Inspiration
Survey teams often move between separate tools for questionnaire design, data cleaning, statistical analysis, charts, and report writing. SurveyReport AI creates one understandable workflow from study design to a traceable report.
What it does
SurveyReport AI guides users through five stages:
- Study — define the topic and objectives, generate or edit a questionnaire, preview the digital form, and export XLSForm.
- Data — import CSV/Excel files, inspect variables, and run quality checks.
- Plan — use GPT-5.6 to propose an analysis plan, then edit and validate it before execution.
- Results — run statistics locally with Mini Stat and display tables and modern interactive charts.
- Report — preview and export validated findings to Word.
GPT-5.6 assists with questionnaires, plans, and interpretations. It does not invent or calculate statistics: numerical evidence comes from the local Mini Stat engine and remains subject to human validation.
How we built it
The application is written in Python with a CustomTkinter desktop interface. It orchestrates the proven Mini Stat engine for data import, statistical execution, visualization, questionnaire/XLSForm support, and Word export. GPT-5.6 is selectable through OpenAI or OpenRouter.
How Codex and GPT-5.6 were used
Codex helped audit the existing Mini Stat codebase, document pre-existing work, design the migration architecture, build the new five-stage interface, connect existing capabilities without falsely claiming them as Build Week work, add tests, improve safety boundaries, and accelerate debugging and verification.
GPT-5.6 powers optional assistance for questionnaire design, analysis-plan proposals, methodological discussion, and interpretation drafting. The active model is visible in the interface.
What existed before Build Week
Mini Stat already provided local project management, questionnaires and collection, Excel/CSV import, statistical analyses, charts, and Word reporting. These are documented transparently in the repository.
Build Week additions
- A new SurveyReport AI interface centered on five clear stages
- A continuous questionnaire-to-report workflow
- Visible GPT-5.6 model selection
- Human validation of analysis plans before calculations
- Modern charts embedded directly in Results
- Report preview and clearer separation of local evidence, AI-generated text, and human approval
- Migration and provenance documentation
Challenges
The main challenge was integrating a mature desktop analysis engine into a simpler experience while preserving working behavior and avoiding regressions. We also designed explicit boundaries so AI can assist reasoning without silently modifying statistical evidence.
What's next
Next steps include deeper Kobo synchronization, expanded Evidence Lock traceability, richer report templates, and optional ARIA Scholar API services for literature reviews and advanced academic writing assistance.
Log in or sign up for Devpost to join the conversation.