Skip to main content

Quickstart

This page takes you from a fresh install to your first PipelineResult in about five minutes using ragaxis.ldk, the launch development kit of pre-wired components.

1. Install RAG Axis​

uv pip install "ragaxis[ldk]"

See Installation for details on extras and supported Python versions.

2. Configure Adapters​

RAG Axis needs an LLMAdapter, an EmbedderAdapter, and a VectorStoreAdapter. The LDK provides reference implementations so you can get started without writing your own:

from ragaxis.ldk import default_pipeline

pipeline = default_pipeline(
    llm_model="gpt-4o-mini",
    embedding_model="text-embedding-3-small",
)

3. Ingest a Document​

from ragaxis.core import Document

doc = Document(
    id="doc-1",
    text="RAG Axis treats every failure mode as a named, typed result.",
    metadata={"source": "readme"},
)

pipeline.ingest([doc])

4. Run a Query​

result = pipeline.run("What does RAG Axis do with failure modes?")

print(result.answer)
print(result.cost_report)

result is a PipelineResult. It carries the generated answer, the retrieved chunks with provenance, and a CostReport with per-stage tokens, latency, and estimated cost.

Next Steps​