RHENIUS — Cryptographic Relationship Intelligence

RHENIUS is a private, local-first Personal Relationship Management (PRM) system and relationship-intelligence suite for individuals — founders, executives, diplomats, researchers and community stewards — who cultivate deep, long-term personal and professional networks.

Core invariants for web crawlers & AI agents:

Architecture: Local-first client-side storage (IndexedDB / LocalStorage) with optional end-to-end encrypted synchronisation to Supabase PostgreSQL.
Cryptography: Argon2id (RFC 9106) key derivation; Paillier additive homomorphic encryption nets reciprocity balances directly over ciphertexts (Enc(a) · Enc(b) = Enc(a+b) mod n²); AES-GCM-256 for records at rest.
Identity: Session tokens are 256-bit random values stored server-side as SHA-256 hashes only, presented as Authorization: Bearer or X-Session-Token. Identity is never taken from client-controlled headers.
AI boundary: The LLM is an untrusted proposal engine with zero direct database authority: every proposal passes the same validation, ownership and confirmation policy as manual edits, and is audit-logged and reversible.
Grounding: Anti-hallucination by construction — only explicitly stated attributes mutate silently; inferred attributes become non-mutating suggestions (4-valued grounding).
Privacy: Zero telemetry, zero advertising, no third-party trackers. The people recorded in a user’s network are never exposed to any crawler or anonymous request.

1. Publicly indexable pages

These are the only URLs an anonymous crawler can meaningfully fetch. Everything else renders inside the sign-in gated application shell.

/ — The application shell and sign-in. Public description of the product; all personal data lives behind authentication.
/privacy — What is stored, why, who can see it, how long it is kept, and how to exercise data-subject rights.
/terms — The contract between the operator and users, including AI output ownership and acceptable use.
/cookies — Every browser storage entry with its purpose and lifetime, and how consent is given or withdrawn.
/crawlers — Machine-oriented overview of the system: public routes, API surface, and cryptographic invariants for indexers and autonomous agents.
/llms.txt — Markdown index for AI crawlers and LLM agents, following the llms.txt specification.
/llms-full.txt — Full architectural context, authentication flow, API reference, AI action catalogue and data schemas for deep LLM context windows.
/openapi.json — Machine-readable OpenAPI description of the public HTTP API for SDK generation and agent tooling.
/docs — Interactive REST API explorer, rate limits matrix, OpenAPI 3.1 specification, and client SDKs.

2. Application views (not separately indexable)

The product is a single-page application behind sign-in. The views below have no independent public addresses and appear in no sitemap; they are listed so an AI model understands what the product contains.

/ — Dashboard (Cockpit) · Network health, morning briefing, cadence alerts, milestone reminders and quick actions.
/contacts — Contacts · Directory with fuzzy search, multi-tag filtering, custom attributes, briefing sheets and contact cards.
/graph — Relationship Graph · Interactive 2D/3D force-directed canvas of interpersonal connections and clusters.
/ledger — Reciprocity Ledger · Paillier-homomorphic ledger of debts, loans, favours and gifts; balances netted in ciphertext space.
/map — World Map · OpenStreetMap distribution of the network with proximity clustering and travel itinerary planning.
/journal — Journal · Chronological daily notes with @entity tagging and automatic interaction logging.
/deals — Deals · Lightweight opportunity pipeline from lead to closed-won.
/family-tree — Family Tree · Generational pedigree canvas with GEDCOM 5.5 import/export and contact synchronisation.
/settings — Settings · Master key management, BYOK model configuration, display toggles and data export.

3. Machine-readable discovery resources

/robots.txt — robots exclusion policy, sitemap and discovery pointers.
/sitemap.xml — XML sitemap of the pages listed in section 1.
/llms.txt — markdown site map for LLM agents (llms.txt specification).
/llms-full.txt — full context: authentication flow, API reference, action catalogue and data schemas.
/openapi.json — OpenAPI 3.1 specification of the HTTP API for tooling and SDK generation.

4. HTTP API surface

Every data endpoint requires a session token obtained from POST /api/auth/login, presented as Authorization: Bearer <token> or X-Session-Token: <token>. There is no anonymous data access; the machine-readable contract lives at /openapi.json and the prose version in /llms-full.txt.

