⚡ Internal Platform Docs

The MELTS.LA Documentation Hub

Comprehensive reference for the Melts.la ED telehealth platform — canonical core documents, KG schema, system diagrams, operational workflows, and the full knowledge base tree.

331

KG Nodes

561

KG Edges

Node Types

21+

Product Stacks

65+

Deliverables

Core Documents

25 canonical documents from ed gem/00_CANONICAL_CORE — click any card to expand.

📋

About Us

v1.0 · 2026-02-24

Internal identity document — operating model, clinical posture, product posture, ethical/compliance boundaries.

What Melts.la is

Melts.la is a California-focused digital health platform for erectile dysfunction (ED) care delivered through an asynchronous telehealth model.

Collect medically relevant information through a structured intake, enabled by AI / multi-modal intake
Support clinician review with explainable decision logic
Enable fulfillment via a partner pharmacy relationship

Operating Model

Clinical care (PC / clinicians)

Care is provided by a partner physician group / professional corporation (PC). Clinicians review intake data, approve/modify/decline treatment, and are accountable for prescribing decisions.

Fulfillment (pharmacy)

Generation Pharmacy (503A) is positioned as the primary fulfillment partner for compounded products.

Platform responsibilities

Intake design, patient experience, consent capture
Policy/rules engine implementation (hard stops + decision logic)
Audit trail integrity
Patient education and follow-up workflow coordination

Clinical Posture

ED is multifactorial; evaluation should include medical/sexual/psychosocial history. PDE5 inhibitors are first-line. Adjuncts (PT-141, Oxytocin, Apomorphine) offered only under explicit eligibility criteria.

Product Posture

Form factor and experience (sublingual films, mini-melts, troches, ODT/RDT)
Vardenafil-centric options where clinically appropriate
Modular adjunct pathways gated by clinical rules

Core Identity

Who: DTC digital health platform (CA-first) for ED
How: Asynchronous, explainable intake + clinician review + rapid fulfillment
Why: Improve safety, personalization, and patient comfort beyond commodity ED telehealth

📑

White Paper Draft

v0.1 · 2026-01-08

Formal defensible narrative — clinical evidence, operating model, formulary strategy, AI decision support, regulatory posture.

Abstract

This paper outlines an asynchronous, California-first DTC ED telehealth model designed to preserve clinical rigor and audit defensibility while reducing stigma and friction. The model supports a patient-specific, adjunct-forward formulary under explicit eligibility gates, conservative evidence handling, and clinician oversight.

Clinical Foundation

Medical history emphasizing cardiovascular status, contraindicated medications
Sexual history (onset, persistence, situational vs generalized pattern)
Psychosocial context (stress, anxiety, relationship strain)
Validated measurement: IIEF-5 / SHIM and EHS

Formulary Strategy (Adjunct-Forward)

Foundation: PDE5 inhibitors

First-line pharmacotherapy. Differentiation via agent selection, dose strategy, and patient-specific form factor.

PT-141 (bremelanotide) — Desire/arousal adjunct

Erectogenic activity in controlled studies. Requires explicit side-effect counseling and conservative titration.

Apomorphine — Rescue/alternative lane

Benefit vs placebo with dose-limiting adverse events (nausea). Surfaced under tighter criteria.

Oxytocin — Anxiety/connection adjunct

Mechanistic plausibility but mixed human outcomes. Conservative claims with explicit uncertainty labeling.

AI Decision Support

Summarize intake into clinician note draft
Propose recommended lane with evidence citations
Cannot bypass hard stops or invent evidence
Full audit trail: triggered rule IDs, inputs, rationale, clinician approval

⚖️

Decision Matrix

v1.2 · 2026-02-17

Weighted priority framework for resolving tradeoffs — 26 dimensions across 4 tiers (Existential → Table Stakes).

Overview

Weighted priority system for resolving tradeoffs between clinical, UX, AI, compliance, and business considerations.

