feat: add cluster name stop start and reset functions to skypilot

app.py

@@ -1,3 +1,4 @@
+import sky
 import boto3
 from transformers import AutoTokenizer
 from botocore.exceptions import NoCredentialsError
@@ -178,7 +179,15 @@ def load_tools():
     print(f"all_tools_list: {all_tools_list}")  # Debugging line
     return gr.update(choices=all_tools_list)
 
-def set_environ(OPENAI_API_KEY: str = "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",
+def start_sky_pilot(cluster_name: str):
+    sky.start(cluster_name)
+
+def stop_sky_pilot(cluster_name: str):
+    sky.stop(cluster_name)
+
+def status_sky_pilot(cluster_name: str):
+    return sky.status(cluster_name)
+def set_environ(OPENAI_API_KEY: str = "",
                 WOLFRAMALPH_APP_ID: str = "",
                 WEATHER_API_KEYS: str = "",
                 BING_SUBSCRIPT_KEY: str = "",
@@ -331,26 +340,25 @@ def fetch_tokenizer(model_name):
         return f"Error loading tokenizer: {str(e)}"
 
 # Add this function to handle the button click
+import sky
+
 def deploy_on_sky_pilot(model_name: str, tokenizer: str, accelerators: str):
-    # Create serving.yaml for SkyPilot deployment
+    # Create a SkyPilot Task
-    serving_yaml = {
+    #TODO have ai generate a yaml file for the configuration the user desires add this as a tool 
-        "resources": {
+    task = sky.Task(
-            "accelerators": accelerators
+        setup="conda create -n vllm python=3.9 -y\nconda activate vllm\ngit clone https://github.com/vllm-project/vllm.git\ncd vllm\npip install .\npip install gradio",
-        },
+        run="conda activate vllm\necho 'Starting vllm api server...'\npython -u -m vllm.entrypoints.api_server --model $MODEL_NAME --tensor-parallel-size $SKYPILOT_NUM_GPUS_PER_NODE --tokenizer $TOKENIZER 2>&1 | tee api_server.log &\necho 'Waiting for vllm api server to start...'\nwhile ! `cat api_server.log | grep -q 'Uvicorn running on'`; do sleep 1; done\necho 'Starting gradio server...'\npython vllm/examples/gradio_webserver.py",
-        "envs": {
+        envs={
             "MODEL_NAME": model_name,
             "TOKENIZER": AutoTokenizer.from_pretrained(model_name)
         },
-        "setup": "conda create -n vllm python=3.9 -y\nconda activate vllm\ngit clone https://github.com/vllm-project/vllm.git\ncd vllm\npip install .\npip install gradio",
+        resources={
-        "run": "conda activate vllm\necho 'Starting vllm api server...'\npython -u -m vllm.entrypoints.api_server --model $MODEL_NAME --tensor-parallel-size $SKYPILOT_NUM_GPUS_PER_NODE --tokenizer $TOKENIZER 2>&1 | tee api_server.log &\necho 'Waiting for vllm api server to start...'\nwhile ! `cat api_server.log | grep -q 'Uvicorn running on'`; do sleep 1; done\necho 'Starting gradio server...'\npython vllm/examples/gradio_webserver.py"
+            "accelerators": accelerators
         }
+    )
 
-    # Write serving.yaml to file
+    # Launch the task on SkyPilot
-    with open('serving.yaml', 'w') as f:
+    sky.launch(task,cluster_name=cluster_name)
-        yaml.dump(serving_yaml, f)
-
-    # Deploy on SkyPilot
-    os.system("sky launch serving.yaml")
 
 # Add this line where you define your Gradio interface
 
@@ -384,9 +392,9 @@ with gr.Blocks() as demo:
             SCENEX_API_KEY = gr.Textbox(label="Scenex api key:", placeholder="Key to use sceneXplain", type="text")
             STEAMSHIP_API_KEY = gr.Textbox(label="Steamship api key:", placeholder="Key to use image generation", type="text")
             HUGGINGFACE_API_KEY = gr.Textbox(label="Huggingface api key:", placeholder="Key to use models in huggingface hub", type="text")
-        HUGGINGFACE_TOKEN = gr.Textbox(label="HuggingFace Token:", placeholder="Token for huggingface", type="text"),
-        AMADEUS_ID = gr.Textbox(label="Amadeus id:", placeholder="Id to use Amadeus", type="text")
             AMADEUS_KEY = gr.Textbox(label="Amadeus key:", placeholder="Key to use Amadeus", type="text")
+            AMADEUS_ID = gr.Textbox(label="Amadeus ID:", placeholder="Amadeus ID",
+                                    type="text")
             AWS_ACCESS_KEY_ID = gr.Textbox(label="AWS Access Key ID:", placeholder="AWS Access Key ID", type="text")
             AWS_SECRET_ACCESS_KEY = gr.Textbox(label="AWS Secret Access Key:", placeholder="AWS Secret Access Key", type="text")
             AWS_DEFAULT_REGION = gr.Textbox(label="AWS Default Region:", placeholder="AWS Default Region", type="text")
@@ -423,12 +431,21 @@ with gr.Blocks() as demo:
                                 info="Choose the model to solve your question, Default means ChatGPT."
                             )
                             tokenizer_output = gr.outputs.Textbox(label="Tokenizer")
+                            cluster_name = gr.outputs.Textbox(label="Cluster")
                             model_chosen.change(fetch_tokenizer, outputs=tokenizer_output)
                             available_accelerators = ["A100", "V100", "P100", "K80", "T4", "P4"]
                             accelerators = gr.Dropdown(available_accelerators, label="Accelerators:")
                             buttonDeploy = gr.Button("Deploy on SkyPilot")
 
-                        buttonDeploy.click(deploy_on_sky_pilot, [model_chosen, tokenizer_output, accelerators, HUGGINGFACE_TOKEN])
+                            buttonDeploy.click(deploy_on_sky_pilot, [model_chosen, tokenizer_output, accelerators, HUGGINGFACE_API_KEY])
+                            buttonStart = gr.Button("Start SkyPilot")
+                            buttonStart.click(start_sky_pilot, [cluster_name])
+
+                            buttonStop = gr.Button("Stop SkyPilot")
+                            buttonStop.click(stop_sky_pilot, [cluster_name])
+
+                            buttonStatus = gr.Button("Check SkyPilot Status")
+                            buttonStatus.click(status_sky_pilot, [cluster_name])
                     with gr.Row():
                         tools_search = gr.Textbox(
                             lines=1,
@@ -470,7 +487,6 @@ with gr.Blocks() as demo:
             SCENEX_API_KEY,
             STEAMSHIP_API_KEY,
             HUGGINGFACE_API_KEY,
-        HUGGINGFACE_TOKEN,
             AMADEUS_ID,
             AMADEUS_KEY,
         ], outputs=key_set_btn)