Selemene Core

Selemene Core

Cross-cutting rules for every Selemene skill.

Two report surfaces

1. Deterministic reports — crates/noesis-vedic-api/src/report_generator/

   • birth_report.rs
   • compatibility_report.rs
   • transit_report.rs
   • Inputs: BirthDetails, CompatibilityPair, date ranges.
   • Output: GeneratedReport JSON with sections and summary.

2. Narrative witness readings — packages/witness-pipeline/src/orchestrator/integrated.ts

   • IntegratedReadingOrchestrator
   • Inputs: ReportGenerationRequest with subjects, report_level, report_mode.
   • Output: assembled markdown + source-pack (manifest + reading + reflection questions).

Bridge CLI contract

The existing @selemene/bridge CLI at bridges/cli/src/cli.ts exposes:

• init — interactive wizard, writes .selemenerc.json
• generate — fetches Rust + TS OpenAPI specs and emits Claude/OpenAI/LangChain tool definitions
• check — health check
• doctor — diagnostics

The skill cluster reads .selemenerc.json for rustUrl, tsUrl, and apiKey. It does not regenerate tool definitions.

Output manifest

Every report run must produce:

{
  "report_type": "birth|compatibility|transit|witness",
  "created_at": "ISO-8601",
  "subject_count": 1,
  "engines_used": ["vedic"],
  "artifact_path": "/absolute/path/to/artifact",
  "witness_prompt": "..."
}

Tone rule

All generated prompts and summaries are non-prescriptive mirrors. No medical, financial, or deterministic life advice. Questions only.