Update src/agent.py
Browse files- src/agent.py +28 -179
src/agent.py
CHANGED
|
@@ -1,18 +1,30 @@
|
|
| 1 |
-
from fastapi import APIRouter, HTTPException, Request
|
|
|
|
| 2 |
from pydantic import BaseModel
|
| 3 |
-
from typing import List, Optional, Dict
|
| 4 |
import time
|
| 5 |
import os
|
| 6 |
import re
|
| 7 |
import json
|
|
|
|
| 8 |
from datetime import datetime
|
| 9 |
|
| 10 |
-
# Import
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 11 |
try:
|
| 12 |
-
from langchain_openai import ChatOpenAI
|
| 13 |
from langchain_anthropic import ChatAnthropic
|
|
|
|
| 14 |
except ImportError:
|
| 15 |
-
|
| 16 |
|
| 17 |
router = APIRouter()
|
| 18 |
|
|
@@ -33,180 +45,17 @@ class AgentResponse(BaseModel):
|
|
| 33 |
sources: List[dict] = []
|
| 34 |
model_used: str = "gpt"
|
| 35 |
response_time: float = 0
|
|
|
|
|
|
|
|
|
|
|
|
|
| 36 |
|
| 37 |
-
# ==========
|
| 38 |
-
STAFF_DIRECTORY = [
|
| 39 |
-
{
|
| 40 |
-
"full_name": "Dr. Abdulla Al Hefeiti",
|
| 41 |
-
"role": "Library Director / Assistant Provost, Libraries",
|
| 42 |
-
"details": "He can help with strategic and institutional matters, library leadership, and partnerships. Email: abdulla.alhefeiti@ku.ac.ae | Phone: +971 2 312 3331",
|
| 43 |
-
"tokens": ["abdulla", "hefeiti", "abdulla al hefeiti", "al hefeiti"],
|
| 44 |
-
},
|
| 45 |
-
{
|
| 46 |
-
"full_name": "Nikesh Narayanan",
|
| 47 |
-
"role": "Research & Access Services Librarian",
|
| 48 |
-
"details": "He can help with research support, Khazna institutional repository, ORCID, SciVal, Scopus, research impact, AI tools for research, bibliometrics, scholarly communication, and LibGuides. Email: nikesh.narayanan@ku.ac.ae | Phone: +971 2 312 3980",
|
| 49 |
-
"tokens": ["nikesh", "narayanan", "nikesh narayanan"],
|
| 50 |
-
},
|
| 51 |
-
{
|
| 52 |
-
"full_name": "Jason Fetty",
|
| 53 |
-
"role": "Medical Librarian",
|
| 54 |
-
"details": "He can help with medical and health sciences research, systematic reviews, PubMed, Embase, CINAHL, UpToDate, and clinical databases. Email: jason.fetty@ku.ac.ae | Phone: +971 2 312 4722",
|
| 55 |
-
"tokens": ["jason", "fetty", "jason fetty"],
|
| 56 |
-
},
|
| 57 |
-
{
|
| 58 |
-
"full_name": "Walter Brian Hall",
|
| 59 |
-
"role": "Digital & Technology Services Librarian / Systems Librarian",
|
| 60 |
-
"details": "He can help with library website, systems, technology, digital infrastructure, ORCID, STEAM, coding, Open Access publishing, APC funding, general library services, and public services. Email: walter.hall@ku.ac.ae | Phone: +971 2 312 3163",
|
| 61 |
-
"tokens": ["walter", "brian", "hall", "walter brian hall"],
|
| 62 |
-
},
|
| 63 |
-
]
|
| 64 |
|
| 65 |
-
#
|
| 66 |
-
SOCIAL_RE = re.compile(r"^(hi|hello|hey|good morning|good afternoon|good evening|how are you|who are you|what can you do|
|
| 67 |
PURE_GREETING_RE = re.compile(r"^(hi|hello|hey|good morning|good afternoon|good evening)[!.\s]*$", re.IGNORECASE)
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
|
| 72 |
-
_CLASSIFIER_SYSTEM = """You are the routing brain of LibBee...""" # [Truncated for brevity - copy from app81.py lines ~875-915]
|
| 73 |
-
|
| 74 |
-
# ========== HELPER FUNCTIONS ==========
|
| 75 |
-
def _normalize_name_query(text: str):
|
| 76 |
-
return [t for t in re.sub(r"[^a-z0-9 ]+", " ", (text or "").lower()).split() if t]
