Neurocognitive System v0.2 foundation cleanup

[sinethxyz]

Summary

Reframes the historical Utopia prototype as Neurocognitive System and cleans the repository without pretending the old implementation was production-complete.

Architecture

• archives the original Utopia architecture/database documents unchanged
• adds a cleaned architecture based on Direction → Sensing → Observation → Evidence → Inference → Policy → Outcome → Calibration
• documents multimodal sensing and future EEG/Neurosity-class extension points
• adds RFCs for the system boundary, observation/evidence/inference separation, and sensor adapters
• adds an explicit historical terminology map
• records the 2026-09-23 static audit

Safety / trust boundary

• adds SECURITY.md
• states clearly that the historical API is not production-hardened or fully authenticated
• documents sensitive-data and neurophysiology boundaries
• scopes vector retrieval and embedding identity by operator
• enforces nested path/body resource-ID consistency

Correctness / reproducibility

• preserves zero-valued evidence instead of collapsing zero to null
• aligns Alembic with the declared sync PostgreSQL driver
• sources migration connectivity from runtime settings
• adds a migration for operator-scoped embeddings
• updates API identity to Neurocognitive System while retaining the historical utopia package
• adds CI for Python compilation, migrations, and tests

Intent

This is a foundation cleanup, not a rewrite.

The historical implementation remains inspectable. The new docs distinguish:

1. what was actually implemented,
2. what the old architecture intended,
3. what the cleaned architecture proposes next.

The implementation is historical. The questions were not.