Update generate_swamiji_intelligent.py

generate_swamiji_intelligent.py

@@ -61,15 +61,21 @@ def load_swamiji_prompt(analyzer, base_query: str) -> str:
 
 def generate_answer(question, temperature=0.4, top_p=0.85, max_tokens=600):
 
-    url = "https://0ae2-136-107-23-173.ngrok-free.app/generate"
+    analyzer = VocabularyAnalyzer()
 
-    payload = {
-        "prompt": question
-    }
+    system_prompt = load_swamiji_prompt(analyzer, question)
 
-    r = requests.post(url, json=payload, timeout=300)
+    full_prompt = f"<s>[INST] {system_prompt}\n\n{question} [/INST]"
 
-    if r.status_code == 200:
-        return r.json()["response"]
+    model = load_model()
 
-    return "Error contacting model server"
+    response = model(
+        full_prompt,
+        temperature=temperature,
+        top_p=top_p,
+        max_new_tokens=max_tokens,
+    )
+
+    filtered_response = analyzer.filter_response(response)
+
+    return filtered_response