ragnerock

Python SDK and CLI for the Ragnerock API.

• A SQLAlchemy-shaped Session API for documents, annotations, operators (agents), workflows, jobs, and SQL queries over annotations.
• GitOps apply — declare secrets, provider configs, schemas, skills, agents, workflows, endpoints, and ingest configs as YAML, and apply them from the CLI or CI.
• A low-level client covering the rest of the platform: agent tools and skills, notebooks and assignments, search, endpoints, ingest sources, saved queries, workflow schedules and change requests, and AI / database / blob provider configs.

pip install ragnerock
pip install 'ragnerock[pandas]'           # QueryResult.to_pandas()
pip install 'ragnerock[gcp,aws,azure]'    # cloud secret managers for GitOps Secrets

Python

from ragnerock import Document, Session, create_engine

engine = create_engine("ragnerock://token:rgnk_...@api.ragnerock.com/my_project")

with Session(engine) as session:
    doc = Document(file_path="./report.pdf", name="q1-report")
    session.add(doc)
    session.commit()

    result = session.query("SELECT vendor, total FROM invoice_extract WHERE total > 1000")
    df = result.to_pandas()

    # Anything the session doesn't model is on the low-level client
    hits = engine.client.search.search(project_id=engine.project_id, query="revenue guidance")

CLI

export RAGNEROCK_HOST=api.ragnerock.com
export RAGNEROCK_API_TOKEN=rgnk_...
export RAGNEROCK_PROJECT=my_project

ragnerock get docs                              # list documents
ragnerock apply -f manifests/ --plan            # preview a GitOps apply
ragnerock apply -f manifests/                   # apply it
ragnerock run ingest --documents q1.pdf --wait  # run a workflow
ragnerock query "SELECT COUNT(*) FROM invoice_extract"

Documentation

The user guide and API reference live in docs/source/ — start with Connecting, Sessions, GitOps manifests, and Platform APIs. Upgrading from 1.x? See Migrating to 2.0. Example manifests are in examples/gitops/.

Development

uv sync --all-packages
task test-unit     # unit tests
task lint          # ruff
task check         # pyright
task format        # ruff format
task doc           # sphinx (warnings are errors)

Licensed under Apache-2.0.