Upload destinations.json

fix(chatbot): destination-suggestion + map follow-up wrongly hit off-topic guardrail

fix(chatbot): food queries misclassified as activities

fix(chatbot): set awaiting after nearby_attractions so 'có' plans a trip

feat(chatbot): surface real hotels and attractions from KB

fix(chatbot): accumulate dietary restrictions + real-time weather

feat(chatbot): OOD guard for arithmetic + find_hotel slot-filling

feat(chatbot): lightweight dialog-state for follow-up replies

fix(chatbot): guard response templates against empty/unfilled slots

test(security): fix all test failures — zero FAILED, zero plain skips

fix(chatbot): add test coverage for all 17 intents; fix dataset quality

fix(chatbot): remove Google Maps button, map click opens GraphHopper; conversational trip intro text

fix(chatbot): simulate multi-pin route map for destinations without per-activity coords

fix(hotel,chatbot): add WiFi to hotel data + fix chatbot map waypoints

fix(chatbot): preserve KB across HF deploys and fix LLM startup

fix(chatbot): add missing use_mmr param to DestinationRecommender.recommend

fix: patch security vulnerabilities and chatbot API key config

refactor: predict() 851→648 lines — parallel IO + module-level constants + card builders

feat: improve rate limiting — env-configurable, all ML endpoints, Retry-After header

fix: train_test_split guard + asyncio background task leak

fix: resolve remaining HIGH/MEDIUM issues from code review

refactor: fix HIGH issues - constructor injection, i18n, ML cache & preprocessing

security: fix 6 critical vulnerabilities across chatbot and backend

fix: nav active state for car-rental & activities + silent catch bugs

security: remove hardcoded ML service API key - require env var (CHATBOT_ML_SERVICE_API_KEY)

feat: add pagination to HotelListPage, fix i18n in hotel cards, strengthen chatbot guardrails

feat(audit): full project audit — ESLint, tests, i18n, transactions, docs

fix(chatbot): add missing tripadvisor_enricher module to repo

feat(llm): switch from SeaLLM to Qwen2.5-7B-Instruct for LLM space

fix(chatbot): resolve 3 reviewer issues — conversationCount bug, task error logging, CI username variable

fix(chatbot): resolve NameError in guardrail and cache UserMemoryService init failure

feat(chatbot): integrate UserMemoryService into predict endpoint for cross-session memory

feat(chatbot): thread user_memory through LLM pipeline — context builder, inference, openai generator

feat(chatbot): add UserMemoryService for cross-session MongoDB memory

feat(chatbot): enable LLM generator by default (USE_LLM_GENERATOR=True)

fix(chatbot): strip UTF-8 BOM from responses_vi.json and responses_en.json

fix(chatbot): use utf-8-sig encoding to handle UTF-8 BOM in JSON files

feat: real-data dataset enrichment + nearby_attractions + travel_tips intents

feat: MapCard zoom controls + multi-marker, nav icon fix, destination_info intent + DestinationInfoCard component

fix(map): replace blocked OSM iframe with static tiles + fix POI search

feat(ML): add show_map intent + context-aware map routing

Wave 3: expand NER fixture 85->102, time_preference F1 0.40->0.98, people/duration fixes

Wave 2: compare_destinations data + NER fixes + evaluation update

eval(chatbot): Wave 1 retrain + held-out evaluation results

feat(chatbot): Phase 4 Wave 1 - NER fixes + Intent TF-IDF upgrade + OOD rejector

fix(phase3): GlobalExceptionHandler, TopUpRequestDTO validation, RefundService wallet credit-back

deploy: add Vercel, Railway, HuggingFace Spaces deployment configs

feat: security audit fixes, performance improvements & global data pipeline

feat(chatbot): add chain-of-thought logic for personalize_recommend intent

Round 15: Dark mode page backgrounds + Chatbot fixes (NER/weather/events/templates/currency)