#include "tensorflow/core/util/tensor_bundle/tensor_bundle.h"
#include "tensorflow/core/util/tensor_slice_reader.h"

namespace tensorflow {
namespace checkpoint {

class TensorSliceReader;

// A wrapper around BundleReader (for V2 checkpoints) and
// checkpoint::TensorSliceReader (for V1), that is more easily SWIG wrapped for
// other languages.
//
// The class currently only interacts with single-slice (i.e., non-partitioned)

#include "tensorflow/core/util/saved_tensor_slice_util.h"
#include "tensorflow/core/util/tensor_bundle/naming.h"
#include "tensorflow/core/util/tensor_bundle/tensor_bundle.h"

namespace tensorflow {
namespace checkpoint {

class TensorSliceReader;

CheckpointReader::CheckpointReader(const string& filename, TF_Status* status)
    : reader_(nullptr),
      v2_reader_(nullptr),
      var_to_shape_map_(nullptr),

]


app = FastAPI()


@app.get("/stream-raw", response_class=StreamingResponse)
async def stream_raw() -> AsyncIterable[str]:
    """Async generator with no internal await - would hang without checkpoint."""
    i = 0
    while True:
        yield f"item {i}\n"
        i += 1


@app.get("/stream-jsonl")
async def stream_jsonl() -> AsyncIterable[int]:
    """JSONL async generator with no internal await."""

                                                      const char* errMsg);

// TF_NewCheckpointReader() return the CheckpointReader that can be use to
// investigate or load the variable from the checkpoint file
typedef struct TF_CheckpointReader TF_CheckpointReader;
TF_CAPI_EXPORT extern TF_CheckpointReader* TF_NewCheckpointReader(
    const char* filename, TF_Status* status);

variables from your model are also going to be untrusted. That means that if
your code interacts with the filesystem, network, etc. and uses checkpointed
variables as part of those interactions (ex: using a string variable to build a
filesystem path), a maliciously created checkpoint might be able to change the
targets of those operations, which could result in arbitrary
read/write/executions.

}

void TF_MakeInternalErrorStatus(TF_Status* status, const char* errMsg) {
  status->status = tensorflow::errors::Internal(errMsg);
}

struct TF_CheckpointReader : public tensorflow::checkpoint::CheckpointReader {
  using tensorflow::checkpoint::CheckpointReader::CheckpointReader;
  std::vector<std::string> variable_list;
};

TF_CheckpointReader* TF_NewCheckpointReader(const char* filename,

*   `tf.train.Checkpoint`:
    *   Now accepts a `root` argument in the initialization, which generates a
        checkpoint with a root object. This allows users to create a
        `Checkpoint` object that is compatible with Keras `model.save_weights()`
        and `model.load_weights`. The checkpoint is also compatible with the
        checkpoint saved in the `variables/` folder in the SavedModel.

  - [Changelog since v1.32.1](#changelog-since-v1321)
  - [Important Security Information](#important-security-information-2)
    - [CVE-2025-0426: Node Denial of Service via Kubelet Checkpoint API](#cve-2025-0426-node-denial-of-service-via-kubelet-checkpoint-api)
  - [Changes by Kind](#changes-by-kind-11)
    - [Feature](#feature-8)
    - [Bug or Regression](#bug-or-regression-10)
    - [Other (Cleanup or Flake)](#other-cleanup-or-flake-4)

- Added a new feature gate `CheckpointRestore` to enable support to checkpoint containers. If enabled it is possible to checkpoint a container using the newly kubelet API (/checkpoint/{podNamespace}/{podName}/{containerName}). ([#104907](https://github.com/kubernetes/kubernetes/pull/104907), [@adrianreber](https://github.com/adrianreber)) [SIG Node and Testing]

  - [Changelog since v1.30.9](#changelog-since-v1309)
  - [Important Security Information](#important-security-information)
    - [CVE-2025-0426: Node Denial of Service via Kubelet Checkpoint API](#cve-2025-0426-node-denial-of-service-via-kubelet-checkpoint-api)
  - [Changes by Kind](#changes-by-kind-4)
    - [Feature](#feature-2)
    - [Bug or Regression](#bug-or-regression-3)
    - [Other (Cleanup or Flake)](#other-cleanup-or-flake-1)