Update app.py

app.py

@@ -1,73 +1,36 @@
-import gradio as gr
-from huggingface_hub import InferenceClient
-from fastmcp import FastMCPClient
 import os
+import gradio as gr
+from smolagents import MCPClient, InferenceClientModel, CodeAgent
 
-# Set up inference client (LLM)
-client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
-
-# Set up FastMCP client (replace with your actual MCP server URL)
-mcp = FastMCPClient("Jobicy Remote Jobs Agent", url="https://oppaai-job-search-mcp-server.hf.space/gradio_api/mcp/sse")
+# Your MCP server URL (replace with your actual HF space MCP endpoint)
+MCP_URL = "https://oppaai-job-search-mcp-server.hf.space/gradio_api/mcp/sse"
 
-def respond(
-    message,
-    history: list[tuple[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
-):
-    messages = [{"role": "system", "content": system_message}]
+# Initialize MCP client to connect to your MCP server
+mcp_client = MCPClient({"url": MCP_URL, "transport": "streamable-http"})
 
-    for val in history:
-        if val[0]:
-            messages.append({"role": "user", "content": val[0]})
-        if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
+# Fetch available tools from MCP server
+tools = mcp_client.get_tools()
 
-    messages.append({"role": "user", "content": message})
+# Initialize your LLM model from Hugging Face (requires HF token for private/protected models)
+llm_model = InferenceClientModel(
+    model_id="HuggingFaceH4/zephyr-7b-beta",
+    token=os.getenv("HUGGINGFACE_TOKEN"),  # Set this environment variable if needed
+)
 
-    # Tool invocation keyword check (e.g., "search jobs for Python in Canada")
-    if "search jobs" in message.lower():
-        # Simple NLP parse: extract keyword, country, industry
-        # (You can improve this with spaCy or regex later)
-        keyword = message.split("for")[-1].strip().split(" in ")[0].strip()
-        country = message.split(" in ")[-1].strip() if " in " in message else ""
-        res = mcp.call_tool("search_jobs", inputs={"keyword": keyword, "country": country, "limit": 5})
+# Create an agent that uses the tools and the LLM
+agent = CodeAgent(tools=tools, model=llm_model, stream_outputs=True)
 
-        if "error" in res:
-            return f"❌ {res['error']}"
-        if not res.get("jobs"):
-            return "No jobs found with those filters."
 
-        job_lines = [
-            f"**{j['title']}** at {j['company']} ({j['location']})\n{j['salary']}\nPosted: {j['pubDate']}\n[Apply Here]({j['url']})"
-            for j in res["jobs"]
-        ]
-        return "\n\n---\n\n".join(job_lines)
+def respond(message, history):
+    # Agent processes the input message, uses tools if needed, else LLM
+    reply = agent.run(message)
+    return reply
 
-    # Fallback to LLM
-    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 or ""
-        response += token
-        yield response
 
 demo = gr.ChatInterface(
-    respond,
-    additional_inputs=[
-        gr.Textbox(value="You are a friendly assistant who can also help find remote jobs.", 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)"),
-    ],
-    title="Job Search Assistant Chatbot"
+    fn=respond,
+    title="Job Search Assistant (Agent)",
+    description="Ask me about job listings or general questions — I’ll use the job search tool when needed!",
 )
 
 if __name__ == "__main__":