Spaces:
Running
Running
Update rag.py
Browse files
rag.py
CHANGED
@@ -24,6 +24,13 @@ similarity_model = SentenceTransformer('paraphrase-MiniLM-L6-v2')
|
|
24 |
HF_DATASET_REPO = "midrees2806/unmatched_queries" # Your dataset repo
|
25 |
HF_TOKEN = os.getenv("HF_TOKEN") # From Space secrets
|
26 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
27 |
# --- Dataset Loading ---
|
28 |
try:
|
29 |
with open('dataset.json', 'r') as f:
|
@@ -64,7 +71,6 @@ def manage_unmatched_queries(query: str):
|
|
64 |
print(f"Failed to save query: {e}")
|
65 |
|
66 |
# --- Enhanced LLM Query ---
|
67 |
-
|
68 |
def query_groq_llm(prompt, model_name="llama3-70b-8192"):
|
69 |
try:
|
70 |
chat_completion = groq_client.chat.completions.create(
|
@@ -81,14 +87,44 @@ def query_groq_llm(prompt, model_name="llama3-70b-8192"):
|
|
81 |
print(f"Error querying Groq API: {e}")
|
82 |
return ""
|
83 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
84 |
def get_best_answer(user_input):
|
|
|
|
|
|
|
|
|
85 |
user_input_lower = user_input.lower().strip()
|
86 |
-
|
87 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
88 |
if any(keyword in user_input_lower for keyword in ["fee structure", "fees structure", "semester fees", "semester fee"]):
|
89 |
return (
|
90 |
"π° For complete and up-to-date fee details for this program, we recommend visiting the official University of Education fee structure page.\n"
|
91 |
-
"You
|
92 |
"π https://ue.edu.pk/allfeestructure.php"
|
93 |
)
|
94 |
|
@@ -98,7 +134,7 @@ def get_best_answer(user_input):
|
|
98 |
best_match_idx = similarities.argmax().item()
|
99 |
best_score = similarities[best_match_idx].item()
|
100 |
|
101 |
-
#
|
102 |
if best_score < 0.65:
|
103 |
manage_unmatched_queries(user_input)
|
104 |
|
@@ -130,4 +166,8 @@ def get_best_answer(user_input):
|
|
130 |
βοΈ [email protected]
|
131 |
π ue.edu.pk"""
|
132 |
|
133 |
-
|
|
|
|
|
|
|
|
|
|
24 |
HF_DATASET_REPO = "midrees2806/unmatched_queries" # Your dataset repo
|
25 |
HF_TOKEN = os.getenv("HF_TOKEN") # From Space secrets
|
26 |
|
27 |
+
# Greeting words list
|
28 |
+
GREETINGS = [
|
29 |
+
"hi", "hello", "hey", "good morning", "good afternoon", "good evening",
|
30 |
+
"assalam o alaikum", "salam", "namaste", "hola", "bonjour", "hi there",
|
31 |
+
"hey there", "greetings", "howdy"
|
32 |
+
]
|
33 |
+
|
34 |
# --- Dataset Loading ---
|
35 |
try:
|
36 |
with open('dataset.json', 'r') as f:
|
|
|
71 |
print(f"Failed to save query: {e}")
|
72 |
|
73 |
# --- Enhanced LLM Query ---
|
|
|
74 |
def query_groq_llm(prompt, model_name="llama3-70b-8192"):
|
75 |
try:
|
76 |
chat_completion = groq_client.chat.completions.create(
|
|
|
87 |
print(f"Error querying Groq API: {e}")
|
88 |
return ""
|
89 |
|
90 |
+
def handle_submit():
|
91 |
+
user_input = input_field.value.strip()
|
92 |
+
|
93 |
+
if not user_input:
|
94 |
+
show_message("Please enter a question")
|
95 |
+
return
|
96 |
+
|
97 |
+
response = get_best_answer(user_input)
|
98 |
+
|
99 |
+
if response.get('should_scroll', False):
|
100 |
+
scroll_to_answer()
|
101 |
+
|
102 |
+
display_response(response.get('response', ''))
|
103 |
+
|
104 |
def get_best_answer(user_input):
|
105 |
+
# 1. Check for empty input
|
106 |
+
if not user_input.strip():
|
107 |
+
return None # This will be handled in the frontend to prevent submission
|
108 |
+
|
109 |
user_input_lower = user_input.lower().strip()
|
110 |
+
|
111 |
+
# 2. Check for minimum word count (3 words)
|
112 |
+
if len(user_input_lower.split()) < 3 and not any(greet in user_input_lower for greet in GREETINGS):
|
113 |
+
return "Please ask your question properly with at least 3 words."
|
114 |
+
|
115 |
+
# 3. Handle greetings (regardless of word count)
|
116 |
+
if any(greet in user_input_lower for greet in GREETINGS):
|
117 |
+
greeting_response = query_groq_llm(
|
118 |
+
f"You are an official assistant for University of Education Lahore. "
|
119 |
+
f"Respond to this greeting in a friendly and professional manner: {user_input}"
|
120 |
+
)
|
121 |
+
return greeting_response if greeting_response else "Hello! How can I assist you today?"
|
122 |
+
|
123 |
+
# 4. Check if question is about fee
|
124 |
if any(keyword in user_input_lower for keyword in ["fee structure", "fees structure", "semester fees", "semester fee"]):
|
125 |
return (
|
126 |
"π° For complete and up-to-date fee details for this program, we recommend visiting the official University of Education fee structure page.\n"
|
127 |
+
"You'll find comprehensive information regarding tuition, admission charges, and other applicable fees there.\n"
|
128 |
"π https://ue.edu.pk/allfeestructure.php"
|
129 |
)
|
130 |
|
|
|
134 |
best_match_idx = similarities.argmax().item()
|
135 |
best_score = similarities[best_match_idx].item()
|
136 |
|
137 |
+
# Save unmatched queries (threshold = 0.65)
|
138 |
if best_score < 0.65:
|
139 |
manage_unmatched_queries(user_input)
|
140 |
|
|
|
166 |
βοΈ [email protected]
|
167 |
π ue.edu.pk"""
|
168 |
|
169 |
+
# Return the response along with a flag to indicate auto-scrolling should happen
|
170 |
+
return {
|
171 |
+
"response": response,
|
172 |
+
"should_scroll": True # Frontend should use this to trigger auto-scrolling
|
173 |
+
}
|