Update app.py

app.py

@@ -2,28 +2,37 @@ import os, json
 import gradio as gr
 from huggingface_hub import InferenceClient
 
-# Load your prompts
+# Load prompts
 with open("system_prompts.json", encoding="utf-8") as f:
     SYSTEM_PROMPTS = json.load(f)
 
 client = InferenceClient(model=os.getenv("HF_MODEL"), token=os.getenv("HF_TOKEN"))
 
 def run_model(prompt: str, text: str) -> str:
+    """Run Mistral with a system prompt + user text."""
     resp = client.chat.completions.create(
-        messages=[{"role":"system","content":prompt},{"role":"user","content":text}],
-        max_tokens=512, temperature=0.3
+        messages=[{"role":"system","content":prompt}, {"role":"user","content":text}],
+        max_tokens=512,
+        temperature=0.3
     )
     return resp.choices[0].message.content
 
-demo = gr.Blocks()
-with demo:
+# Create Gradio blocks
+with gr.Blocks() as demo:
     gr.Markdown("# 🌱 Ecolinguistic MCP Server")
     for name, prompt in SYSTEM_PROMPTS.items():
-        func_name = name.lower().replace(" ", "_")
-        textbox = gr.Textbox(lines=5, label="Text")
-        output = gr.Textbox(label="Response")
-        gr.Button("Run", api_name=func_name)\
-          .click(lambda txt, p=prompt: run_model(p, txt), inputs=textbox, outputs=output)
+        fn_name = name.lower().replace(" ", "_")
+        description = f"Tool: {name}"
+        
+        # Define a function for each prompt
+        def make_fn(p):
+            def tool_fn(text: str) -> str:
+                return run_model(p, text)
+            return tool_fn
+        
+        tool_fn = make_fn(prompt)
+        
+        # Register as MCP tool using gr.api
+        gr.api(fn=tool_fn, api_name=fn_name, api_description=description)
 
-if __name__ == "__main__":
-    demo.launch(mcp_server=True, share=True)
+demo.launch(mcp_server=True, share=True)