Update app.py

app.py

@@ -23,6 +23,11 @@ MODEL = "nousresearch/deephermes-3-llama-3-8b-preview:free"
 
 app = FastAPI()
 
+# ✅ Greeting checker
+def is_greeting(text):
+    greetings = ["hi", "hello", "hey", "good morning", "good afternoon", "good evening"]
+    return any(g in text.lower() for g in greetings)
+
 # ✅ Context fetcher
 def get_context(query, top_k=5, threshold=0.3):
     query_vec = embed_model.encode([query])
@@ -43,16 +48,20 @@ def chat_fn(message, history):
 
     context = get_context(message)
 
+    # ✅ Allow fallback for greetings
     if not context.strip():
+        if is_greeting(message):
+            return "👋 Hello! How can I assist you today?"
         return "❌ I couldn’t find any relevant info to answer that. Please ask something else."
 
+    # ✅ Construct message payload
     messages = [
         {
             "role": "system",
             "content": (
-                "You are a precise, concise assistant. Use ONLY the provided context to answer."
-                " Do not make up facts. Answer briefly and clearly. If unsure, respond with:"
-                " 'I couldn’t find the answer based on available context.'\n\nContext:\n" + context
+                "You are a helpful, precise assistant. Use the provided context to answer when available."
+                " If no relevant context is found, reply politely or ask the user to rephrase."
+                "\n\nContext:\n" + context
             )
         }
     ]
@@ -66,7 +75,7 @@ def chat_fn(message, history):
     payload = {
         "model": MODEL,
         "messages": messages,
-        "max_tokens": 200,
+        "max_tokens": 300,
     }
 
     try:
@@ -74,7 +83,7 @@ def chat_fn(message, history):
             "https://openrouter.ai/api/v1/chat/completions",
             headers=headers,
             json=payload,
-            timeout=30  # Prevent hanging
+            timeout=30
         )
         response.raise_for_status()
         reply = response.json()["choices"][0]["message"]["content"]