groundcover Python SDK

Official Python SDK for the groundcover API.

Installation

1. Install uv (skip if you already have it)

curl -LsSf https://astral.sh/uv/install.sh | sh

2. Install the SDK

uv init   # skip if you already have a pyproject.toml
uv add "git+https://github.com/groundcover-com/groundcover-sdk-python.git"

Requirements

• Python 3.9+
• A groundcover API key and Backend ID

Quick Start

import groundcover
from groundcover.api.metrics import metrics_query
from groundcover.models.query_request import QueryRequest
from groundcover.models.query_request_query_type import QueryRequestQueryType
from datetime import datetime, timedelta, timezone

with groundcover.Client() as client:
    now = datetime.now(timezone.utc)
    result = metrics_query.sync_detailed(
        client=client,
        body=QueryRequest(
            promql="up",
            query_type=QueryRequestQueryType.INSTANT,
            start=now - timedelta(hours=1),
            end=now,
        ),
    )
    print(result.parsed)

The SDK provides typed API functions as the primary interface. Import the operation module and call sync_detailed() (or asyncio_detailed() for async):

from groundcover.api.logs import search_logs
from groundcover.models.logs_search_request import LogsSearchRequest

result = search_logs.sync_detailed(
    client=client,
    body=LogsSearchRequest(start=start, end=end, query="* | stats count(*)"),
)

Raw HTTP methods (client.get(), client.post(), etc.) remain available for advanced use cases or endpoints not yet available as typed functions.

Configuration

The SDK reads from environment variables by default:

Variable	Description	Required
GC_API_KEY	Your groundcover API key	Yes
GC_BACKEND_ID	Your groundcover Backend ID	Yes
GC_BASE_URL	API base URL (default: https://api.groundcover.com)	No
GC_TRACEPARENT	Default traceparent header for tracing	No

You can also pass configuration explicitly:

client = groundcover.Client(
    api_key="your-api-key",
    backend_id="your-backend-id",
    base_url="https://api.groundcover.com",
    timeout=60.0,
    retry_count=5,
)

Async Support

import groundcover
from groundcover.api.dashboards import get_dashboards

async with groundcover.AsyncClient() as client:
    result = await get_dashboards.asyncio_detailed(client=client)

Monitors (YAML)

Monitor endpoints use YAML content-type, so they have dedicated client helper methods:

with groundcover.Client() as client:
    # Create from YAML string
    client.create_monitor(yaml_string)

    # Get returns parsed YAML as dict
    monitor = client.get_monitor("monitor-id")

    # Update
    client.update_monitor("monitor-id", updated_yaml)

    # Delete uses the typed API
    from groundcover.api.monitors import delete_monitor
    delete_monitor.sync_detailed("monitor-id", client=client)

Workflow SDK removal

Breaking change (AI-461): The public Python SDK no longer exposes workflow create, list, or delete operations for POST /api/workflows/create, POST /api/workflows/list, and DELETE /api/workflows/{id}. This removes Client.create_workflow(), AsyncClient.create_workflow(), and the generated workflow list/delete operations. For temporary compatibility, the unchanged endpoints can be called directly through REST. Workflows are deprecated; migrate notification automation to Destinations and Notification Routes using the workflow migration guide.

Condition Builder

from groundcover.utils import ConditionSet

conditions = (
    ConditionSet()
    .add("namespace", "production")
    .add("workload", "api-server")
    .add_oom_event_conditions()
    .build()
)

response = client.post("/api/logs/v2/search", json={
    "query": "* | stats count(*)",
    "conditions": conditions,
    ...
})

Error Handling

All requests — both typed API calls and raw HTTP methods — raise the same typed exceptions:

import groundcover

try:
    result = get_dashboard.sync_detailed("nonexistent", client=client)
except groundcover.NotFoundError:
    print("Dashboard not found")
except groundcover.AuthenticationError:
    print("Invalid API key")
except groundcover.RateLimitError:
    print("Rate limited, SDK retries automatically on 429/503")
except groundcover.APIError as e:
    print(f"API error {e.status_code}: {e.body}")

Retry Behavior

By default, the SDK retries on HTTP 503 and 429 with exponential backoff and jitter:

• Default retries: 3
• Default retry statuses: 503 (Service Unavailable), 429 (Too Many Requests)
• Backoff: Exponential with jitter, 1s min, 30s max

Development

cd sdk-python
uv sync --extra dev

# Run unit tests
make test-unit

# Run linting
make lint

# Format code
make format

License

Apache 2.0 — see LICENSE.