Spaces:
Runtime error
Runtime error
Update rag_faiss.py
Browse files- rag_faiss.py +92 -50
rag_faiss.py
CHANGED
|
@@ -27,22 +27,83 @@ print("🔹 Loading embedding model...")
|
|
| 27 |
embedder = SentenceTransformer(EMBED_MODEL)
|
| 28 |
|
| 29 |
# -------------------------------
|
| 30 |
-
# SMALL TALK
|
| 31 |
# -------------------------------
|
| 32 |
-
|
| 33 |
-
"
|
| 34 |
-
|
| 35 |
-
|
| 36 |
-
|
| 37 |
-
|
| 38 |
-
|
| 39 |
-
|
| 40 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 41 |
}
|
| 42 |
|
| 43 |
-
def
|
| 44 |
q = query.lower().strip()
|
| 45 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 46 |
|
| 47 |
# -------------------------------
|
| 48 |
# VECTOR STORE
|
|
@@ -52,12 +113,12 @@ class VectorStore:
|
|
| 52 |
print(f"🔹 Loading FAISS index from {INDEX_FILE}")
|
| 53 |
|
| 54 |
if not os.path.exists(INDEX_FILE):
|
| 55 |
-
raise FileNotFoundError("❌
|
| 56 |
-
|
| 57 |
-
self.index = faiss.read_index(INDEX_FILE)
|
| 58 |
|
| 59 |
if not os.path.exists(META_FILE):
|
| 60 |
-
raise FileNotFoundError("❌ meta.pkl not found
|
|
|
|
|
|
|
| 61 |
|
| 62 |
with open(META_FILE, "rb") as f:
|
| 63 |
self.texts = pickle.load(f)
|
|
@@ -103,32 +164,25 @@ def call_openrouter(prompt, max_tokens=300):
|
|
| 103 |
"max_tokens": max_tokens
|
| 104 |
}
|
| 105 |
|
| 106 |
-
|
| 107 |
-
|
| 108 |
-
|
| 109 |
-
|
| 110 |
-
|
| 111 |
-
|
| 112 |
-
|
| 113 |
-
|
| 114 |
-
return r.json()["choices"][0]["message"]["content"]
|
| 115 |
-
|
| 116 |
-
except Exception as e:
|
| 117 |
-
return f"❌ Error contacting OpenRouter: {e}"
|
| 118 |
|
| 119 |
# -------------------------------
|
| 120 |
-
# MAIN
|
| 121 |
# -------------------------------
|
| 122 |
-
def answer_question(vs, question):
|
| 123 |
-
|
| 124 |
-
|
| 125 |
-
|
| 126 |
-
|
| 127 |
-
for key in SMALL_TALK_RESPONSES:
|
| 128 |
-
if key in q:
|
| 129 |
-
return SMALL_TALK_RESPONSES[key]
|
| 130 |
|
| 131 |
-
# 2️⃣ Knowledge-based
|
| 132 |
contexts = vs.search(question, k=4)
|
| 133 |
|
| 134 |
if not contexts:
|
|
@@ -154,15 +208,3 @@ QUESTION: {question}
|
|
| 154 |
ANSWER:
|
| 155 |
"""
|
| 156 |
return call_openrouter(prompt)
|
| 157 |
-
|
| 158 |
-
# -------------------------------
|
| 159 |
-
# EXAMPLE USAGE (for local test)
|
| 160 |
-
# -------------------------------
|
| 161 |
-
if __name__ == "__main__":
|
| 162 |
-
vs = VectorStore()
|
| 163 |
-
|
| 164 |
-
while True:
|
| 165 |
-
user_q = input("\nYou: ")
|
| 166 |
-
if user_q.lower() in ["exit", "quit"]:
|
| 167 |
-
break
|
| 168 |
-
print("Bot:", answer_question(vs, user_q))
|
|
|
|
| 27 |
embedder = SentenceTransformer(EMBED_MODEL)
|
| 28 |
|
| 29 |
# -------------------------------
|
| 30 |
+
# SMALL TALK / IDENTITY INTENTS
|
| 31 |
# -------------------------------
|
| 32 |
+
SMALL_TALK_INTENTS = {
|
| 33 |
+
"greeting": {
|
| 34 |
+
"keywords": [
|
| 35 |
+
"hello", "hi", "hey", "hii", "hai",
|
| 36 |
+
"good morning", "good evening", "good afternoon"
|
| 37 |
+
],
|
| 38 |
+
"answer": "Hello! I’m the official virtual assistant for GUESSS India. How can I help you today?"
|
| 39 |
+
},
|
| 40 |
+
"wellbeing": {
|
| 41 |
+
"keywords": [
|
| 42 |
+
"how are you", "how r you", "how are u",
|
| 43 |
+
"how do you do", "are you okay"
|
| 44 |
+
],
|
| 45 |
+
"answer": "I’m doing well, thank you! How can I assist you with GUESSS India today?"
|
| 46 |
+
},
|
| 47 |
+
"identity": {
|
| 48 |
+
"keywords": [
|
| 49 |
+
"who are you", "what are you",
|
| 50 |
+
"tell me about yourself", "introduce yourself"
|
| 51 |
+
],
|
| 52 |
+
"answer": "I’m a virtual assistant created to provide accurate and reliable information about GUESSS India."