Pre-Layer Hard Rules (Non-Rankable)

Active nitrate or popper use → HARD STOP
Recent MI/stroke → HARD STOP
Unstable angina → HARD STOP
No fabricated statistics or citations

Weight Tiers

Tier	Weight Range	Classification
T1	100–85	Existential
T2	80–60	Strategic
T3	45–25	Operational
T4	15–5	Table Stakes

Top 10 Dimensions

#	Dimension	Tier	Wt
1	LA_Local_Personalization	T1	100
2	Regulatory_Defensibility	T1	95
3	Clinical_Appropriateness	T1	90
4	Compounding_Justification_Strength	T1	85
5	Personalization_Accuracy	T2	80
6	AI_Explainability	T2	78
7	Psychological_Comfort	T2	75
8	Modality_Flexibility	T2	72
9	Data_Privacy_Security	T2	68
10	Emotional_Engagement	T2	65

Full matrix contains 26 dimensions. See source document for complete listing.

🏗️

Architecture Overview

v1.0 · 2026-03-20

High-level system components — patient UX, intake orchestration, policy engine, AI support, clinician review, pharmacy fulfillment.

System Components

Patient Experience Layer — Web/mobile UI: onboarding, consent, intake, education, checkout
Intake Orchestration — Question engine, instrument scoring (SHIM, EHS, CV screen), consent capture
Policy & Rules Engine — Two layers: Hard stops + Decision layer. All logged with rule IDs.
AI Support Layer — RAG over KB. Cannot override hard stops. Outputs require truth/evidence/rule citations.
Clinician Review Layer — Dashboard with intake summary, recommendations, safety triggers. Approve/modify/decline.
Pharmacy Fulfillment — Rx intake, compounding workflow, shipping, pharmacist counseling.
Data & Analytics — Immutable audit logs, experiment flags, outcomes tracking, drift monitoring.

Kill Switch Pattern

Hard-stop rules: AI cannot bypass, UI cannot hide, clinician overrides are explicit and logged.

KB → Policy → Product Mapping

Every AI recommendation must reference at least one Truth ID, one Rule ID, and ideally one Evidence ID.

🎯

Vision & Goals

v2.0 · 2026-02-24

Mission, vision, core values, goals (G1–G5), success metrics, and explicit non-goals for the ED launch.

Mission

Deliver clinically rigorous, evidence-aligned ED care through an asynchronous, patient-centered telehealth experience.

Core Values

Safety first — Non-negotiable hard-stop rules
Evidence integrity — No fabricated claims; uncertain areas labeled
Explainability + auditability — Every recommendation traceable
Patient dignity + privacy — Reduce stigma, sensitive data handling
Patient autonomy — Within clinical boundaries
Local accountability — LA-first execution

Goals

G1 — Clinical quality: Guideline alignment, severity measurement, red-flag triage
G2 — Regulatory defensibility: Async evaluation adequacy, compounding rationale, audit trail
G3 — Patient experience: Stigma reduction, efficient intake, multi-modal support
G4 — Operational excellence: Fast time-to-fulfillment, minimize back-and-forth
G5 — Learning system: Structured outcomes, controlled experimentation, safety monitoring

🧹

Data Hygiene Rules

v0.1 · 2026-01-08

Repository hygiene, canonicalization pipeline, evidence integrity, scope discipline, and KG readiness conventions.

Content Taxonomy

_DELIVERABLES/ — Canonical outputs only (truths, evidence, rules, synthesis)
_INTERNAL_SOURCES/ — Drafts, brainstorms, internal memos
_EXTERNAL_SOURCES/ — Guidelines PDFs, peer-reviewed papers, regulatory sources
_ARCHIVE/ — Older versions for traceability

Canonicalization Pipeline

A claim becomes canonical only through: Evidence Table → Canonical Truth → Rules/Policy → Synthesis

Evidence Integrity (Non-Negotiable)

No fabricated evidence
Competitor marketing ≠ clinical evidence
Verbatim excerpts ≤ 25 words with source + date

Two-Layer Logic Model

Layer 1: Hard stops — Evaluated first, non-overridable by AI
Layer 2: Decision matrix — Used only when multiple options remain valid

📐

Schema Rule Sheet

v1.0 · 2026-02-24

One-page KG rules for the AI architect — node types, edge types, ID patterns, logic model, and audit requirements.

