A Python wrapper for the data.gouv.fr API that allows you to interact easily with datasets and resources across all three platforms (production www, demo, and dev). Install it through PyPI:
pip install datagouv-clientRequirements: Python >= 3.10
from datagouv import Dataset, Resource, Topic
# Get a dataset and its resources
dataset = Dataset("5d13a8b6634f41070a43dff3")
print(f"Dataset: {dataset.title}")
print(f"Resources: {len(dataset.resources)}")
# Download a resource
resource = dataset.resources[0]
resource.download("my_file.csv")
# Get a topic and its elements
topic = Topic("68b6e6dbdac745f47d4ff6e0")
elements = topic.elements
datasets = topic.datasetsIf you only want to retrieve existing objects (aka you don't want to modify them on datagouv), here is what a workflow could look like:
from datagouv import Dataset, Resource, Organization
dataset = Dataset("5d13a8b6634f41070a43dff3") # you can find a dataset's id in the `Informations` tab of its landing page
# you can now access a bunch of info about the dataset
print(dataset.title)
print(dataset.description)
print(dataset.created_at)
print(dataset.organization) # this is an instance of Organization
print(dataset) # this displays all the attributes of the dataset as a dict
# and of course its resources, which are all Resource instances
for res in dataset.resources:
print(res.title)
print(res.url) # this is the download URL of the resource
print(res.id) # the id of the resource itself
print(res.dataset_id) # the id of the dataset the resource belongs to
print(res) # this displays all the attributes of the resource as a dict
# if you are only interested in a specific resource
resource = Resource("f868cca6-8da1-4369-a78d-47463f19a9a3") # you can find a resource's id in its `Métadonnées` tab
print(resource)
# you can also access a dataset from one of its resources
d = resource.dataset # this returns an instance of Dataset
# you can also download a resource locally (**Note:** if it doesn't exist, parent path will be created)
resource.download("./file.csv") # this saves the resource in your working directory as "file.csv"
# and a subset or all resources of a dataset (**Note:** if it doesn't exist, parent path will be created)
# the files are named `resource_id.format` (for instance f868cca6-8da1-4369-a78d-47463f19a9a3.csv)
d.download_resources(
folder="data", # if not specified, saves them into your working directory
resources_types=["main", "documentation"], # default is only main resources
)
organization = Organization("646b7187b50b2a93b1ae3d45") # you can find an organization's id in the `Informations` tab of its landing page, in "Informations techniques"
# you can loop through the organization's datasets, which are Dataset instances
for dat in organization.datasets:
print(f"{dat.title} has {len(dat.resources)} resources")Note: If you encounter errors during API calls, the client will raise appropriate exceptions (e.g.,
PermissionErrorfor authentication issues,httpx.HTTPErrorfor API errors).
Note: If you want to get objects from demo or dev, you must use a client:
from datagouv import Client, Dataset, Resource
dataset = Dataset("5d13a8b6634f41070a43dff3", _client=Client("demo"))You can also access objects' metrics (views, downloads) with the get_monthly_traffic_metrics function:
for month_metrics in Dataset("5d13a8b6634f41070a43dff3").get_monthly_traffic_metrics(
start_month="2025-01", # optional, goes back as far as possible if not set
end_month="2025-06", # optional, until today if not set
):
print(month_metrics)The metrics differ depending on the object:
- for datasets:
{
"__id": 43110395,
"dataset_id": "6789251f3a805425afee55e6",
"metric_month": "2025-01",
"monthly_visit": 233,
"monthly_download_resource": 3
}- for resources:
{
"__id": 58728461,
"resource_id": "5ffa8553-0e8f-4622-add9-5c0b593ca1f8",
"dataset_id": "5c4ae55a634f4117716d5656",
"metric_month": "2025-04",
"monthly_download_resource": 5669
}- for organizations:
{
"__id": 7,
"organization_id": "646b7187b50b2a93b1ae3d45",
"metric_month": "2023-07",
"monthly_visit_dataset": 27196,
"monthly_download_resource": 1085933,
"monthly_visit_reuse": 123,
"monthly_visit_dataservice": 456
}If you want to modify objects on the datagouv platforms, you will need to create an authenticated client:
from datagouv import Client
client = Client(
environment="www", # here you can set which platform the client will interact with, default is production
api_key="MY_SECRET_API_KEY", # your API key, that grants your rights on the platform
)Note: You can find your API key on https://www.data.gouv.fr/fr/admin/me/ (don't forget to change the prefix to get the key from the right environment).
