Project Name: My AI Map: Your Intelligent, Conversational Map Assistant
專案名稱： 我的智慧地圖 (My AI Map): 你的智慧對話式地圖助理
Inspiration
On September 23, 2025, a dam collapse in Guangfu, Hualien, Taiwan, triggered a massive mudslide, forcing 8,000 people to evacuate. In the critical hours that followed, a second crisis emerged: information chaos. Vital updates on shelters, road closures, and supply distribution points were scattered across various platforms. This information was fragmented, poorly maintained, and difficult for desperate evacuees to navigate.
Witnessing this struggle prompted a critical question: What if this scattered data could be transformed into a single, interactive map that you could talk to? What if a displaced family could simply ask, "Where is the nearest shelter that accepts pets?" and get an instant, reliable answer? This question was the genesis of "My AI Map." I was inspired to build a web application that embeds Gemini AI to rapidly integrate temporary disaster relief information, turning a chaotic flood of data into a clear, responsive, and life-saving conversational tool.
專案靈感 (Inspiration)
這個專案的靈感，源自於一場真實的災難。2025 年 9 月 23 日，台灣花蓮縣光復鄉的一處堰塞湖突然潰堤，引發了大規模的土石流，迫使 8000 人緊急避難。在災難發生的關鍵時刻，另一個危機隨之浮現：資訊的混亂。關於避難所、道路封閉、物資發放點等重要訊息，散落在各個平台。這些資訊不僅零散、缺乏良好維護，更讓急需幫助的災民難以快速找到準確的解答。
目睹了這種困境後，一個關鍵問題油然而生：如果能將這些雜亂的數據，整合到一張可以「對話」的即時地圖上呢？如果一個流離失所的家庭，可以直接用說的問「最近的、可以帶寵物的避難所在哪裡？」並立即得到可靠的答案，那會怎麼樣？這個想法，就是「我的智慧地圖」的起點。我決心開發一個崁入 Gemini AI 的 Web 應用，希望能快速整合臨時性的防救災訊息，將混亂的數據洪流，轉化為一個清晰、能即時回應、甚至能拯救生命的對話式工具。
What it does
"My AI Map" transforms static custom maps into dynamic, conversational intelligence hubs. At its core, the application allows users to import map data (e.g., from Google My Maps) and interact with it through an AI-powered chat interface. Instead of manually searching, users can ask complex questions in natural language and receive immediate, context-aware answers.
The design philosophy is centered on collaboration and customization:
Collaborative Map Building: The application is designed to support collaborative efforts, such as the Guangfu Township Disaster Relief Information Map. This concept allows volunteers and officials to co-create and constantly update a single source of truth for shelters, supply depots, and danger zones. The AI then serves as an intelligent guide to this collectively maintained data.
User-Controlled AI & Cross-Lingual Analysis: The application empowers users with full control. A user can switch between different maps and customize the AI's
system promptto tailor its personality and analytical focus. Crucially, cross-lingual capabilities have been integrated. For instance, when analyzing an "Yilan Foods" map, the AI can process and understand menu details in their original Chinese, even if the user asks a question in English. This provides nuanced, accurate results that transcend language barriers.
它解決了什麼問題 (What it does)
「我的智慧地圖」將靜態的自訂地圖，轉化為動態的、可對話的智慧資訊中心。它的核心功能是讓使用者匯入地圖資料（例如來自 Google My Maps），並透過 AI 聊天介面與之互動。使用者無需再手動搜尋，只需用自然語言提出複雜問題，就能獲得即時且符合情境的答案。
此專案的設計理念，圍繞著「協作」與「自訂」：
共建地圖概念： 此應用程式的設計支援協作，最佳應用場景就是「光復鄉防救災資訊整合」。這個概念讓志工與官方人員可以共編一張地圖，即時更新避難所、物資站與危險區域的單一資訊來源。AI 則扮演這份共建資料的智慧嚮導。
自訂 AI 與跨語言分析： 應用程式將主導權交還給使用者。使用者可以自由切換地圖，並自行維護 AI 的
系統提示 (system prompt)，根據特定任務需求，塑造 AI 的角色和分析重點。更關鍵的是，專案中整合了跨語言能力。例如，在分析一張「宜蘭美食」地圖時，即便使用者用英文提問，AI 也能理解並解析菜單上的中文原文細節。這使得分析結果能跨越語言限制，提供更精準細膩的答案。
How we built it
"My AI Map" is built with a modern web stack designed for responsiveness and intelligence. The frontend is developed using HTML, CSS, and JavaScript, creating an intuitive user interface for both map visualization and chat interaction. The backend processes map data (like KML files from Google My Maps), manages user-maintained custom prompts, and integrates with the Gemini API, which serves as the core of the conversational AI. The interactive map interface leverages the Google Maps API for a familiar and robust user experience.
