Update app.py

app.py

@@ -1,57 +1,43 @@
 import os
 import json
-import requests
 import gradio as gr
+from huggingface_hub import InferenceClient
 
 # Load tool prompts
 with open("system_prompts.json", "r", encoding="utf-8") as f:
     SYSTEM_PROMPTS = json.load(f)
 TOOLS = list(SYSTEM_PROMPTS.keys())
 
-# These are your HF Inference Endpoint settings
-HF_API_TOKEN = os.environ["HF_TOKEN"]
-HF_API_URL = "https://api-inference.huggingface.co/models/mistralai/Mistral-7B-Instruct-v0.2"
+# Set these environment variables in your HF Space or local env
+HF_TOKEN = os.environ["HF_TOKEN"]
+HF_MODEL = os.environ.get("HF_MODEL")
 
-def run_mistral(prompt: str, text: str) -> str:
+# Initialize HF InferenceClient
+client = InferenceClient(model=HF_MODEL, token=HF_TOKEN)
+
+def run_model(prompt: str, text: str) -> str:
     """
-    Calls the Hugging Face Inference Endpoint for Mistral using REST.
-    Sends 'prompt' as system message and 'text' as user message.
-    For reference, HF Inference endpoints accept JSON payloads like this :contentReference[oaicite:1]{index=1}.
+    Uses InferenceClient.chat.completions to call mnodel.
     """
-    headers = {
-        "Authorization": f"Bearer {HF_API_TOKEN}",
-        "Content-Type": "application/json"
-    }
-    payload = {
-        "inputs": [
+    resp = client.chat.completions.create(
+        messages=[
             {"role": "system", "content": prompt},
             {"role": "user", "content": text}
         ],
-        "parameters": {
-            "max_new_tokens": 512,
-            "temperature": 0.3,
-            "return_full_text": False
-        }
-    }
-    resp = requests.post(HF_API_URL, headers=headers, json=payload, timeout=60)
-    resp.raise_for_status()
-    out = resp.json()
-    # HF returns a list for chat models
-    if isinstance(out, list) and "generated_text" in out[0]:
-        return out[0]["generated_text"]
-    # else a dict
-    return out.get("generated_text", json.dumps(out))
+        max_tokens=512,
+        temperature=0.3
+    )
+    return resp.choices[0].message.content
 
 def ecoling_tool(tool: str, text: str) -> str:
-    # Looks up the prompt and calls the endpoint
-    return run_mistral(SYSTEM_PROMPTS[tool], text)
+    return run_model(SYSTEM_PROMPTS[tool], text)
 
 demo = gr.Interface(
     fn=ecoling_tool,
     inputs=[gr.Dropdown(TOOLS, label="Tool"), gr.Textbox(lines=8, label="Input Text")],
     outputs=[gr.Textbox(label="Response")],
     title="🌱 Ecolinguistic MCP Server",
-    description="Accessible via UI and standard MCP; powered by a HF Mistral endpoint."
+    description="UI + MCP server"
 )
 
 if __name__ == "__main__":