|
| 77 |
-
|
| 78 |
-
def _match_staff_name(question: str) -> Optional[Dict]:
|
| 79 |
-
tokens = _normalize_name_query(question)
|
| 80 |
-
if not tokens:
|
| 81 |
-
return None
|
| 82 |
-
for staff in STAFF_DIRECTORY:
|
| 83 |
-
staff_tokens = set()
|
| 84 |
-
for tok in staff.get("tokens", []):
|
| 85 |
-
staff_tokens.update(_normalize_name_query(tok))
|
| 86 |
-
if all(tok in staff_tokens for tok in tokens):
|
| 87 |
-
return staff
|
| 88 |
-
return None
|
| 89 |
-
|
| 90 |
-
def _staff_name_answer(staff: dict) -> str:
|
| 91 |
-
return f"Did you mean <strong>{staff['full_name']}</strong>?<br><br><strong>{staff['full_name']}</strong> is the <strong>{staff['role']}</strong>. {staff['details']}"
|
| 92 |
-
|
| 93 |
-
def _looks_library_hours_question(question: str) -> bool:
|
| 94 |
-
return bool(HOURS_RE.search(question.lower()))
|
| 95 |
-
|
| 96 |
-
def _hours_redirect_answer() -> str:
|
| 97 |
-
return """
|
| 98 |
-
Here are the current <strong>Khalifa University Library</strong> hours:<br><br>
|
| 99 |
-
<strong>Habshan Library (SAN Campus):</strong><br>
|
| 100 |
-
Sunday – Wednesday: 8:00 AM – 7:00 PM<br>
|
| 101 |
-
Thursday: 8:00 AM – 4:00 PM<br>
|
| 102 |
-
Friday: Closed<br>
|
| 103 |
-
Saturday: 12:00 PM – 8:00 PM<br><br>
|
| 104 |
-
⚠️ Check official page for latest times: <a href="https://library.ku.ac.ae/hours">library.ku.ac.ae/hours</a>
|
| 105 |
-
"""
|
| 106 |
-
|
| 107 |
-
async def _llm_classify(question: str, history: list, model: str = "gpt"):
|
| 108 |
-
"""Copy this function from app81.py lines ~918-960"""
|
| 109 |
-
# Simplified version for now:
|
| 110 |
-
q = question.lower()
|
| 111 |
-
if SOCIAL_RE.match(q):
|
| 112 |
-
return {"intent": "social", "casual_answer": "Hello! I'm LibBee, the KU Library AI Assistant."}
|
| 113 |
-
elif any(k in q for k in MEDICAL_KEYWORDS):
|
| 114 |
-
return {"intent": "search_medical"}
|
| 115 |
-
elif LIBRARY_CUE_RE.search(q):
|
| 116 |
-
return {"intent": "library_info"}
|
| 117 |
-
else:
|
| 118 |
-
return {"intent": "general"}
|
| 119 |
-
|
| 120 |
-
# ========== MAIN ENDPOINT ==========
|
| 121 |
-
@router.post("", response_model=AgentResponse)
|
| 122 |
-
async def agent_query(request: Request, req: AgentRequest):
|
| 123 |
-
start = time.time()
|
| 124 |
-
|
| 125 |
-
question = req.question.strip()
|
| 126 |
-
|
| 127 |
-
# 1. Instant handlers (from app81.py Layer 0)
|
| 128 |
-
|
| 129 |
-
# Staff name match
|
| 130 |
-
staff_match = _match_staff_name(question)
|
| 131 |
-
if staff_match:
|
| 132 |
-
return AgentResponse(
|
| 133 |
-
answer=_staff_name_answer(staff_match),
|
| 134 |
-
intent="library_info",
|
| 135 |
-
tools_used=["staff_name_match"],
|
| 136 |
-
model_used=req.model,
|
| 137 |
-
response_time=time.time() - start
|
| 138 |
-
)
|
| 139 |
-
|
| 140 |
-
# Hours question
|
| 141 |
-
if _looks_library_hours_question(question):
|
| 142 |
-
return AgentResponse(
|
| 143 |
-
answer=_hours_redirect_answer(),
|
| 144 |
-
intent="library_info",
|
| 145 |
-
tools_used=["hours_hardcoded"],
|
| 146 |
-
model_used=req.model,
|
| 147 |
-
response_time=time.time() - start
|
| 148 |
-
)
|
| 149 |
-
|
| 150 |
-
# Greeting
|
| 151 |
-
if PURE_GREETING_RE.match(question):
|
| 152 |
-
return AgentResponse(
|
| 153 |
-