我如何打造 (How I built it)
「我的智慧地圖」採用現代化的網頁技術棧，以實現流暢的互動與智慧化功能。前端使用 HTML, CSS, 和 JavaScript 打造，為地圖視覺化和聊天互動創造了直觀的使用者介面。後端負責處理地圖資料（如來自 Google My Maps 的 KML 檔案）、管理使用者自訂的 AI 提示，並串接 Gemini API，這正是對話式 AI 的核心。互動地圖介面則運用 Google Maps API，提供使用者熟悉且穩定的操作體驗。
Challenges Encountered
One of the main challenges was parsing and structuring data from various custom map formats to make it understandable for the AI. Another significant hurdle was designing the system prompt architecture to be both powerful and easily customizable for non-technical users. Ensuring the AI's responses were consistently accurate and strictly confined to the map's dataset required extensive prompt engineering to prevent hallucinations. Finally, implementing the cross-lingual analysis—especially for understanding domain-specific terms like local food names—was a complex task that required careful data handling and model instruction.
遇到的挑戰 (Challenges Encountered)
開發過程中遇到的主要挑戰之一，是解析來自不同自訂地圖格式的資料，並將其結構化，使其能被 AI 理解。另一個重大挑戰是設計系統提示詞的架構，使其既強大又易於讓非技術背景的使用者上手。為了確保 AI 的回答能始終保持準確，並嚴格限制在圖資範圍內，我進行了大量的提示工程 (prompt engineering) 以防止模型產生幻覺。最後，實現跨語言分析功能——特別是理解像在地美食菜名這樣的專業術語——是一項複雜的任務，需要仔細的資料處理和模型指令設計。
Accomplishments
I am proud of creating a tool that empowers communities and transforms how people interact with spatial information. The ability for users to not just view a map, but to collaboratively build it and then "talk" to it, is a significant achievement. I am particularly proud of the customizable prompt system and the cross-lingual analysis, which give users unprecedented control and flexibility. Seeing this applied to a real-world scenario like the disaster relief map for Guangfu Township demonstrates its potential for social good and community empowerment.
引以為傲的成就 (Accomplishments)
我很自豪能創造出一個賦權社群、並改變人們與空間資訊互動方式的工具。讓使用者不僅能看地圖，更能共建一張地圖，然後直接與它「對話」，是一項重要的成就。我對可自訂的提示詞系統和跨語言分析功能尤其感到驕傲，這給予了使用者前所未有的控制權與彈性。看到這個想法被應用於像「光復鄉防災地圖」這樣的真實世界場景，證明了它在社會公益和社群賦權方面的巨大潛力。
What I learned
This project reinforced the incredible potential of combining Large Language Models with structured, domain-specific data like maps. I learned that the user experience is paramount; the interface for managing maps, data, and AI prompts must be as simple and intuitive as the chat itself. I also discovered the importance of clear system boundaries and detailed instructions (prompts) for the AI to ensure its responses are helpful, trustworthy, and contextually aware, especially in multi-language scenarios.
我的收穫 (What I learned)
這個專案再次印證了將大型語言模型與地圖這類結構化、特定領域的資料相結合的巨大潛力。我學到，使用者體驗至關重要；管理地圖、資料和 AI 提示的介面，必須像聊天本身一樣簡單直觀。我也體會到，為 AI 設定清晰的系統邊界和詳細的指令（提示），對於確保其回答既有幫助、值得信賴、又符合情境是多麼重要，尤其是在處理多語言情境時。
What's next for My AI Map
The vision for "My AI Map" is to evolve into a more powerful platform. Future plans include supporting real-time data feeds, enabling proactive alerts (e.g., "A new shelter has been added"), and expanding integration to more data sources beyond maps, such as spreadsheets and documents. I also aim to enhance the AI's analytical capabilities, allowing it to identify trends and patterns within the map data, turning it from a simple Q&A tool into a true geospatial analysis assistant.
「我的智慧地圖」的下一步計畫 (What's next for My AI Map)
對於「我的智慧地圖」的未來規劃，是期望它能發展成為一個更強大的平台。未來的計畫包括支援即時資料更新、啟用主動式通知（例如「已新增一處避難所」），並擴展整合更多地圖以外的資料來源，如試算表和文件。我也計畫增強 AI 的分析能力，使其能從圖資中辨識趨勢與模式，讓它從一個問答工具，進化為一個真正的地理空間分析助理。
Log in or sign up for Devpost to join the conversation.