Update src/agent.py

src/agent.py

@@ -1,8 +1,18 @@
 from fastapi import APIRouter, HTTPException, Request
 from pydantic import BaseModel
-from typing import List, Optional
+from typing import List, Optional, Dict
 import time
 import os
+import re
+import json
+from datetime import datetime
+
+# Import OpenAI for classification
+try:
+    from langchain_openai import ChatOpenAI
+    from langchain_anthropic import ChatAnthropic
+except ImportError:
+    pass
 
 router = APIRouter()
 
@@ -18,32 +28,185 @@ class AgentRequest(BaseModel):
 class AgentResponse(BaseModel):
     answer: str
     intent: str = "general"
-    tools_used: list = []
-    response_time: float = 0
+    tools_used: List[str] = []
+    search_results: List[dict] = []
+    sources: List[dict] = []
     model_used: str = "gpt"
+    response_time: float = 0
+
+# ========== STAFF DIRECTORY DATA (from app81.py) ==========
+STAFF_DIRECTORY = [
+    {
+        "full_name": "Dr. Abdulla Al Hefeiti",
+        "role": "Library Director / Assistant Provost, Libraries",
+        "details": "He can help with strategic and institutional matters, library leadership, and partnerships. Email: abdulla.alhefeiti@ku.ac.ae | Phone: +971 2 312 3331",
+        "tokens": ["abdulla", "hefeiti", "abdulla al hefeiti", "al hefeiti"],
+    },
+    {
+        "full_name": "Nikesh Narayanan",
+        "role": "Research & Access Services Librarian",
+        "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",
+        "tokens": ["nikesh", "narayanan", "nikesh narayanan"],
+    },
+    {
+        "full_name": "Jason Fetty",
+        "role": "Medical Librarian",
+        "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",
+        "tokens": ["jason", "fetty", "jason fetty"],
+    },
+    {
+        "full_name": "Walter Brian Hall",
+        "role": "Digital & Technology Services Librarian / Systems Librarian",
+        "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",
+        "tokens": ["walter", "brian", "hall", "walter brian hall"],
+    },
+]
+
+# ========== REGEX PATTERNS (from app81.py) ==========
+SOCIAL_RE = re.compile(r"^(hi|hello|hey|good morning|good afternoon|good evening|how are you|who are you|what can you do|thanks|thank you|bye|goodbye)\b", re.IGNORECASE)
+PURE_GREETING_RE = re.compile(r"^(hi|hello|hey|good morning|good afternoon|good evening)[!.\s]*$", re.IGNORECASE)
+HOURS_RE = re.compile(r"\b(library hours?|opening hours?|when.*open|when.*close|open today|closed today)\b", re.IGNORECASE)
+LIBRARY_CUE_RE = re.compile(r"\b(library|librarian|borrow|primo|database|article|book|ill|interlibrary|full text)\b", re.IGNORECASE)
+MEDICAL_KEYWORDS = ['medicine','medical','health','clinical','nursing','pharmacy','pubmed','embase']
+
+_CLASSIFIER_SYSTEM = """You are the routing brain of LibBee...""" # [Truncated for brevity - copy from app81.py lines ~875-915]
+
+# ========== HELPER FUNCTIONS ==========
+def _normalize_name_query(text: str):
+    return [t for t in re.sub(r"[^a-z0-9 ]+", " ", (text or "").lower()).split() if t]
+
+def _match_staff_name(question: str) -> Optional[Dict]:
+    tokens = _normalize_name_query(question)
+    if not tokens:
+        return None
+    for staff in STAFF_DIRECTORY:
+        staff_tokens = set()
+        for tok in staff.get("tokens", []):
+            staff_tokens.update(_normalize_name_query(tok))
+        if all(tok in staff_tokens for tok in tokens):
+            return staff
+    return None
+
+def _staff_name_answer(staff: dict) -> str:
+    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']}"
+
+def _looks_library_hours_question(question: str) -> bool:
+    return bool(HOURS_RE.search(question.lower()))
 
