中文版本

This project is developing an AI Travel Assistant that generates daily itineraries from user constraints and preferences, incorporates weather-aware adjustments and routing, supports booking with a one-click confirm flow, and can quickly regenerate a plan when new constraints are provided.

1. AI-driven personalized trip generation: Generate daily trip plans under constraints (destination, budget, party size, date range, preferences).

2. Weather-aware dynamic planning: Monitor and incorporate destination weather for the relevant day(s), adjusting activities intelligently.

3. Iterative LLM feedback loop: Support full re-prompting to regenerate itineraries when users provide new constraints or are not satisfied with the plan.

4. Contextual Trip Insights for smarter decisions: Provide clear Trip Insights (local news and practical tips) to improve decision quality and the overall travel experience.

5. Unified booking orchestration: Offer one-click confirmation for hotels/transport/tickets (Mockup) as a demonstration of an integrated booking workflow.

6. Route optimization with visual mapping: Optimize routing and visualize navigation between locations by calculating travel time and distance (walking/transport/driving where applicable), and providing interactive map visualization.

In summary: These features combine LLM intelligence + real-world data pipelines (weather, mapping, routing) to deliver a dynamic, practical, and user-centric travel planning experience.

https://aitrip.dingzh.cc

Travel planning is typically fragmented across multiple tools for research, weather, routing, and booking. When weather or availability changes, users must manually revisit each step to fix conflicts. AI Trip Assistant consolidates this workflow: start from user preferences and constraints, produce a structured day-by-day itinerary, factor in real-world signals such as destination weather and route feasibility, and allow rapid regeneration when constraints change. On top of planning, it provides unified booking orchestration with a one-click confirm demo flow, and Trip Insights (local news and tips) to raise situational awareness. The goal is to make travel smarter, more resilient, and lower-effort for users.

• Frontend: React 19, Vite, TypeScript
• Backend: Spring Boot 3.5, Java 21, Maven
• Database: PostgreSQL 16 (Neon Serverless)
• AI Agent Model: OpenAI gpt-4o-mini

We integrated 11 advanced technologies to significantly improve system performance (faster static delivery with CDN), scalability (EC2 + serverless PostgreSQL), and maintainability (Dockerized CI/CD automation).

• Use Alibaba Cloud CDN to accelerate access to the EC2 server (Singapore), prioritizing cached content on edge nodes (but effectiveness is limited—when edge nodes have no cache they still need to access the Singapore origin. Consider later deploying domestic cloud servers to create dual-origin servers).
• Backend servers located on EC2 (Singapore) act on behalf of users to send requests to the LLM (OpenAI), avoiding the problem of users in Mainland China being unable to directly request the LLM.
• For external resources that cannot be directly accessed from Mainland China, such as GeoDB Cities and Unsplash images, use Nginx reverse proxy on EC2 (Singapore) to issue requests from the EC2 side, avoiding direct requests from users in Mainland China.
• Adapt to Amap (Gaode): use a very small Google resource as a probe when users access; within Mainland China default to Amap, overseas default to Google Maps; use Google Geocode to obtain geocoding and then provide that geocoding to Amap to solve Amap API’s poor support for searching place names in English. Limitation: Amap cannot provide route planning for overseas cities (scenario: user is in Mainland China but planning travel in an overseas city); likewise, Google Maps’ route planning support for Mainland China cities is also poor.
• Deploy dual-end OSS origin servers to synchronize static resources and accelerate access with a CDN; set CDN rules so overseas users default to origin OSS (Singapore) and Mainland China users default to origin OSS (Hangzhou), speeding up access for both domestic and overseas users.

• Live Demo: EC2 Deployment + Alicdn Speedup
• UI Mockup: Figma Prototype
• Jira Project: Kanban & Sprint Board
• Stage 1 Report: Overleaf Document

Users can explore the major features and start planning a trip instantly via “Try Free”.

Destination autocomplete, date selection, budget, currency, and preferences input enable highly personalized planning.

Manage multiple saved itineraries, revisit plans, or create new trips conveniently.

AI-generated day schedule, real-time weather, destination photos and contextual travel insights enhance decision quality.

All activities become booking tasks. Demonstrates complete booking readiness with batch confirmation.

Estimated distance and duration between attractions, supporting better route optimization and mobility planning.

Users can add or adjust preference to regenerate itineraries seamlessly with the LLM.

Profile editing, avatar upload, and verified email ensure a personalized and secure experience. Gender and age will be used as preference if applicable.

• java 21
• git
• nvm (not necessary but recommended) + Node 20

Nvm is a nodejs manager that lets you install and switch between different nodejs versions locally. After you install it, you can use it to install and switch between different nodejs versions.

# if you don't have the requirement of switching node versions, you can ignore this tip and just install node directly
nvm install 20.19.4
nvm use 20.19.4

Check that everything is installed:

java --version
git --version
node --version

git clone https://github.sydney.edu.au/25S2-ELEC5620-Wed9-11-Group61/ELEC5620_AI_Trip_Assistant.git
git checkout <your development branch> 

Or you can use any other git gui client, like Github Desktop, TortoiseGit.

This project uses Neon to host PostgreSQL.

Please create a database in the Neon Console, and then fill in the corresponding values for spring.datasource.url, spring.datasource.username, and spring.datasource.password in api/src/main/resources/application.yaml.

This project uses OpenAI GPT-4o-Mini as the default LLM model to generate trip plans, you can choose your own model or use the default model.

1. Create a new OpenAI account in https://platform.openai.com/account/api-keys.
2. Add your API key to application.yaml spring.ai.openai.api-key.
3. Add your model name to application.yaml spring.ai.openai.model-name.

OpenAI GPT-4o-Mini docs: https://platform.openai.com/docs/models/gpt-4o-mini

1. Create a new S3 bucket in AWS Console.
2. Add your bucket region to application.yaml aws.s3.region.
3. Add your bucket name to application.yaml aws.s3.bucket-name.
4. Add your dir name in bucket to application.yaml aws.s3.dir-name.
5. Add your cdn domain to application.yaml aws.s3.cdn.
6. Create a new IAM user and generate an Access Key ID and Secret Access Key.
7. Add the Access Key ID and Secret Access Key to your environment variables with the key name AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY.

AWS S3 docs: https://docs.aws.amazon.com/AmazonS3/latest/userguide/create-bucket-configure.html

If you are in China, aliyun oss service is a better choice for you: https://help.aliyun.com/zh/oss/developer-reference/oss-java-sdk/

Please refer to the following guides to set up the external APIs.

• GeoDB API: http://geodb-cities-api.wirefreethought.com/
• Google Maps API: https://developers.google.com/maps/documentation/javascript/get-api-key
• AMAP API: https://lbs.amap.com/
• Unsplash API: https://unsplash.com/developers
• OpenWeather API: https://openweathermap.org/api
• SendGrid Email API: https://www.twilio.com/docs/sendgrid/for-developers/sending-email/api-getting-started

Make sure you're inside of ELEC5620_AI_Trip_Assistant

cd web
npm install 
npm run dev

Make sure you're inside of ELEC5620_AI_Trip_Assistant

cd api
.\mvnw spring-boot:run # Windows
./mvnw spring-boot:run # Linux/Mac

cd external-service
.\mvnw spring-boot:run # Windows
./mvnw spring-boot:run # Linux/Mac

Or you can use any other IDE to run this project, like IntelliJ IDEA.

Go to a web browser and access http://localhost:5173/