Sailu explores the potential for AI-mediated semantic divination through ancient dream interpretation practices. The system uses AI to fill gaps between sparse historical data points through anchored analogies, demonstrating how computational methods can recreate interpretive frameworks from incomplete or fragmentary ancient sources.
Architecture
The application uses a React frontend with TypeScript and TailwindCSS, supported by a FastAPI backend with PostgreSQL database and JWT authentication. The system processes dream submissions through a six-stage interpretation pipeline managed by a central orchestrator pattern.
Processing Flow
When a user submits a dream narrative, the system:
- Divination: Uses OpenAI GPT models to extract symbols and elements from the dream text
- Ancient Rules: Searches the knowledge base via semantic embeddings to retrieve relevant historical interpretation frameworks
- Cultural Context: Analyzes cultural and historical background using the 135+ ETCSL texts and scholarly sources
- Historical Dreams: Finds similar ancient dream accounts from Mesopotamian, Egyptian, and Greek collections using vector similarity
- Ritual Recommendations: Suggests traditional practices based on extracted symbols and cultural context
- Final Synthesis: Combines all analyses into a comprehensive interpretation with confidence scoring, rule applicability, and historical cross-referencing
The system uses sentence transformers for semantic matching, NLTK VADER for sentiment analysis, and hybrid retrieval combining semantic and symbol-based approaches.
Knowledge Base
The system integrates extensive scholarly sources including Mesopotamian cuneiform tablets, Egyptian dream papyri, Greek oneiromantic traditions, and 135+ texts from the Electronic Text Corpus of Sumerian Literature. Content includes 41 historical dreams, organized encyclopedia entries across four categories, three ancient dreambooks, and documented interpretation rules from multiple traditions.
All content uses 768-dimensional embeddings for semantic search, enabling cross-cultural analysis and contextual matching between contemporary dreams and historical precedents.
AI-Mediated Interpretation
When processing contemporary dreams, the AI generates novel reactions to novel data by drawing connections to documented ancient precedents, effectively expanding the interpretive framework beyond what the limited historical sources explicitly contain. The system performs actual interpretation rather than merely describing ancient practices, exposing implicit rules that emerge only through practical application.
Output Capabilities
The system generates multi-format reports including styled HTML presentations, downloadable PDFs, academic-friendly Markdown, and machine-readable JSON. Processing handles complex dreams without length limitations, delivering complete analytical depth for professional consultation standards.
View Sample Interpretation - Experience the complete dream interpretation system.