Skip to main content

OpenStateGraph

Who this page is for: you are adding OpenStateGraph to a project you already have, from a package index. This is the distribution's own page; there is no checkout in it.

A compiler from a vendor-neutral workflow.json to a LangGraph StateGraph. Not an execution engine — the thing it produces is a plain compiled LangGraph object that runs, streams, checkpoints and deploys anywhere Python runs, with or without this package's editor.

pip install "openstategraph[ollama]"

That is the line that works today, and it names no index because it needs none: this is on PyPI. Nothing to pin — 0.3.0 is the first final release, so an unpinned install resolves it. ../docs/releasing.md is the page that says how a version gets here.

You can install the identical artifact from a checkout — pip install -e "backend[ollama]" from the repository root, or build the wheel with python3 -m build backend — which is what CI's clean-install job does, into an empty virtualenv outside the repository.

from openstategraph import load_workflow

workflow = load_workflow("path/to/my-workflow")
print(workflow.ask("How many invoices are there?"))

load_workflow takes the one thing an adopter actually has — the package folder holding workflow.json — and wires the package's own tools/, functions/, middlewares/, skills/ and knowledge/ before compiling. Anything it could not resolve lands on .warnings and logs a WARNING, because the alternative failure mode is a workflow that answers confidently without the tools it was drawn with.

workflow.graph is the escape hatch: the compiled LangGraph object, with nothing of ours in the way.

Install footprint

The core is four dependencies — langgraph, langchain, langchain-core, pydantic. Never "four packages": package is a settled word here for workflows/<slug>/ (CLAUDE.md's lexicon), and backend/pyproject.toml names this line as the place the other sense kept being copied to. Everything else is an extra, because a consumer of load_workflow should not install a web server or three provider SDKs to run a graph in their own process.

Extra Install it for
[anthropic] / [openai] / [ollama] a model string starting anthropic: / openai: / ollama:
[deep] a document containing an agent.deep node
[sqlite] settings.checkpointer: "sqlite", or OPENSTATEGRAPH_MEMORY_PATH
[server] the editor's HTTP API (openstategraph.api.main) — the web layer only, so pair it with a provider extra ([server,ollama]) or the editor opens onto workflows it cannot run. Includes [mcp], so openstategraph mcp works after this extra alone
[mcp] the MCP transport (openstategraph.mcp_server) on its own, with no web server
[all] everything above — what a checkout of this repo wants
[dev] contributors (pytest, ruff)

Each of those is imported lazily at its one call site, and a missing one raises an ImportError naming the exact pip install line rather than degrading quietly.

Public surface and stability

openstategraph.__all__ and openstategraph.abc.__all__ are the semver-public surface, plus openstategraph.errors and the workflow.json schema itself. Everything under openstategraph.api.* and openstategraph.mcp_server is internal and carries no stability guarantee. See docs/stability.md for the deprecation policy — the short version is that pre-1.0, a breaking change bumps the minor, never the patch.

MIT licensed. Full documentation lives in the repository's docs/.

Metadata

Release files for openstategraph 0.3.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for openstategraph 0.3.0
File Size Uploaded Provenance Actions
openstategraph-0.3.0.tar.gz 6.6 MB Without provenance Details for openstategraph-0.3.0.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for openstategraph 0.3.0
File Interpreter ABI Platform Actions
openstategraph-0.3.0-py3-none-any.whl Python 3 none any Details for openstategraph-0.3.0-py3-none-any.whl

Total release size: 11.5 MB

Release files / openstategraph-0.3.0.tar.gz

Download URL openstategraph-0.3.0.tar.gz
Size 6.6 MB
Tags Source
SHA-256 checksum
How to use checksums
8bf05e180aebb344eaec264b00e4e51cf0c06bf34617c21b8373202017072c43
BLAKE2b-256 checksum
How to use checksums
7fbb0e8119d59ef2867a107c4e582349d1b4fe5bac59f779404c6e6d265dac2b
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.13.9

Release files / openstategraph-0.3.0-py3-none-any.whl

Download URL openstategraph-0.3.0-py3-none-any.whl
Size 4.9 MB
Tags Python 3
SHA-256 checksum
How to use checksums
bc6638049c5186433416cdba8530e5f2c7a7617f6db9f61cc5d22b39c22b8b68
BLAKE2b-256 checksum
How to use checksums
6942eafb9915ce2029a9d3a59edbe7e06694feed583678f57192a10acf378f2a
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/7.0.0 CPython/3.13.9

Security for openstategraph 0.3.0

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page