fix: use SmolLM2-1.7B via HF Inference API - Tiny Titan eligible

app.py

@@ -11,6 +11,12 @@ from chemistry import (
     calculate_lye, score_recipe, find_substitutes,
     cure_timeline, profile_summary, list_oils, OILS
 )
+import os
+from huggingface_hub import InferenceClient
+
+client = InferenceClient("HuggingFaceTB/SmolLM2-1.7B-Instruct")
+
+MODEL = "HuggingFaceTB/SmolLM2-1.7B-Instruct"
 
 # ---------------------------------------------------------------------------
 # Config — change MODEL to "Qwen/Qwen2.5-7B-Instruct" on HF Space
@@ -88,19 +94,18 @@ def detect_mode(text: str) -> str:
 # ---------------------------------------------------------------------------
 
 def ask_model(prompt: str, context: str) -> str:
-    """Call local Ollama model."""
     full_prompt = f"{context}\n\nUser asked: {prompt}"
     try:
-        response = ollama.chat(
-            model=MODEL,
+        response = client.chat_completion(
             messages=[
                 {"role": "system", "content": SYSTEM_PROMPT},
                 {"role": "user", "content": full_prompt}
-            ]
+            ],
+            max_tokens=500,
         )
-        return response["message"]["content"]
+        return response.choices[0].message.content
     except Exception as e:
-        return f"Model error: {str(e)}. Make sure Ollama is running."
+        return f"Model error: {str(e)}"
 
 
 # ---------------------------------------------------------------------------

requirements.txt

@@ -1,3 +1,2 @@
 gradio
-ollama
 huggingface_hub