answer="Hi! I'm <strong>LibBee</strong>, the Khalifa University Library AI Assistant. I can help you find articles and books, search databases, and answer library questions. What would you like help with?",
|
| 154 |
-
intent="social_greeting",
|
| 155 |
-
tools_used=[],
|
| 156 |
-
model_used=req.model,
|
| 157 |
-
response_time=time.time() - start
|
| 158 |
-
)
|
| 159 |
-
|
| 160 |
-
# 2. LLM Classification (Layer 1)
|
| 161 |
-
try:
|
| 162 |
-
classification = await _llm_classify(question, req.history, req.model)
|
| 163 |
-
intent = classification.get("intent", "general")
|
| 164 |
-
|
| 165 |
-
# 3. Route by intent
|
| 166 |
-
if intent == "social":
|
| 167 |
-
return AgentResponse(
|
| 168 |
-
answer=classification.get("casual_answer", "Hello! How can I help?"),
|
| 169 |
-
intent="social",
|
| 170 |
-
model_used=req.model,
|
| 171 |
-
response_time=time.time() - start
|
| 172 |
-
)
|
| 173 |
-
|
| 174 |
-
elif intent == "library_info":
|
| 175 |
-
# For now, return generic library info response
|
| 176 |
-
# Later: Add RAG here
|
| 177 |
-
return AgentResponse(
|
| 178 |
-
answer="I can help with library information. This is where your RAG logic will go.",
|
| 179 |
-
intent="library_info",
|
| 180 |
-
tools_used=["get_library_info"],
|
| 181 |
-
model_used=req.model,
|
| 182 |
-
response_time=time.time() - start
|
| 183 |
-
)
|
| 184 |
-
|
| 185 |
-
elif intent in ["search_academic", "search_medical"]:
|
| 186 |
-
return AgentResponse(
|
| 187 |
-
answer=f"I'll help you search for {'medical' if intent == 'search_medical' else 'academic'} literature on: {question}",
|
| 188 |
-
intent=intent,
|
| 189 |
-
tools_used=["search_primo"],
|
| 190 |
-
model_used=req.model,
|
| 191 |
-
response_time=time.time() - start
|
| 192 |
-
)
|
| 193 |
-
|
| 194 |
-
else:
|
| 195 |
-
return AgentResponse(
|
| 196 |
-
answer="I'm the KU Library AI Assistant. I can help you find articles, books, and answer library questions.",
|
| 197 |
-
intent="general",
|
| 198 |
-
model_used=req.model,
|
| 199 |
-
response_time=time.time() - start
|
| 200 |
-
)
|
| 201 |
-
|
| 202 |
-
except Exception as e:
|
| 203 |
-
return AgentResponse(
|
| 204 |
-
answer=f"Error processing request: {str(e)}",
|
| 205 |
-
intent="error",
|
| 206 |
-
model_used=req.model,
|
| 207 |
-
response_time=time.time() - start
|
| 208 |
-
)
|
| 209 |
-
|
| 210 |
-
@router.get("/test")
|
| 211 |
-
async def test():
|
| 212 |
-
return {"status": "ok", "message": "Agent is working!"}
|
|
|
|
| 1 |
+
from fastapi import APIRouter, HTTPException, Request, Response
|
| 2 |
+
from fastapi.responses import JSONResponse
|
| 3 |
from pydantic import BaseModel
|
| 4 |
+
from typing import List, Optional, Dict, Any
|
| 5 |
import time
|
| 6 |
import os
|
| 7 |
import re
|
| 8 |
import json
|
| 9 |
+
import secrets
|
| 10 |
from datetime import datetime
|
| 11 |
|
| 12 |
+
# Import services
|
| 13 |
+
from src.config import get_settings
|
| 14 |
+
from src.services.cache_service import CacheService
|
| 15 |
+
from src.services.staff_service import (
|
| 16 |
+
match_staff_name, match_staff_role, staff_name_answer, staff_role_answer,
|
| 17 |
+
build_staff_index, STAFF_DIRECTORY
|
| 18 |
+
)
|
| 19 |
+
from src.services.rag_service import RAGService
|
| 20 |
+
|
| 21 |
+
# Try to import LLM libraries
|
| 22 |
try:
|
| 23 |
+
from langchain_openai import ChatOpenAI, OpenAIEmbeddings
|
| 24 |