|
| 53 |
+
},
|
| 54 |
+
"name": {
|
| 55 |
+
"keywords": [
|
| 56 |
+
"what is your name", "your name",
|
| 57 |
+
"do you have a name", "who should i call you"
|
| 58 |
+
],
|
| 59 |
+
"answer": "I don’t have a personal name, but you can think of me as the GUESSS India Assistant."
|
| 60 |
+
},
|
| 61 |
+
"capabilities": {
|
| 62 |
+
"keywords": [
|
| 63 |
+
"what do you do", "what can you do",
|
| 64 |
+
"how can you help", "what help can you provide"
|
| 65 |
+
],
|
| 66 |
+
"answer": (
|
| 67 |
+
"I answer questions related to GUESSS India, including surveys, programs, "
|
| 68 |
+
"campus ambassadors, podcasts, and general information."
|
| 69 |
+
)
|
| 70 |
+
},
|
| 71 |
+
"trust": {
|
| 72 |
+
"keywords": [
|
| 73 |
+
"can i trust you", "is your information reliable",
|
| 74 |
+
"are you reliable", "is this official"
|
| 75 |
+
],
|
| 76 |
+
"answer": "Yes. My responses are based on approved and verified information provided by the GUESSS India team."
|
| 77 |
+
},
|
| 78 |
+
"human": {
|
| 79 |
+
"keywords": [
|
| 80 |
+
"are you human", "are you a real person",
|
| 81 |
+
"are you ai", "are you a bot"
|
| 82 |
+
],
|
| 83 |
+
"answer": "No. I’m an AI-based assistant designed to share verified information from GUESSS India."
|
| 84 |
+
},
|
| 85 |
+
"availability": {
|
| 86 |
+
"keywords": [
|
| 87 |
+
"are you available", "are you always available",
|
| 88 |
+
"when are you available"
|
| 89 |
+
],
|
| 90 |
+
"answer": "Yes. I’m available 24/7 on the GUESSS India website."
|
| 91 |
+
},
|
| 92 |
+
"thanks": {
|
| 93 |
+
"keywords": [
|
| 94 |
+
"thank you", "thanks", "thank u", "thx"
|
| 95 |
+
],
|
| 96 |
+
"answer": "You’re welcome! If you have more questions about GUESSS India, I’m here to help."
|
| 97 |
+
}
|
| 98 |
}
|
| 99 |
|
| 100 |
+
def handle_small_talk(query: str):
|
| 101 |
q = query.lower().strip()
|
| 102 |
+
for intent in SMALL_TALK_INTENTS.values():
|
| 103 |
+
for kw in intent["keywords"]:
|
| 104 |
+
if kw in q:
|
| 105 |
+
return intent["answer"]
|
| 106 |
+
return None
|
| 107 |
|
| 108 |
# -------------------------------
|
| 109 |
# VECTOR STORE
|
|
|
|
| 113 |
print(f"🔹 Loading FAISS index from {INDEX_FILE}")
|
| 114 |
|
| 115 |
if not os.path.exists(INDEX_FILE):
|
| 116 |
+
raise FileNotFoundError("❌ index.faiss not found")
|
|
|
|
|
|
|
| 117 |
|
| 118 |
if not os.path.exists(META_FILE):
|
| 119 |
+
raise FileNotFoundError("❌ meta.pkl not found")
|
| 120 |
+
|
| 121 |
+
self.index = faiss.read_index(INDEX_FILE)
|
| 122 |
|
| 123 |
with open(META_FILE, "rb") as f:
|
| 124 |
self.texts = pickle.load(f)
|
|
|
|
| 164 |
"max_tokens": max_tokens
|
| 165 |
}
|
| 166 |
|
| 167 |
+
r = requests.post(
|
| 168 |
+
OPENROUTER_URL,
|
| 169 |
+
headers=headers,
|
| 170 |
+
json=payload,
|
| 171 |
+
timeout=60
|
| 172 |
+
)
|
| 173 |
+
r.raise_for_status()
|
| 174 |
+
return r.json()["choices"][0]["message"]["content"]
|
|
|
|
|
|
|
|
|
|
|
|
|
| 175 |
|
| 176 |
# -------------------------------
|
| 177 |
+
# MAIN ANSWER FUNCTION
|
| 178 |
# -------------------------------
|
| 179 |
+
def answer_question(vs, question: str):
|
| 180 |
+
# 1️⃣ Small talk / identity
|
| 181 |
+
small_talk = handle_small_talk(question)
|
| 182 |
+
if small_talk:
|
| 183 |
+
return small_talk
|
|
|
|
|
|
|
|
|
|
| 184 |
|
| 185 |
+
# 2️⃣ Knowledge-based (RAG)
|
| 186 |
contexts = vs.search(question, k=4)
|
| 187 |
|
| 188 |
if not contexts:
|
|
|
|
| 208 |
ANSWER:
|
| 209 |
"""
|
| 210 |
return call_openrouter(prompt)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|