-# Temporary: Simple echo to test connectivity
+def _hours_redirect_answer() -> str:
+    return """
+    Here are the current <strong>Khalifa University Library</strong> hours:<br><br>
+    <strong>Habshan Library (SAN Campus):</strong><br>
+    Sunday – Wednesday: 8:00 AM – 7:00 PM<br>
+    Thursday: 8:00 AM – 4:00 PM<br>
+    Friday: Closed<br>
+    Saturday: 12:00 PM – 8:00 PM<br><br>
+    ⚠️ Check official page for latest times: <a href="https://library.ku.ac.ae/hours">library.ku.ac.ae/hours</a>
+    """
+
+async def _llm_classify(question: str, history: list, model: str = "gpt"):
+    """Copy this function from app81.py lines ~918-960"""
+    # Simplified version for now:
+    q = question.lower()
+    if SOCIAL_RE.match(q):
+        return {"intent": "social", "casual_answer": "Hello! I'm LibBee, the KU Library AI Assistant."}
+    elif any(k in q for k in MEDICAL_KEYWORDS):
+        return {"intent": "search_medical"}
+    elif LIBRARY_CUE_RE.search(q):
+        return {"intent": "library_info"}
+    else:
+        return {"intent": "general"}
+
+# ========== MAIN ENDPOINT ==========
 @router.post("", response_model=AgentResponse)
 async def agent_query(request: Request, req: AgentRequest):
-    """
-    Main agent endpoint.
-    For now, it echoes back. You'll migrate your app81.py logic here.
-    """
     start = time.time()
     
-    # TODO: Migrate your classification logic from app81.py here
-    # For now, just return a welcome message
+    question = req.question.strip()
     
-    response_time = time.time() - start
+    # 1. Instant handlers (from app81.py Layer 0)
     
-    return AgentResponse(
-        answer=f"Hello! You asked: '{req.question}'. I'm ready to help once you migrate the full logic from app81.py.",
-        intent="test",
-        response_time=response_time,
-        model_used=req.model
-    )
-
-@router.post("/test")
-async def test_endpoint():
-    """Quick test that doesn't require complex logic"""
-    return {"status": "ok", "message": "Agent endpoint is working!"}
+    # Staff name match
+    staff_match = _match_staff_name(question)
+    if staff_match:
+        return AgentResponse(
+            answer=_staff_name_answer(staff_match),
+            intent="library_info",
+            tools_used=["staff_name_match"],
+            model_used=req.model,
+            response_time=time.time() - start
+        )
+    
+    # Hours question
+    if _looks_library_hours_question(question):
+        return AgentResponse(
+            answer=_hours_redirect_answer(),
+            intent="library_info",
+            tools_used=["hours_hardcoded"],
+            model_used=req.model,
+            response_time=time.time() - start
+        )
+    
+    # Greeting
+    if PURE_GREETING_RE.match(question):
+        return AgentResponse(
+            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?",
+            intent="social_greeting",
+            tools_used=[],
+            model_used=req.model,
+            response_time=time.time() - start
+        )
+    
+    # 2. LLM Classification (Layer 1)
+    try:
+        classification = await _llm_classify(question, req.history, req.model)
+        intent = classification.get("intent", "general")
+        
+        # 3. Route by intent
+        if intent == "social":
+            return AgentResponse(
+                answer=classification.get("casual_answer", "Hello! How can I help?"),
+                intent="social",
+                model_used=req.model,
+                response_time=time.time() - start
+            )
+        
+        elif intent == "library_info":
+            # For now, return generic library info response
+            # Later: Add RAG here
+            return AgentResponse(
+                answer="I can help with library information. This is where your RAG logic will go.",
+                intent="library_info",
+                tools_used=["get_library_info"],
+                model_used=req.model,
+                response_time=time.time() - start
+            )
+            
+        elif intent in ["search_academic", "search_medical"]:
+            return AgentResponse(
+                answer=f"I'll help you search for {'medical' if intent == 'search_medical' else 'academic'} literature on: {question}",
+                intent=intent,
+                tools_used=["search_primo"],
+                model_used=req.model,
+                response_time=time.time() - start
+            )
+        
+        else:
+            return AgentResponse(
+                answer="I'm the KU Library AI Assistant. I can help you find articles, books, and answer library questions.",
+                intent="general",
+                model_used=req.model,
+                response_time=time.time() - start
+            )
+            
+    except Exception as e:
+        return AgentResponse(
+            answer=f"Error processing request: {str(e)}",
+            intent="error",
+            model_used=req.model,
+            response_time=time.time() - start
+        )
+
+@router.get("/test")
+async def test():
+    return {"status": "ok", "message": "Agent is working!"}