Update src/streamlit_app.py

src/streamlit_app.py

@@ -1060,22 +1060,22 @@ with tabs[4]:
                 if "llm_result" not in st.session_state:
                     st.session_state["llm_result"] = None
                 
-                if st.button("Get AI Recommendation (tiny local LLM)", key="ai_reco"):
+                if st.button("Get AI Recommendation (via HF API)", key="ai_reco"):
                     summary = st.session_state.get("automl_summary", {})
                     if not summary:
                         st.warning("Please run AutoML first to generate context.")
                         st.stop()
                     try:
-                        from transformers import pipeline
-                        st.info("Loading compact instruction-tuned model (in-memory)...")
+                        import requests, json
+                        st.info("Contacting Hugging Face Inference API (Mixtral-8x7B-Instruct)…")
                 
-                        
-                        assistant = pipeline("text2text-generation", model="google/flan-t5-small")
+                        API_URL = "https://api-inference.huggingface.co/models/mistralai/Mixtral-8x7B-Instruct-v0.1"
+                        headers = {"Authorization": f"Bearer {st.secrets['HF_TOKEN']}"}
                 
                         prompt = f"""
                         You are an ML model tuning advisor.
-                        Based on this AutoML summary, suggest 3 concise steps to improve performance
-                        if overfitting, underfitting, or data-quality issues are seen.
+                        Based on this AutoML summary, suggest 3 concise, actionable steps
+                        to improve model performance if overfitting, underfitting, or data-quality issues are observed.
                 
                         Use case: {summary.get('use_case')}
                         Target: {summary.get('target')}
@@ -1084,11 +1084,32 @@ with tabs[4]:
                         Leaderboard: {summary.get('leaderboard')}
                         """
                 
-                        result = assistant(prompt, max_new_tokens=100)[0]["generated_text"]
-                        st.session_state["llm_result"] = result
-                        log("LLM in-memory recommendation generated successfully.")
+                        payload = {
+                            "inputs": prompt,
+                            "parameters": {"max_new_tokens": 200, "temperature": 0.7}
+                        }
+                
+                        response = requests.post(API_URL, headers=headers, json=payload, timeout=60)
+                        response.raise_for_status()
+                        result = response.json()
+                
+                        if isinstance(result, list) and "generated_text" in result[0]:
+                            text = result[0]["generated_text"]
+                        elif isinstance(result, dict) and "generated_text" in result:
+                            text = result["generated_text"]
+                        else:
+                            text = json.dumps(result, indent=2)
+                
+                        st.session_state["llm_result"] = text.strip()
+                        log("HF API recommendation generated successfully.")
+                        st.success("AI Recommendation (Mixtral-8x7B-Instruct):")
+                        st.markdown(st.session_state["llm_result"])
+                
                     except Exception as e:
-                        st.session_state["llm_result"] = f" LLM generation failed: {e}"
+                        err_msg = f"HF Inference API call failed: {e}"
+                        st.error(err_msg)
+                        log(err_msg)
+
                 
                 # Persist output even after rerun
                 if st.session_state["llm_result"]: