> ## Documentation Index
> Fetch the complete documentation index at: https://docs.costgraph.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Python

> Python SDK examples for the CostGraph Pricing API

## Install

The Python client is not on PyPI. Install it from the public repository,:

```bash theme={null}
pip install git+https://github.com/baselinehq/pricingapi-client-python.git
```

## Client setup

```python theme={null}
import os

import pricing_api_client
from pricing_api_client.api.pricing_api import PricingApi
from pricing_api_client.api.marketplace_api import MarketplaceApi
from pricing_api_client.models.instance import Instance
from pricing_api_client.models.github_com_baselinehq_golang_shared_types_vm import GithubComBaselinehqGolangSharedTypesVM

configuration = pricing_api_client.Configuration(
    host="https://pricing.baselinehq.cloud",
    api_key={"ApiKeyAuth": os.environ["COSTGRAPH_API_KEY"]},
)

with pricing_api_client.ApiClient(configuration) as api_client:
    pricing = PricingApi(api_client)
    marketplace = MarketplaceApi(api_client)
```

## Compute pricing

```python theme={null}
instance = Instance(
    region="nyc1",
    instance_type="s-2vcpu-2gb",
    usage_type="ONDEMAND",
    provider="DigitalOcean",
    operating_system="linux",
    service="Droplet",
    vm=GithubComBaselinehqGolangSharedTypesVM(cpu_cores=2, ram_gb=2),
)

price = pricing.v1_pricing_compute_post(instance=instance)
print(price.cost_per_hour)
```

## Recommendations

```python theme={null}
from pricing_api_client.api.recommendations_api import RecommendationsApi

instance = Instance(
    region="nyc1",
    instance_type="s-4vcpu-16gb-amd",
    usage_type="ONDEMAND",
    provider="DigitalOcean",
    operating_system="linux",
    service="Droplet",
    vm=GithubComBaselinehqGolangSharedTypesVM(cpu_cores=4, ram_gb=16),
)

recommendations = RecommendationsApi(api_client)
recs = recommendations.v1_recommendations_compute_post(instance=instance)
```

## Disk pricing

```python theme={null}
from pricing_api_client.models.disk import Disk

disk = Disk(
    provider="AWS",
    service="AmazonEBS",
    region="us-east-1",
    type="gp3",
    usage_type="ONDEMAND",
    capacity_gb=100,
)

disk_price = pricing.v1_pricing_disks_post(instance=disk)
```

## Marketplace

Publish your own rates. Re-posting the same natural key with a new cost keeps its id and reprices everything pinned to it.

```python theme={null}
from pricing_api_client.models.compute_price import ComputePrice
from pricing_api_client.models.compute_prices_request import ComputePricesRequest

entry = ComputePrice(
    provider="",
    service="compute",
    region="on-prem",
    instance_type="bare-metal-s",
    usage_type="ONDEMAND",
    operating_system="linux",
    cpu_cores=64,
    ram_gb=512,
    cost_per_hour=2.5,
)

published = marketplace.v1_marketplace_compute_post(
    body=ComputePricesRequest(entries=[entry])
)
price_id = published.data[0].id
```

List, patch, and retire:

```python theme={null}
page = marketplace.v1_marketplace_compute_get(limit=100)

updated = marketplace.v1_marketplace_compute_id_patch(
    id=price_id, body={"cost_per_hour": 3.0}
)

marketplace.v1_marketplace_compute_id_delete(id=price_id)
```

The same five verbs exist for disks, databases, and models. A model patch retires the old row and returns the new id, because model prices are versioned.

Errors on `/v1` are problem documents; the SDK raises `pricing_api_client.exceptions.ApiException` with `status` and the parsed body on `e.body`.