the embedding and retrieval operators.

End-to-end scenarios¶

Production-shaped Dags that combine several of the patterns above into one pipeline. See End-to-End Pipelines for the architecture behind each one.

Example	What it shows
example_llamaindex_rag.py	Three RAG shapes with LlamaIndex: a single load-embed-retrieve-answer Dag, a production-shaped split into scheduled indexing plus on-demand query Dags, and multi-source RAG.
example_llamaindex_10k.py, example_langchain_10k.py	SEC 10-K financial analysis against live EDGAR filings: one Dag indexes filings on a schedule, the other decomposes a comparison question at runtime and fans out retrieval with Dynamic Task Mapping. One variant per RAG library.
example_llm_survey_analysis.py	Natural-language querying of a CSV via `LLMSQLQueryOperator` and `AnalyticsOperator`, interactive and scheduled variants.
example_llm_survey_agentic.py	The same survey data, analyzed by fanning a multi-dimensional research question into an agentic multi-query synthesis pipeline.
example_aip_progress_tracker.py	The same use case – tracking Airflow Improvement Proposal progress – solved two ways, a deterministic pipeline and an autonomous agent, to compare the tradeoff.

Reliability¶

Guide	What it shows
LLM Retry Policies	Classifying task failures with an LLM to decide retry, fail, or delay. Source: example_llm_retry_policy.py.