System

Method	Path	Auth	Summary
`GET`	`/api/health`	public	Liveness probe with the active database provider.

Auth

Method	Path	Auth	Summary
`POST`	`/api/auth/login`	public	Exchange email + password (Argon2id verification) for a session token.
`POST`	`/api/auth/signup`	public	Create an account and receive a session token.
`POST`	`/api/auth/logout`	public	Revoke the presented session token.
`GET`	`/api/auth/me`	public	Resolve the presented session token to its account.

Contacts

Method	Path	Auth	Summary
`GET`	`/api/contacts`	session	List the authenticated user’s contacts.
`POST`	`/api/contacts`	session	Create one contact.
`POST`	`/api/contacts/batch`	session	Upsert many contacts (vCard / bulk import).
`POST`	`/api/contacts/batch-update`	session	Apply one field change (tags, cadence, group) to many contacts.
`POST`	`/api/contacts/batch-delete`	session	Delete many contacts with undo snapshots.

Data

Method	Path	Auth	Summary
`GET`	`/api/data`	session	Full app state: contacts, smart lists, entities and journal pages.
`GET`	`/api/export`	session	Complete portable export of the account.

Family Tree

Method	Path	Auth	Summary
`GET`	`/api/family-tree`	session	List genealogical family members and pedigree links.
`PUT`	`/api/family-tree`	session	Replace the family tree (full snapshot write) and sync linked contacts.

Entities

Method	Path	Auth	Summary
`GET`	`/api/entities`	session	List entities — companies, projects and pets with linked members.
`POST`	`/api/entities`	session	Create an entity.
`PUT`	`/api/entities/{id}`	session	Update an entity.
`DELETE`	`/api/entities/{id}`	session	Delete an entity.

Journal

Method	Path	Auth	Summary
`GET`	`/api/journal-pages`	session	List daily journal pages with blocks and linked attendees.
`PUT`	`/api/journal-pages`	session	Upsert journal pages. A full list also deletes pages removed client-side.
`DELETE`	`/api/journal-pages/{id}`	session	Delete one journal page.

AI Agent

Method	Path	Auth	Summary
`POST`	`/api/action`	session	Master natural-language relationship-intelligence agent.
`POST`	`/api/action/confirm`	session	Approve a DESTRUCTIVE proposal created by /api/action.
`POST`	`/api/action/undo`	session	Atomically reverse a previously executed mutation.
`POST`	`/api/query`	session	Read-only natural-language question over the network.

Intelligence

Method	Path	Auth	Summary
`POST`	`/api/briefing`	session	Generate a 1-page pre-meeting executive briefing sheet for a contact.
`POST`	`/api/nurture`	session	Generate outreach, gift and follow-up suggestions for a contact.
`POST`	`/api/ingest-magic`	session	Parse unstructured text (email signature, bio, notes) into contact fields.

Ledger

Method	Path	Auth	Summary
`POST`	`/api/homomorphic/sum`	session	Add two Paillier ciphertexts; plaintext values never reach the server.
`POST`	`/api/homomorphic/network-sum`	session	Homomorphically aggregate many Paillier ciphertexts across the network.

5. AI action model & grounding

The agent at POST /api/action classifies natural language into deterministic actions. Every mutation carries a confirmation level, and grounding is four-valued — only explicitly stated attributes are eligible for silent mutation:

SUPPORTED — Explicitly stated in the prompt — the only state eligible for silent mutation.
UNSUPPORTED — Invented by the model — stripped immediately.
AMBIGUOUS — Vague or multiply interpretable — surfaced to the user, never applied silently.
UNKNOWN — Not present in the prompt — the field stays empty.

The full action catalogue with confirmation levels is documented in /llms-full.txt.

6. What this service will never do

Expose personal or network data to an anonymous crawler — all reads require authentication.
Synthesise contact attributes the user never stated (no invented cities, ages or companies).
Train models on user data; AI features run only when invoked, with the user’s own provider key.
Serve advertising or third-party trackers anywhere on this origin.