HOUSE OF SAJU
In developmentAI personalized content
House of Saju
Make calculated insights easier to read.
A reading experience that adds LLM explanations to calendar-based and rule-based fortune and saju reports.
HOW IT COMES TOGETHER
The core workflow- 01
Input and calendar
- 02
Rule-based calculation
- 03
LLM explanation
- 04
Report assembly
01 / THE QUESTION
The problem we set out to solve.
Traditional saju data and scores can be difficult for newcomers to understand. If generated explanations also alter the underlying calculations, the results lose consistency.
02 / OUR APPROACH
Our approach.
Calculation and narration have separate roles. A rules engine determines the result, and the LLM refines its explanation. Report generation is linked to payment events and processing states to reduce duplicate work.
- 01
Rule-based results
A traditional calendar and rules engine provide the results used in free fortune readings.
- 02
Readable explanations
OpenAI refines the wording while preserving the original scores and calculated results.
- 03
Report generation
Structured saju reports are generated after payment events, with processing states stored in a repository.
- 04
Fallback and deduplication
Fallback responses cover missing API configuration, while state management prevents repeated requests from generating duplicate reports.
03 / INTELLIGENCE AT WORK
Where AI does the work.
The LLM is an explanation layer over the calculation engine. Its role is limited so generated wording does not change the underlying scores or rules.
04 / WHERE WE ARE
Where the project stands.
Free readings and saju reports are under active development. We are improving consistency between calculated results and AI explanations, alongside duplicate-request and failure handling.
Saju and fortune readings are for entertainment and self-reflection.
