TL;DR: The Python client is not reading the kubeconfig you think it is, or the file is malformed. Set KUBECONFIG to the right file (or fix current-context in it), and load_kube_config works.

```text
kubernetes.config.config_exception.ConfigException: Invalid kube-config file. Expected key current-context in kube-config
```

## Fix it

1. Check which file kubectl uses: kubectl config view --minify. Expected: valid output with a current-context.
2. Point the client at the same file: export KUBECONFIG to that path, or pass config_file= explicitly to load_kube_config. Expected: no more ConfigException.
3. If the file itself is bad, fix current-context to name an existing context: kubectl config get-contexts, then kubectl config use-context [name]. Expected: kubectl config current-context prints the name.
4. Verify from Python: python -c "from kubernetes import config; config.load_kube_config(); print('ok')". Expected: ok.

## When this applies
- load_kube_config() raises ConfigException.
- kubectl works but the Python client does not (different file being read).

## When it doesn't
- The error is ApiException 401: the config loaded; the cluster rejected your credentials.
- Running inside a pod: use config.load_incluster_config() instead.

## Compatibility
- kubernetes client any version.

## Why it happens
kubectl merges ~/.kube/config with KUBECONFIG, while the Python client defaults to ~/.kube/config only. A KUBECONFIG env var set for kubectl, or a malformed file kubectl tolerates, breaks the client.

## Edge cases
- Multiple documents or Windows CRLF line endings can trip the YAML parser; normalize the file.
- In CI, write the kubeconfig from a secret to a temp file and pass config_file= explicitly.