DEV: first

.gitignore

@@ -0,0 +1 @@
+__pycache__

README.md

@@ -11,7 +11,6 @@ license: apache-2.0
 short_description: Podcast Generator MCP Server
 tags:
 - Agents-MCP-Hackathon
+- agent-demo-track
 - mcp-server-track
 ---
-
-An example chatbot using [Gradio](https://gradio.app), [`huggingface_hub`](https://huggingface.co/docs/huggingface_hub/v0.22.2/en/index), and the [Hugging Face Inference API](https://huggingface.co/docs/api-inference/index).

app.py

@@ -1,64 +1,64 @@
+import os
+
 import gradio as gr
-from huggingface_hub import InferenceClient
 
-"""
-For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
-"""
-client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
+from smolagents import CodeAgent, ToolCallingAgent, MCPClient, InferenceClientModel
+from openai import OpenAI
+
+
+model_name = None
+workspace = "imessam"
+environment = None
+app_name = "example-vllm-openai-compatible"
+function_name = "serve"
+
+api_key = os.getenv("MODAL_API_KEY")
 
+client = OpenAI(api_key=api_key)
+
+prefix = workspace + (f"-{environment}" if environment else "")
+
+client.base_url = (
+    f"https://{prefix}--{app_name}-{function_name}.modal.run/v1"
+)
 
-def respond(
-    message,
-    history: list[tuple[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-):
-    messages = [{"role": "system", "content": system_message}]
+print(str(client.base_url.host))
 
-    for val in history:
-        if val[0]:
-            messages.append({"role": "user", "content": val[0]})
-        if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
+model = client.models.list().data[0]
+model_id = model.id
 
-    messages.append({"role": "user", "content": message})
+
+def generate_podcast(prompt : str, history: list) -> str:
 
     response = ""
 
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
+    try:
+        mcp_client = MCPClient([
+            {"url": "https://agents-mcp-hackathon-websearch.hf.space/gradio_api/mcp/sse", "transport": "sse"},
+            {"url": "https://agents-mcp-hackathon-footballmatchesbydate.hf.space/gradio_api/mcp/sse", "transport": "sse"}] # This is the MCP Server we created in the previous section
+        )
+        tools = mcp_client.get_tools()
+
+        model = InferenceClientModel()
+        agent = CodeAgent(tools=[*tools], model=model)
+
 
-        response += token
-        yield response
+        response = str(agent.run(prompt))
+        
 
+    finally:
+        mcp_client.disconnect()
+
+
+    return response
 
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
 demo = gr.ChatInterface(
-    respond,
-    additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
-        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=0.95,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
-        ),
-    ],
+    fn=generate_podcast,
+    type="messages",
+    examples=["Generate a podcast about AI"],
+    title="Podcast Generator Agent and MCP Server",
+    description="This is an agent that uses MCP tools to generate a podcast, and can be used as an MCP server.",
 )
 
-
 if __name__ == "__main__":
-    demo.launch()
+    demo.launch(mcp_server=True)

modal_setup.py

@@ -0,0 +1,99 @@
+import modal
+import os
+
+
+app_name : str = "example-vllm-openai-compatible"
+
+app = modal.App(name=app_name)
+
+
+
+print(f"setting up container image ...")
+
+vllm_image = (
+    modal.Image.debian_slim(python_version="3.12")
+    .pip_install(
+        "vllm==0.7.2",
+        "huggingface_hub[hf_transfer]==0.26.2",
+        "flashinfer-python==0.2.0.post2",  # pinning, very unstable
+        extra_index_url="https://flashinfer.ai/whl/cu124/torch2.5",
+    )
+    .env({"HF_HUB_ENABLE_HF_TRANSFER": "1"})  # faster model transfers
+)
+
+vllm_image = vllm_image.env({"VLLM_USE_V1": "1"})
+
+print(f" done setting up container image.")
+
+
+
+
+MODELS_DIR = "/llamas",
+MODEL_NAME = "neuralmagic/Meta-Llama-3.1-8B-Instruct-quantized.w4a16"
+MODEL_REVISION = "a7c09948d9a632c2c840722f519672cd94af885d"
+
+
+print(f" downloading model weights...")
+
+
+hf_cache_vol = modal.Volume.from_name("huggingface-cache", create_if_missing=True)
+vllm_cache_vol = modal.Volume.from_name("vllm-cache", create_if_missing=True)
+
+
+print(f" done downloading model weights.")
+
+
+
+print(f"building engine...")
+
+N_GPU = 1  # tip: for best results, first upgrade to more powerful GPUs, and only then increase GPU count
+
+MINUTES = 60  # seconds
+
+VLLM_PORT = 8000
+
+
+@app.function(
+    image = vllm_image,
+    secrets=[modal.Secret.from_name("api_key")],
+    gpu=f"H100:{N_GPU}",
+    scaledown_window=15 * MINUTES,  # how long should we stay up with no requests?
+    timeout=10 * MINUTES,  # how long should we wait for container start?
+    volumes={
+        "/root/.cache/huggingface": hf_cache_vol,
+        "/root/.cache/vllm": vllm_cache_vol,
+    },
+)
+@modal.concurrent(
+    max_inputs=100
+)  # how many requests can one replica handle? tune carefully!
+@modal.web_server(port=VLLM_PORT, startup_timeout=50 * MINUTES)
+def serve():
+    import subprocess
+
+    API_KEY = os.environ["MODAL_API_KEY"]
+
+    cmd = [
+        "vllm",
+        "serve",
+        "--uvicorn-log-level=info",
+        MODEL_NAME,
+        "--revision",
+        MODEL_REVISION,
+        "--host",
+        "0.0.0.0",
+        "--port",
+        str(VLLM_PORT),
+        "--api-key",
+        API_KEY,
+        "--enable-auto-tool-choice"
+        " ",
+        "--tool-call-parser",
+        "llama3_json"
+    ]
+
+    subprocess.Popen(" ".join(cmd), shell=True)
+
+
+print(f"done building engine.")
+

requirements.txt

@@ -1 +1,4 @@
-huggingface_hub==0.25.2
+gradio[mcp]
+modal
+openai
+openai-agents