Skip to content

Latest commit

 

History

History
188 lines (141 loc) · 7.52 KB

README.md

File metadata and controls

188 lines (141 loc) · 7.52 KB

Zeroshot Python SDK

Install from PyPI with Python 3.11 or newer:

python -m pip install the-open-engine-zeroshot

Import the package as zeroshot. Pip selects the wheel for your OS and CPU. See the Python SDK guide for setup and examples.

Zeroshot enforces your team's QA process on coding agents as a graph you define. zeroshot is a typed async client for its run engine. Each platform wheel contains the matching zeroshot executable. Python supplies async calls and typed result objects; the executable remains authoritative for graph validation, run state, template expansion, and runtime/provider rules.

import asyncio

from zeroshot import Client, UniformRuntime

async def main() -> None:
    runtime = UniformRuntime(
        harness="codex",
        provider="openrouter",
        model="PROVIDER_MODEL_ID",
        effort="max",
    )

    async with Client(runtime=runtime) as client:
        result = await client.run(
            "Implement the requested change and run the relevant tests.",
            wait_timeout=21_600,
        )

    result.raise_for_failure()


if __name__ == "__main__":
    asyncio.run(main())

Client() defaults to LocalTarget() and Preset("software-change"), but it never guesses a harness, provider, or model. run() submits one durable graph run and waits for its terminal result. submit() returns a Run immediately for status, resumable watch/log streams, waiting, or durable force-stop control.

Local, direct, and hosted targets

Local runs use the folder captured when the client opens. Set LocalTarget.workspace or per-run cwd= to select another folder; relative cwd paths use that captured workspace. Agents write in place. Without Git delivery, plain folders work; push needs committed Git and a remote, while PR and merge need a supported forge (GitHub today).

from zeroshot import Client, UniformRuntime


async def run_local(runtime: UniformRuntime) -> None:
    async with Client(runtime=runtime) as client:
        result = await client.run("Update the parser.", cwd="/path/to/folder")

    result.raise_for_failure()

The same client reaches an unauthenticated Docker deployment or another direct target by changing only the target:

from zeroshot import Client, DirectTarget, UniformRuntime


async def run_direct(runtime: UniformRuntime) -> None:
    target = DirectTarget("http://127.0.0.1:8080")

    async with Client(target=target, runtime=runtime) as client:
        result = await client.run("Inspect the repository and report success.")

    result.raise_for_failure()

Docker/direct and hosted targets require a committed Git source. Per-run repository= accepts a credential-free HTTPS/SSH Git URL or GitHub owner/name; branch= and revision= override source selection. source_connection= selects a named connection used only for checkout and delivery.

HostedTarget reuses a named target and login from the CLI. Merge plans are immutable DAGs over one repository, branch, and hosted profile. That profile must contain exactly one builtin.git-delivery.merge@2 node; pull-request delivery isn't accepted for plans. Agent bindings must not declare GH_TOKEN; only the Git delivery binding may declare it.

from datetime import UTC, datetime, timedelta

from zeroshot import Client, HostedTarget, MergePlanRequest, MergePlanRun


async def run_plan() -> None:
    request = MergePlanRequest(
        title="Release checkout update",
        repository="the-open-engine/zeroshot",
        branch="main",
        profile="org:software-change",
        expires_at=(datetime.now(UTC) + timedelta(days=1)).isoformat(),
        runs={
            "backend": MergePlanRun(input={"task": "Update the API."}),
            "frontend": MergePlanRun(input={"task": "Update the client."}),
            "integrate": MergePlanRun(
                input={"task": "Verify the combined changes and run the release tests."},
                needs=("backend", "frontend"),
            ),
        },
        submission_key="checkout-release-1",
    )

    async with Client(target=HostedTarget("platform")) as client:
        plan = await client.submit_plan(request)
        status = await plan.wait(wait_timeout=21_600)

    if not status.succeeded:
        raise RuntimeError(f"merge plan ended as {status.state}")

Plan input is static JSON. A dependency controls when a node may start; it doesn't copy output into another node. Run IDs are assigned once during atomic submission. After dependencies merge, Platform materializes a node against an exact source revision. When materialization completes, readyAt is set and the node's queue deadline is the earlier of plan expiry and seven days after readyAt. A merge conflict is repaired by the conflicting run itself in the same checkout and delivery loop. A descendant cannot repair a failed predecessor; it ends as dependency_failed. MergePlan.watch() polls the aggregate Platform status and emits changed snapshots because the v1 plan API has no streaming endpoint.

Exact graph and runtime control

Use GraphSpec and RuntimePlan to pass exact JSON values unchanged. Python does not traverse or validate either document:

from collections.abc import Mapping
from typing import Any

from zeroshot import GraphSpec, RunRequest, RuntimePlan


def exact_request(
    graph_document: Mapping[str, Any],
    runtime_document: Mapping[str, Any],
) -> RunRequest:
    return RunRequest(
        title="Repair checkout",
        graph=GraphSpec.from_dict(graph_document),
        initial_input={"ticket": "OE-123"},
        runtime=RuntimePlan.from_dict(runtime_document),
        submission_key="oe-123-attempt-1",
    )

await client.list_presets() and await client.get_preset(...) query the bundled Rust catalog. The Python package contains no copied preset registry, provider/model registry, graph schema, or semantic validator. Every submission runs the executable's preflight before a local controller starts or a direct target is contacted.

Durable observation

Run.watch(after=...) and Run.logs(after=..., execution=...) resume strictly after opaque native cursors. A wait_timeout expiry raises RunWaitTimeout carrying the still-active Run; cancelling the Python await also detaches without stopping it. Run.force_stop() is the only operation that changes run lifetime. Each LogEvent.timestamp is the positive, JavaScript-safe Unix epoch millisecond captured at the producer and preserved unchanged by durable replay.

Environment values

With environment=None, submission values come from os.environ. An explicit mapping replaces ambient credential variables; native saved connections and local harness login still apply. Zeroshot selects declared runtime and source-connection fields; transport credentials stay outside agent environments unless independently declared by a node. The executable owns provider defaults.

Packaging and versions

The package is released only as platform wheels because each wheel bundles one matching native executable; there is no first-run download, Node dependency, PyO3 ABI, or auth dependency. The distribution name is the-open-engine-zeroshot, the import package remains zeroshot, and py.typed is included.

An SDK release tag is zeroshot-python-vZEROSHOT_SDK, for example zeroshot-python-v8.0.0_1; its PEP 440 package version is 8.0.0.post1. Revision 1 is released automatically after the corresponding Zeroshot release. Later SDK-only revisions may be released independently against the same immutable Zeroshot release.