Once your client is set up, you can instantiate datasets and resources from it. Of course, you will only be allowed to modify objects according to your rights (so objects created by you or an organization you are part of):
dataset = client.dataset("5d13a8b6634f41070a43dff3")
# this is also a Dataset instance, with all the same attributes as above, but since you're authenticated, you have access to new methods
dataset.update({"title": "A brand new title"}) # update the dataset online with the payload you give, and also update the attributes of the object
print(dataset.title) # -> "A brand new title"
dataset.delete() # delete the dataset, use with caution!
# you can also modify the extras
dataset.update_extras(payload)
dataset.delete_extras(payload)
# the methods are the same for resources
for idx, res in enumerate(dataset.resources):
res.update({"title": f"Resource n°{idx + 1}"})
print(res.title) # -> "Resource n°X"
# delete every third resource
if idx % 3 == 0:
res.delete()With an authenticated client, you are also allowed to create datasets and resources on the environment you specified:
dataset = client.dataset().create(
{
"title": "New dataset",
"description": "A description is required",
"organization": "646b7187b50b2a93b1ae3d45", # the organization that will own the dataset
},
) # this creates a dataset with the values you specified, and instantiates a Dataset
dataset.update({"tags": ["environment", "water"]})
# alternatively you can create a dataset from an organization, and it will be attached to it
organization = client.organization("646b7187b50b2a93b1ae3d45")
dataset = organization.create_dataset(
{
"title": "New dataset",
"description": "A description is a required",
}
)There are two types of resources on datagouv:
static: a file is uploaded directly on the platformremote: reference the URL of a file that is stored somewhere else on the internet
You have two options to create a resource (of any type):
- from the client itself, by specifying the id of the dataset you want to include it into (you must have the rights on the dataset):
# to create a static resource from a file
resource = client.resource().create_static(
file_to_upload="path/to/your/file.txt",
payload={"title": "New static resource"},
dataset_id="5d13a8b6634f41070a43dff3",
) # this creates a static resource with the values you specified, and instantiates a Resource
# to create a remote resource from an url
resource = client.resource().create_remote(
payload={"url": "http://example.com/file.txt", "title": "New remote resource"},
dataset_id="5d13a8b6634f41070a43dff3",
) # this creates a remote resource with the values you specified, and instantiates a Resource- from the dataset you want to include it into (you must have the rights on the dataset), in which case you don't have to specify the
dataset_id:
dataset = client.dataset("5d13a8b6634f41070a43dff3")
# to create a static resource from a file
resource = dataset.create_static(
file_to_upload="path/to/your/file.txt",
payload={"title": "New static resource"},
) # this creates a static resource with the values you specified, and instantiates a Resource
# to create a remote resource from an url
resource = dataset.create_remote(
payload={"url": "http://example.com/file.txt", "title": "New remote resource"},
) # this creates a remote resource with the values you specified, and instantiates a Resource
# to update the file of a static resource
resource.update({"title": "New title"}, file_to_upload="path/to/your/new_file.txt")Note: If you are not planning to use an object's attributes, you may prevent the initial API call using
fetch=False, in order not to unnecessarily ping the API.
dataset = client.dataset("5d13a8b6634f41070a43dff3", fetch=False)
print(dataset.title) # -> this will fail because the attributes are not set from the initial call
# but you can update the object as usual
dataset.update({"title": "New title"})
print(dataset.title) # -> "New title" because the attributes are set from the responseMany datagouv endpoints are paginated, which can make it tedious to retrieve all objects. An instance of Client has a method to create an iterator from any endpoint that returns paginated data:
for obj in client.get_all_from_api_query(
"api/1/datasets/?organization=534fff81a3a7292c64a77e5c", # get all datasets from a specific organization
mask="data{id,title,resources{id,title}}", # you can apply a mask to retrieve only specific fields of the objects
cast_as=Dataset, # you can get the results as objects to manipulate them more easily
):
print(f"Dataset {obj['title']} has {len(obj['resources'])} resources") # if cast_as is not used, otherwise `obj.id` and `obj.resources`You can also check if resources have been updated more recently than others:
# Check if any resource in a dataset has been updated more recently than a specific resource
resource = Resource("f868cca6-8da1-4369-a78d-47463f19a9a3")
has_newer_updates = resource.check_if_more_recent_update("5d13a8b6634f41070a43dff3")Contributions and feedback are welcome! Main guidelines:
- as few API calls as possible (use responses to create/update objects)
- build on the existing
Remember to format, lint, and sort imports with Ruff before committing (checks will remind you anyway):
pip install .[dev]
ruff check --fix .
ruff format .The release process uses bump'X.