CiteQuery

Team Name

Croc and Chips

Team Members

Project Description

Team Name: Croc and Chips

Team Members:
Team Member 1: Max (Ngoc Cuong Hoang)
Team Member 2: Kingsley (Kar Keat Koh)
Team Member 3: Felix (Tien Minh Nguyen)
Team Member 4: Avery (Le Quynh Nhu Doan)
Team Member 5: Jane (Phuong Tran Tran)
Team Member 6: Echo (Tianhui Ke)
Team Member 7: Emma (ThiXuan Thanh Tran)

Project Description

CiteQueryChatbot – Accurate & Trustworthy AI Assistant for Government Teams

Government agencies have huge amounts of data — HR, finance, operations, and more.
But when workers need a quick insight, they often face:
- Slow manual searching through spreadsheets or reports
- Confusing data structures that require technical knowledge
- AI chatbots that “hallucinate” (make up wrong answers)

For government decisions, where every answer must be accurate and auditable, that’s a serious problem.

Our Solution – CiteQuery

CiteQueryChatbot combines natural language processing (NLP) and SQL-based querying to give precise, explainable answers.

Take a look

Here’s how it works:

1. Upload Dataset - The user opens the program, enters their API key, and uploads a dataset (e.g., HR leave records).

2. Get a list of Questions instantly - Example: “What is the average leave days by department?”

3. Question selection - Select the question from the list of suggested questions for fast answer

4. SQL Query Generation and Visual results - Our RAG (Retrieve–Augment–Generate) pipeline converts the natural question into an exact SQL query - The system retrieves results directly from the SQLite database — no guessing, no hallucinations -Visual Results - Outputs both a important data and a bar graph for better understanding

5. Transparent Logging - Every question, SQL query, and answer is saved to a log for auditing and traceability.

Key Features

• ✅ Precise Answers – Converts natural language to SQL, retrieves exact data
  
• ✅ Suggestions for Questions – Suggests clearer questions before querying
  
• ✅ Audit Log – SQL query and logging available for future audit and traceability.
  
• ✅ Graph + Data Output – Presents insights visually and in raw numbers
  
• ✅ Auditability – Saves query history so results can be reviewed later
  
• ✅ Privacy-Friendly – Works locally with uploaded datasets
  

Why It Matters

• Reduces guesswork – eliminates AI hallucinations
  
• Saves time – no more digging through massive spreadsheets manually
  
• Supports better decisions – clear, reliable, and traceable insights
  
• Works for any team – HR, finance, operations, policy
  

Future Plans

• Support more visualizations beyond bar charts (pie, line, scatter)
  
• Expand dataset compatibility (PostgreSQL, live APIs)
  
• Add role-based permissions for multi-user teams
  
• Integrate with government reporting dashboards
• Collect all user-entered prompts to identify what topics and questions are most relevant to each department
  
• Dataset access control based on user authority levels to protect sensitive information
  
• Integrate Microsoft GraphRag for more accurate retrieval and reasoning, with support for Azure-hosted databases commonly used by government agencies
  

We can do it

CiteQueryChatbot helps government workers make fast, confident, and auditable decisions by turning questions into trusted answers.

🚀 Let’s make government data work better, together!

Please help us to improve by filling in the survey

https://tally.so/r/n98p5E

SNEAK PEEK~ The thought process