tests/test_atlas_functionality.py

#!/usr/bin/env python3
"""
Atlas Functionality Test Suite
Tests the chat endpoint and search decision waterfall logic
"""

import asyncio
import json
import requests
from typing import Dict, Any, Optional, List
import time

# Test configuration
BASE_URL = "http://localhost:7860"
CHAT_ENDPOINT = f"{BASE_URL}/chat"
HEALTH_ENDPOINT = f"{BASE_URL}/"

class AtlasTestSuite:
    def __init__(self):
        self.results = []
        self.session_id = None
        
    def log_result(self, test_name: str, success: bool, details: str = "", response_data: Dict = None):
        """Log test result"""
        result = {
            "test": test_name,
            "success": success,
            "details": details,
            "timestamp": time.time()
        }
        if response_data:
            result["response_data"] = response_data
        self.results.append(result)
        
        status = "✅ PASS" if success else "❌ FAIL"
        print(f"{status} {test_name}: {details}")
        
    def test_health_endpoint(self) -> bool:
        """Test the health endpoint"""
        try:
            response = requests.get(HEALTH_ENDPOINT, timeout=10)
            if response.status_code == 200:
                data = response.json()
                self.log_result("Health Check", True, "Server responding correctly", data)
                return True
            else:
                self.log_result("Health Check", False, f"Status code: {response.status_code}")
                return False
        except Exception as e:
            self.log_result("Health Check", False, f"Exception: {str(e)}")
            return False
    
    def test_anonymous_chat_no_search(self) -> bool:
        """Test anonymous chat request without search"""
        try:
            payload = {
                "prompt": "Hello, how are you?",
                "max_new_tokens": 100,
                "use_search": False,
                "temperature": 0.7
            }
            
            response = requests.post(CHAT_ENDPOINT, json=payload, timeout=30)
            
            if response.status_code == 200:
                data = response.json()
                # Check response structure
                if "response" in data and len(data["response"]) > 0:
                    # Store session ID for follow-up tests
                    self.session_id = response.headers.get("X-Session-ID")
                    self.log_result("Anonymous Chat (No Search)", True, 
                                  f"Response length: {len(data['response'])} chars, Session: {self.session_id}", 
                                  data)
                    return True
                else:
                    self.log_result("Anonymous Chat (No Search)", False, "Empty or missing response")
                    return False
            else:
                self.log_result("Anonymous Chat (No Search)", False, f"Status: {response.status_code}")
                return False
                
        except Exception as e:
            self.log_result("Anonymous Chat (No Search)", False, f"Exception: {str(e)}")
            return False
    
    def test_anonymous_chat_with_search(self) -> bool:
        """Test anonymous chat request with search enabled"""
        try:
            payload = {
                "prompt": "What is artificial intelligence?",
                "max_new_tokens": 150,
                "use_search": True,
                "temperature": 0.7
            }
            
            response = requests.post(CHAT_ENDPOINT, json=payload, timeout=45)
            
            if response.status_code == 200:
                data = response.json()
                # Check response structure
                has_response = "response" in data and len(data["response"]) > 0
                has_search_decision = "search_decision" in data
                has_cache_info = "cache_info" in data
                
                details = f"Response: {has_response}, Decision: {has_search_decision}, Cache: {has_cache_info}"
                
                if has_response:
                    self.log_result("Anonymous Chat (With Search)", True, details, data)
                    return True
                else:
                    self.log_result("Anonymous Chat (With Search)", False, "Missing response")
                    return False
            else:
                self.log_result("Anonymous Chat (With Search)", False, f"Status: {response.status_code}")
                return False
                
        except Exception as e:
            self.log_result("Anonymous Chat (With Search)", False, f"Exception: {str(e)}")
            return False
    
    def test_authenticated_chat(self) -> bool:
        """Test authenticated chat request"""
        try:
            payload = {
                "prompt": "Hello, I'm a test user. Can you help me?",
                "max_new_tokens": 100,
                "use_search": False,
                "temperature": 0.7,
                "user_id": "test_user_001"
            }
            
            response = requests.post(CHAT_ENDPOINT, json=payload, timeout=30)
            
            if response.status_code == 200:
                data = response.json()
                if "response" in data and len(data["response"]) > 0:
                    self.log_result("Authenticated Chat", True, 
                                  f"Response length: {len(data['response'])} chars", data)
                    return True
                else:
                    self.log_result("Authenticated Chat", False, "Empty or missing response")
                    return False
            else:
                self.log_result("Authenticated Chat", False, f"Status: {response.status_code}")
                return False
                
        except Exception as e:
            self.log_result("Authenticated Chat", False, f"Exception: {str(e)}")
            return False
    
    def test_conversation_follow_up(self) -> bool:
        """Test conversation follow-up to check search decision waterfall"""
        if not self.session_id:
            self.log_result("Conversation Follow-up", False, "No session ID available")
            return False
            
        try:
            # First establish conversation history
            payload1 = {
                "prompt": "What is machine learning?",
                "max_new_tokens": 100,
                "use_search": True,
                "temperature": 0.7
            }
            
            headers = {"X-Session-ID": self.session_id}
            response1 = requests.post(CHAT_ENDPOINT, json=payload1, headers=headers, timeout=45)
            
            if response1.status_code != 200:
                self.log_result("Conversation Follow-up", False, f"First request failed: {response1.status_code}")
                return False
            
