Skip to content

About

Python SDK for the Arango AI Suite (pre-alpha) — AutoGraph end-to-end workflow through the platform gateway

Topics

Resources

Security policy

Stars

1 star

Watchers

0 watching

Forks

Repository files navigation

arango-ai-sdk

Work in Progress

Python SDK for the Arango AI Suite. It wraps the suite's wire contracts so a developer never handles endpoints, auth headers, JSON shapes, error codes, or polling loops directly.

The suite covers four services:

Service Role
ACP Authentication and project management (system-wide)
FileManager File upload and category/scope labeling (system-wide)
AutoGraph Ingestion: corpus build, strategizing, orchestration
Retriever (AutoRag) Querying the built ContextGraph

Status: work in progress. The version 1 surface is in the repository. Publishing to public PyPI is a later step. Licensed under the Apache License 2.0.

Install from this repository

Requires Python 3.12 or newer and uv.

git clone https://github.com/arangoml/arango-ai-sdk
cd arango-ai-sdk
make install

make install runs uv sync and creates a virtual environment in .venv. Without make, as on a default Windows setup, run uv sync directly. Activate the environment, then run any script that imports the SDK from that shell:

source .venv/bin/activate          # macOS and Linux
.venv\Scripts\activate             # Windows
import arango_ai_sdk

print(arango_ai_sdk.__version__)

Install from the private index

The package is also published to the Arango private PyPI (Google Artifact Registry). Public PyPI availability is a separate, later step.

Authentication to the registry uses Google Artifact Registry keyring support:

pip install keyring keyrings.google-artifactregistry-auth
gcloud auth login   # or supply a service-account key

With pip:

pip install arango-ai-sdk \
  --extra-index-url https://europe-west3-python.pkg.dev/arango-ml/arango-ml-internal-pypi/simple/

With uv, declare the index explicitly in the consuming project's pyproject.toml:

[[tool.uv.index]]
name = "arango_private"
url = "https://europe-west3-python.pkg.dev/arango-ml/arango-ml-internal-pypi/simple/"
explicit = true

[tool.uv.sources]
arango-ai-sdk = { index = "arango_private" }

Then:

import arango_ai_sdk

print(arango_ai_sdk.__version__)

Development

The project is managed with uv.

uv sync                  # create the environment
uv run pre-commit install  # enable the git hooks
make check               # lint + typecheck + dependency check + tests

Individual targets:

Command What it runs
make lint ruff check
make format ruff format
make typecheck ty check
make deptry dependency hygiene check
make test pytest
make build uv build (wheel + sdist)
make docs Sphinx user docs, see docs/README.md

Versioning and release channels

The version is derived from the git tag — it is not stored in any file, and there is nothing to bump by hand. Which channel a build belongs to is encoded in the version itself, and installers enforce the separation: pip and uv skip pre-releases unless asked. Assuming the latest tag is 0.1.0rc1:

Git state Version produced Channel Who installs it
Any untagged commit 0.1.0rc1.post1.dev3+g1a59664 dev us, automatically
Tagged v0.1.0rc1 0.1.0rc1 QA QA, by asking for it
Tagged v0.1.0 0.1.0 production everyone

Releasing is therefore a tag, not a merge.

git tag -m "QA candidate" v0.1.0rc1   # tags are signed in arangoml org, so -m is required
git push origin v0.1.0rc1             # CI publishes 0.1.0rc1

License

Apache License 2.0. See LICENSE.

About

Python SDK for the Arango AI Suite (pre-alpha) — AutoGraph end-to-end workflow through the platform gateway

Topics

Resources

Security policy

Stars

1 star

Watchers

0 watching

Forks

Releases

Packages

Used by

Contributors

Languages