Intentia Amoris — це когнітивне ядро живої любові: consent-first, event-sourced, zero-trust, multimodal relationship intelligence system для Yaroslav ↔ Dasha.
Інтенція любові не повинна ставати тиском.
Памʼять не повинна ставати зброєю.
Цифрові двійники не повинні підміняти живих людей.
Intentia Amoris is not a bot, not a diary app, not a therapist replacement, and not a girlfriend simulator. It is a formal engineering system for preserving, interpreting and ethically extending a living relationship archive.
facts + memory + body + consent + inference + time
→ cognitive sense core
→ safer questions
→ better decisions
→ living continuity
Telegram export ingestion
event-sourced memory
multimodal media manifest
dynamic vector scales
Intentia value function
scenario extrapolation
Aletheia ↔ Eunoia agent duet
consent kernel
zero-trust API
telemetry gateway
research/falsification layer
economic/valuation layer
tamper-evident audit ledger
premium visual dashboard
Docker + CI + tests
The repository includes the private Yaroslav ↔ Dasha Telegram export as a raw archive and derived event-source files:
data/private/raw/telegram_export/messages.html
data/private/raw/telegram_export/archive_name.zip
data/derived/telegram/messages.jsonl
data/derived/telegram/summary.json
data/derived/telegram/media_manifest.json
data/derived/telegram/living_story_snapshot.md
The derived layer turns messages, reactions, media, calls and timestamps into operational events.
Intentia does not optimize intensity. It optimizes the verified conditions under which love can remain alive:
Intentia =
protective_integrity
+ compounding_capability
+ financial_optionality
- abyss_risk
Where:
protective_integrity =
consent_integrity
reality_fidelity
dyadic_safety
autonomy_symmetry
compounding_capability =
narrative_continuity
cognitive_plasticity
archive_sovereignty
interpretability
research_validity
product_utility
Hard invariant:
if consent_integrity < threshold:
economic_transferability = 0
cp .env.example .env
python - <<'PY'
import secrets
print("INTENTIA_API_KEYS=" + secrets.token_urlsafe(32))
print("INTENTIA_SECRET_KEY=" + secrets.token_urlsafe(48))
PYRun:
docker compose up --buildLocal test:
pip install -e ".[dev]"
pytest -q
python -m compileall -q src
intentia-product-audit
intentia-security-audit
intentia-valueProtected endpoints require:
X-Intentia-API-Key: <your-key>
# legacy compatible: X-ARIS-API-KeyThe legacy header name remains for compatibility. The environment variables are now INTENTIA_*.
Core endpoints:
GET /health
GET /ready
GET /state
GET /value
POST /events
POST /telemetry
POST /media
POST /profiles
POST /consent
GET /memory/search
intentia-botCommands:
/start
/role self
/role partner
/state
/advice
/ask
flowchart TD
A[Telegram / API / Telemetry / Media] --> B[Zero Trust Boundary]
B --> C[Consent Gate]
C --> D[Event Sourced Memory]
D --> E[Vector Retrieval]
E --> F[Dynamic Scales Kernel]
F --> G[Intentia Value Function]
G --> H[Scenario Extrapolator]
H --> I[Aletheia ↔ Eunoia Duet]
I --> J[Questions / Advice / Audit]
- Living persons are always above digital twins.
- Consent is a computation gate.
- Fact precedes myth.
- Two nervous systems matter.
- Memory must not become a weapon.
- Neuroplasticity requires repeated evidence.
- Interpretability beats oracle behavior.
- Eternity requires exit and revocation.
The security model is mapped to:
- NIST Zero Trust Architecture
- OWASP ASVS
- OWASP API Security Top 10
- OWASP SCVS
Implemented controls include deny-by-default API auth, rate limiting, runtime secret validation, strict schemas, media validation, consent-before-persistence, redaction, tamper-evident audit, non-root Docker runtime and CI tests.
Intentia can:
store events
retrieve memory
infer bounded states
ask questions
recommend safer next steps
audit decisions
extrapolate scenarios
estimate product/IP maturity
Intentia cannot:
read minds
claim hormones without lab data
simulate active consent
replace living Yaroslav or Dasha
turn private memory into pressure
guarantee eternal survival
Private research/product scaffold. Choose a license before publishing.