            # Now test follow-up question (should trigger different search logic)
            payload2 = {
                "prompt": "Can you tell me more about that?",
                "max_new_tokens": 100,
                "use_search": True,
                "temperature": 0.7,
                "history": [
                    {"role": "user", "content": "What is machine learning?"},
                    {"role": "assistant", "content": response1.json()["response"][:200]}
                ]
            }
            
            response2 = requests.post(CHAT_ENDPOINT, json=payload2, headers=headers, timeout=45)
            
            if response2.status_code == 200:
                data = response2.json()
                has_search_decision = "search_decision" in data
                decision_info = ""
                
                if has_search_decision:
                    decision = data["search_decision"]
                    decision_info = f"Should search: {decision.get('should_search')}, " \
                                  f"Confidence: {decision.get('confidence')}, " \
                                  f"Flow: {decision.get('flow_type', 'unknown')}"
                
                self.log_result("Conversation Follow-up", True, 
                              f"Follow-up successful. {decision_info}", data)
                return True
            else:
                self.log_result("Conversation Follow-up", False, f"Status: {response2.status_code}")
                return False
                
        except Exception as e:
            self.log_result("Conversation Follow-up", False, f"Exception: {str(e)}")
            return False
    
    def test_search_decision_patterns(self) -> bool:
        """Test different search decision patterns"""
        test_cases = [
            {
                "name": "Elaboration Request",
                "prompt": "Tell me more about that",
                "history": [{"role": "assistant", "content": "AI is a field of computer science."}],
                "expected_search": False
            },
            {
                "name": "New Information Request", 
                "prompt": "What is the latest news about AI?",
                "history": [],
                "expected_search": True
            },
            {
                "name": "Clarification Request",
                "prompt": "What do you mean by that?",
                "history": [{"role": "assistant", "content": "Machine learning uses algorithms."}],
                "expected_search": False
            }
        ]
        
        success_count = 0
        
        for case in test_cases:
            try:
                payload = {
                    "prompt": case["prompt"],
                    "max_new_tokens": 50,
                    "use_search": True,
                    "temperature": 0.7,
                    "history": case.get("history", [])
                }
                
                response = requests.post(CHAT_ENDPOINT, json=payload, timeout=30)
                
                if response.status_code == 200:
                    data = response.json()
                    if "search_decision" in data:
                        actual_search = data["search_decision"].get("should_search")
                        expected = case["expected_search"]
                        
                        if actual_search == expected:
                            self.log_result(f"Search Pattern: {case['name']}", True, 
                                          f"Correctly decided {'to search' if actual_search else 'not to search'}")
                            success_count += 1
                        else:
                            self.log_result(f"Search Pattern: {case['name']}", False, 
                                          f"Expected {expected}, got {actual_search}")
                    else:
                        self.log_result(f"Search Pattern: {case['name']}", False, "No search decision in response")
                else:
                    self.log_result(f"Search Pattern: {case['name']}", False, f"Status: {response.status_code}")
                    
            except Exception as e:
                self.log_result(f"Search Pattern: {case['name']}", False, f"Exception: {str(e)}")
        
        return success_count == len(test_cases)
    
    def test_force_search_parameter(self) -> bool:
        """Test force_search parameter override"""
        try:
            payload = {
                "prompt": "Tell me more",  # Would normally not search
                "max_new_tokens": 50,
                "use_search": True,
                "force_search": True,  # Should override decision
                "temperature": 0.7,
                "history": [{"role": "assistant", "content": "Here's some information about AI."}]
            }
            
            response = requests.post(CHAT_ENDPOINT, json=payload, timeout=30)
            
            if response.status_code == 200:
                data = response.json()
                if "search_decision" in data:
                    decision = data["search_decision"]
                    if decision.get("should_search") and "forced" in decision.get("reason", "").lower():
                        self.log_result("Force Search Override", True, "Force search parameter worked correctly")
                        return True
                    else:
                        self.log_result("Force Search Override", False, "Force search not detected in decision")
                        return False
                else:
                    self.log_result("Force Search Override", False, "No search decision in response")
                    return False
            else:
                self.log_result("Force Search Override", False, f"Status: {response.status_code}")
                return False
                
        except Exception as e:
            self.log_result("Force Search Override", False, f"Exception: {str(e)}")
            return False
    
    def run_all_tests(self):
        """Run all tests and print summary"""
        print("🚀 Starting Atlas Functionality Tests")
        print("=" * 50)
        
        test_methods = [
            self.test_health_endpoint,
            self.test_anonymous_chat_no_search,
            self.test_anonymous_chat_with_search,
            self.test_authenticated_chat,
            self.test_conversation_follow_up,
            self.test_search_decision_patterns,
            self.test_force_search_parameter
        ]
        
        passed = 0
        total = len(test_methods)
        
        for test_method in test_methods:
            try:
                if test_method():
                    passed += 1
                time.sleep(1)  # Brief pause between tests
            except Exception as e:
                print(f"❌ Test {test_method.__name__} crashed: {str(e)}")
        
        print("\n" + "=" * 50)
        print(f"📊 Test Summary: {passed}/{total} tests passed")
        
        if passed == total:
            print("🎉 All tests passed! Atlas is functioning correctly.")
        else:
            print(f"⚠️  {total - passed} tests failed. Check the details above.")
        
        return passed == total

if __name__ == "__main__":
    tester = AtlasTestSuite()
    success = tester.run_all_tests()
    
    # Save detailed results
    with open("test_results.json", "w") as f:
        json.dump(tester.results, f, indent=2)
    
    print(f"\n📄 Detailed results saved to test_results.json")
    exit(0 if success else 1)