fix: enable datasets,post_training and eval_tsaks works

Author: gyliu513

src/llama_stack_client/lib/cli/datasets/list.py

@@ -21,7 +21,7 @@ def list_datasets(ctx):
     console = Console()
     headers = ["identifier", "provider_id", "metadata", "type", "purpose"]
 
-    datasets_list_response = client.datasets.list()
+    datasets_list_response = client.beta.datasets.list()
     if datasets_list_response:
         table = Table()
         for header in headers:

src/llama_stack_client/lib/cli/datasets/register.py

@@ -68,7 +68,7 @@ def register(
         if not url:
             raise click.BadParameter("URL is required when dataset path is not specified")
 
-    response = client.datasets.register(
+    response = client.beta.datasets.register(
         dataset_id=dataset_id,
         source={"uri": url},
         metadata=metadata,

src/llama_stack_client/lib/cli/datasets/unregister.py

@@ -16,5 +16,5 @@
 def unregister(ctx, dataset_id: str):
     """Remove a dataset"""
     client = ctx.obj["client"]
-    client.datasets.unregister(dataset_id=dataset_id)
+    client.beta.datasets.unregister(dataset_id=dataset_id)
     click.echo(f"Dataset '{dataset_id}' unregistered successfully")

src/llama_stack_client/lib/cli/eval/run_benchmark.py

@@ -96,16 +96,16 @@ def run_benchmark(
     client = ctx.obj["client"]
 
     for benchmark_id in benchmark_ids:
-        benchmark = client.benchmarks.retrieve(benchmark_id=benchmark_id)
+        benchmark = client.alpha.benchmarks.retrieve(benchmark_id=benchmark_id)
         scoring_functions = benchmark.scoring_functions
         dataset_id = benchmark.dataset_id
 
-        results = client.datasets.iterrows(dataset_id=dataset_id, limit=-1 if num_examples is None else num_examples)
+        results = client.beta.datasets.iterrows(dataset_id=dataset_id, limit=-1 if num_examples is None else num_examples)
 
         output_res = {}
 
         for i, r in enumerate(tqdm(results.data)):
-            eval_res = client.eval.evaluate_rows(
+            eval_res = client.alpha.eval.evaluate_rows(
                 benchmark_id=benchmark_id,
                 input_rows=[r],
                 scoring_functions=scoring_functions,

src/llama_stack_client/lib/cli/eval/run_scoring.py

@@ -77,15 +77,15 @@ def run_scoring(
     output_res = {}
 
     if dataset_id is not None:
-        dataset = client.datasets.retrieve(dataset_id=dataset_id)
+        dataset = client.beta.datasets.retrieve(dataset_id=dataset_id)
         if not dataset:
             click.BadParameter(
                 f"Dataset {dataset_id} not found. Please register using llama-stack-client datasets register"
             )
 
         # TODO: this will eventually be replaced with jobs polling from server vis score_bath
         # For now, get all datasets rows via datasets API
-        results = client.datasets.iterrows(dataset_id=dataset_id, limit=-1 if num_examples is None else num_examples)
+        results = client.beta.datasets.iterrows(dataset_id=dataset_id, limit=-1 if num_examples is None else num_examples)
         rows = results.rows
 
     if dataset_path is not None:

src/llama_stack_client/lib/cli/eval_tasks/eval_tasks.py

@@ -49,12 +49,12 @@ def register(
         except json.JSONDecodeError as err:
             raise click.BadParameter("Metadata must be valid JSON") from err
 
-    response = client.eval_tasks.register(
-        eval_task_id=eval_task_id,
+    response = client.alpha.benchmarks.register(
+        benchmark_id=eval_task_id,
         dataset_id=dataset_id,
         scoring_functions=scoring_functions,
         provider_id=provider_id,
-        provider_eval_task_id=provider_eval_task_id,
+        provider_benchmark_id=provider_eval_task_id,
         metadata=metadata,
     )
     if response:

src/llama_stack_client/lib/cli/eval_tasks/list.py

@@ -21,7 +21,7 @@ def list_eval_tasks(ctx):
     client = ctx.obj["client"]
     console = Console()
     headers = []
-    eval_tasks_list_response = client.eval_tasks.list()
+    eval_tasks_list_response = client.alpha.benchmarks.list()
     if eval_tasks_list_response and len(eval_tasks_list_response) > 0:
         headers = sorted(eval_tasks_list_response[0].__dict__.keys())
 

src/llama_stack_client/lib/cli/post_training/post_training.py

@@ -44,7 +44,7 @@ def supervised_fine_tune(
     client = ctx.obj["client"]
     console = Console()
 
-    post_training_job = client.post_training.supervised_fine_tune(
+    post_training_job = client.alpha.post_training.supervised_fine_tune(
         job_uuid=job_uuid,
         model=model,
         algorithm_config=algorithm_config,
@@ -66,7 +66,7 @@ def get_training_jobs(ctx):
     client = ctx.obj["client"]
     console = Console()
 
-    post_training_jobs = client.post_training.job.list()
+    post_training_jobs = client.alpha.post_training.job.list()
     console.print([post_training_job.job_uuid for post_training_job in post_training_jobs])
 
 
@@ -80,7 +80,7 @@ def get_training_job_status(ctx, job_uuid: str):
     client = ctx.obj["client"]
     console = Console()
 
-    job_status_reponse = client.post_training.job.status(job_uuid=job_uuid)
+    job_status_reponse = client.alpha.post_training.job.status(job_uuid=job_uuid)
     console.print(job_status_reponse)
 
 
@@ -94,7 +94,7 @@ def get_training_job_artifacts(ctx, job_uuid: str):
     client = ctx.obj["client"]
     console = Console()
 
-    job_artifacts = client.post_training.job.artifacts(job_uuid=job_uuid)
+    job_artifacts = client.alpha.post_training.job.artifacts(job_uuid=job_uuid)
     console.print(job_artifacts)
 
 
@@ -107,7 +107,7 @@ def cancel_training_job(ctx, job_uuid: str):
     """Cancel the training job"""
     client = ctx.obj["client"]
 
-    client.post_training.job.cancel(job_uuid=job_uuid)
+    client.alpha.post_training.job.cancel(job_uuid=job_uuid)
 
 
 # Register subcommands