from langchain_anthropic import ChatAnthropic
|
| 25 |
+
LLM_AVAILABLE = True
|
| 26 |
except ImportError:
|
| 27 |
+
LLM_AVAILABLE = False
|
| 28 |
|
| 29 |
router = APIRouter()
|
| 30 |
|
|
|
|
| 45 |
sources: List[dict] = []
|
| 46 |
model_used: str = "gpt"
|
| 47 |
response_time: float = 0
|
| 48 |
+
corrected_query: Optional[str] = None
|
| 49 |
+
natural_query: Optional[str] = None
|
| 50 |
+
database_query: Optional[str] = None
|
| 51 |
+
ask_librarian: Optional[bool] = None
|
| 52 |
|
| 53 |
+
# ========== CONSTANTS FROM APP81.PY ==========
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 54 |
|
| 55 |
+
# Regex patterns
|
| 56 |
+
SOCIAL_RE = re.compile(r"^(hi|hello|hey|good morning|good afternoon|good evening|how are you|hows it going|how are you doing|who are you|what are you|tell me about yourself|what can you do|what do you do|how can you help|what help can you provide|are you a bot|are you a robot|are you ai|are you an ai|thank you|thanks|thx|bye|goodbye|see you|lol|haha|you are silly|are you silly|are you dumb|stupid bot|dumb bot|idiot bot|joke|tell me a joke)\b", re.IGNORECASE)
|
| 57 |
PURE_GREETING_RE = re.compile(r"^(hi|hello|hey|good morning|good afternoon|good evening)[!.\s]*$", re.IGNORECASE)
|
| 58 |
+
GREETING_FOLLOWUP_RE = re.compile(r"^(yes|yes please|yes check|sure|okay|ok|yeah|yep|please do|go ahead)$", re.IGNORECASE)
|
| 59 |
+
HOURS_RE = re.compile(r"\b(library hours?|opening hours?|closing hours?|open hours?|open today|open tomorrow|open now|open on|open this|closed today|closed tomorrow|closed on|closed this|when.*open|when.*close|when.*does.*library|what time.*open|what time.*close|what time.*library|opening time|closing time|is.*library.*open|is.*campus.*library|is.*habshan.*open|is.*san.*open|is.*library.*closed|will.*library.*open|will.*it.*open|does.*library.*open|library.*open.*today|library.*close.*today|habshan.*open|san.*open|main campus.*open|ramadan hours?|weekend hours?|friday hours?|saturday hours?|sunday hours?|library.*schedule|schedule.*library|what.*hours|till.*when|until.*when|what day.*open|which day.*open|what day.*closed|which day.*closed)\b", re.IGNORECASE)
|
| 60 |
+
KU_CAMPUS_RE = re.compile(r"\b(where is the library|where is.*library|library.*location|location.*library|which building|what building|which floor|main campus library|san campus library|habshan library|sas al nakhl library|library.*address|address.*library|how to get to.*library|how.*reach.*library|how.*find.*library|library.*campus(es)?|campus(es)?.*library|how many.*librar|how many.*branch|library.*branch(es)?|branch(es)?.*library|which campus.*library|library.*which campus|library.*map|map.*library|how many libraries|what libraries|which libraries|(ku|khalifa).*(has|have|got).*(librar|branch)|(librar|branch).*(ku|khalifa).*(has|have)|(all|both|two|2).*(libraries|library|branches)|tell me about.*(campus|libraries|branches)|describe.*(campus|libraries)|(what|where).*(campus|campuses).*(librar))\b", re.IGNORECASE)
|
| 61 |
+
RESEARCH_CUE_RE = re.compile(r"\b(find|search|look for|show me|get me|give me|locate|recommend|suggest|articles?|papers?|books?|journals?|studies|literature|research|systematic review|evidence|sources|database|databases|peer reviewed|open access|advances?|approach(es)?|method(s|ology)?|*
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|