Loading Bar
Activity Indicator to let the user that data is being fetched
Response with attached resource link
Response that gives contact information that are from government sources

Inspiration

Pakistan is a country with incredible tourism potential, but foreign travelers often struggle to find accurate and reliable information. The inspiration for PakGuider came from the desire to simplify and enhance the travel experience for visitors to Pakistan by providing them with trusted, government-backed details.

What it does

PakGuider is an AI-powered chatbot that provides real-time responses to tourism-related questions about Pakistan. It helps users access government documents, important contacts, travel tips, visa policies, vaccination requirements, and more. The bot uses official resources to ensure accuracy and relevance.

How we built it

The core of the application was built using Streamlit for the user interface, integrated with Snowflake for storing and querying official government documents. I utilized OCR technology to extract data from PDFs and images, ensuring that even scanned or image-based documents could be processed. The chatbot is powered by Mistral LLM, which uses a combination of AI and real-time Snowflake Cortex Search queries to answer user questions.

Challenges we ran into

OCR Technology: Extracting accurate information from scanned PDFs and images was challenging. The OCR needed to be highly reliable to capture text from various document formats. Snowflake Integration: The integration of Snowflake with the system was delayed for three days due to version compatibility issues and connection errors. Session Management: Managing the state between the assistant's responses and the user's questions in Streamlit was tricky, especially when trying to keep track of conversation history. Finding Relevant Documents: It was time-consuming to search and identify the most relevant documents from a large database of official documents. VS Code and Snowflake: The connection between Snowflake and the local development environment (VS Code) had initial connectivity issues, which took time to resolve.

Accomplishments that we're proud of

Pakistan first chatbot that utilizes government information to guide tourist regarding the country . Attaches the link of the pdf from where the information is utilized, so it has context.

What we learned

Overcoming OCR challenges taught us how important it is to clean and preprocess images before applying text recognition. Dealing with Snowflake’s configuration taught me the importance of detailed troubleshooting and testing with the right versions. Importance of reading documentation. We gained deeper insights into managing stateful applications, especially when working with conversational AI systems.

What's next for Pakistan Tourist Guide Chatbot

Expansion of Knowledge Base: I plan to integrate more government documents and local tourism content to broaden the chatbot’s range. Collaborations with Hotels / Trip planning agencies: I plan to collaborate with them to get a user base . Multi-language Support: Adding more languages to make the bot more accessible to international tourists. Real-time Updates: Automating the update process for any new government publications or travel guidelines. Mobile App: Develop a mobile application to make PakGuider more accessible to travelers on-the-go.