The Lekko SDK for Python
The lekko_client
package is published on PyPI.
pip install lekko_client
Creates a client that fetches configs from Lekko backend and caches them in memory. Configs are kept up to date via polling.
import lekko_client
lekko_client.initialize(lekko_client.CachedServerConfig(
owner_name="<REPOSITORY_OWNER>",
repo_name="<REPOSITORY_NAME>",
api_key="<API_KEY>", # Optional - defaults to "LEKKO_API_KEY" ENV Var
lekko_uri="<URI>", # Optional - defaults to "prod.api.lekko.dev:443"
context={}, # Optionally provide context dict to be merged into each get request
))
str_config = lekko_client.get_string("my_namespace", "my_config", {"context_key": "context_val"})
Creates a client that reads configs from a git repository on disk and caches them in memory. Configs are kept up to date via a file watcher.
import lekko_client
lekko_client.initialize(lekko_client.CachedGitConfig(
owner_name="<REPOSITORY_OWNER>",
repo_name="<REPOSITORY_NAME>",
git_repo_path="<GIT_REPO_PATH>",
api_key="<API_KEY>", # Optional - defaults to "LEKKO_API_KEY" ENV Var
lekko_uri="<URI>", # Optional - defaults to "prod.api.lekko.dev:443"
context={}, # Optionally provide context dict to be merged into each get request
))
str_config = lekko_client.get_string("my_namespace", "my_config", {"context_key": "context_val"})
Create a client that communicates with the Lekko API or a Lekko sidecar for remote evaluation
import lekko_client
lekko_client.initialize(lekko_client.APIConfig( # Or lekko_client.SidecarConfig
owner_name="<REPOSITORY_OWNER>",
repo_name="<REPOSITORY_NAME>",
api_key="<API_KEY>", # Optional - defaults to "LEKKO_API_KEY" ENV Var
context={}, # Optionally provide context dict to be merged into each get request
))
str_config = lekko_client.get_string("my_namespace", "my_config", {"context_key": "context_val"})
lekko_client.initialize()
must be invoked prior to calling the lekko_client.get_*()
functions and should not be called multiple times. We recommend invoking it early in your app's lifecycle, for example when constructing your Flask app or as part of FastAPI's lifecycle context manager.
We recommend you invoke lekko_client.close()
during app shutdown. This will ensure all evaluation events are properly tracked and the SDK client is deregistered correctly.
It is also possible to do your own lifecycle management and avoid the lekko_client.initialize()
and lekko_client.get_*()
methods entirely. The above root module-level methods are provided for convenience to suit most use cases, without having to explicitly manage a client instance and pass it around. Feel free to construct any of the clients in lekko_client.clients
manually.
For example:
from lekko_client.clients import APIClient
client = APIClient(
owner_name="<REPOSITORY_OWNER>",
repo_name="<REPOSITORY_NAME>",
api_key="<API_KEY>",
)
...
def use_client(client):
str_config = client.get_string("my_namespace", "my_config", {"context_key": "context_val"})
return str_config
use_client(client)
There are two methods to retrieve protobuf configs, with one allowing you to specify the expected proto message type.
-
get_proto_by_type(key, context, proto_message_type)
will attempt to convert the config to the specified type and raiseMismatchedProtoType
on failure. -
get_proto(key, context)
will attempt to locate the proto message type in the proto symbol database, which tracks imported proto types. If unable to locate the type, it will simply return the message as agoogle.protobuf.any_pb2.Any
For example, if you have a proto config named proto_config
defined as:
{
"@type": "type.googleapis.com/google.protobuf.BoolValue",
"value": false
}
You would see the following behavior:
>>> config_value = client.get_proto("proto", {})
>>> type(config_value)
<class 'google.protobuf.any_pb2.Any'>
>>> config_value.value
b''
But after importing the appropriate type:
>>> from google.protobuf import wrappers_pb2
>>> config_value = client.get_proto("proto", {})
>>> type(config_value)
<class 'google.protobuf.wrappers_pb2.BoolValue'>
>>> config_value.value
False
See: https://github.com/lekkodev/python-sdk/blob/main/example.py