Node Types

EvidenceSource — citation, doi, pmid, source_type
CanonicalTruth — truth_id, statement, confidence, scope
Rule — rule_id, layer (HARD_STOP | DECISION), trigger, action
IntakeQuestion — question_id, text, answer_type, logic_hooks
Score / Instrument — SHIM/IIEF-5, EHS, METS
Drug / API — tadalafil, vardenafil, sildenafil, PT-141, oxytocin, apomorphine
ProductSKU — sku_id, actives, dose, form_factor, tier
Competitor — brand, jurisdictions, offerings, intake_features

ID Stability Rules

Evidence: Fxx-EV###
Truth: Fxx-T###
Rule: Fxx-R### or global R-ED-...
Intake: Q-ED-###
Product: P-ED-###

IDs must not change across versions.

📖

Core README

v1.0 · 2026-02-24

Index of the 00_CORE_DOCS canonical set — purpose, conflict resolution rules, and governance status.

What Lives in Core Docs

Cross-folder "core layer" that stays relatively stable while individual folders evolve:

Vision & Goals
Decision Matrix
Rules (Data Hygiene + Canonicalization)
Architecture Overview
White Paper Draft
About Us

Conflict Resolution

Evidence strength (guideline > narrative review > competitor marketing)
Safety-first rules (clinical hard stops)
Decision matrix (tradeoffs)
Update core docs only when conflict resolved

📎

Decision Matrix Addendum

v0.1 · 2026-01-08

Competitor-derived intake and verification patterns that inform Decision Matrix interpretation.

Observed Competitor Patterns

"For yourself vs for someone else" hard gate (Hims)

If user indicates flow is for someone else → hard-stop + share/referral path. Reinforces regulatory defensibility and identity/patient-attestation layer need.

Identity verification (ID / SSN / driver's license)

Competitors may request identity verification artifacts during onboarding. Both a compliance tool and conversion friction point.

Disqualification messaging style

Should be handled with minimal stigma and clear next steps while remaining consistent with platform scope.

Recommendation

No weight changes yet, but requires implementing identity + self-attestation checks as policy primitives and logging all gating decisions for defensibility.

🧠

Decision Matrix Rationale

v1.1 · 2025-12-23

Defense and justification of the matrix ranking logic — why UX, AI explainability, and compounding novelty outrank operations.

Core Thesis

In stigmatized medical domains, user experience directly determines data quality, which determines clinical outcomes.

Key Defenses

1. Experience Dimensions Rank High

UX is not cosmetic—it is clinical infrastructure. A patient who feels judged will not disclose substance use, anxiety triggers, or relationship dynamics.

2. AI_Explainability (#4) outranks Decision_Inference_Power (#13)

California telehealth law requires defensible clinical rationale. Inference without explainability is liability.

3. Novelty_Signal (#12) rose — compounding vertical integration

Compounding is the economic moat. If the system can't communicate "this is not a generic pill," the premium pricing justification collapses.

4. Competitive_Parity_Coverage (#25) is last

Differentiation beats imitation. Parity is hygiene, not strategy.

5. Operations rank lower than UX/AI

Operational efficiency follows product-market fit. Optimizing provider review before validating intake experience is premature optimization.

📖

AI King Bible — AbacusAI Implementation Guide

v2.0 · 2026-01-15 · 4,451 lines

Comprehensive production-ready implementation guide for AbacusAI Workflows, DeepAgents, AI Studio, RAG, Knowledge Graph, guardrails, and Cloudflare Workers deployment. Platform architecture blueprint for the AI-powered compounding pharmacy.

System Architecture — 5-layer architecture: Patient Touchpoints → Edge Gateway → Workflows Orchestrator → DeepAgent Controllers → Knowledge + Customer Data layers
DeepAgents — 4 agent types: Sales Qualification, Medical Intake (MI-based), Compound Recommendation, Behavior Change
Knowledge Graph & RAG — Document ingestion, entity extraction, vector embedding, hybrid retrieval (semantic + keyword + graph)
Guardrails — Medical accuracy validation, output quality, PII protection, HIPAA compliance
Multimodal UX — Quadrants, sliders, radar charts, voice, avatar agent
Workflows — Foundation (KB), Agent Dev, Quality, Data Management, Experimentation, Integration
Deployment — Cloudflare Workers edge deployment, 8-week roadmap

Source: AI_King_Bible/AbacusAI_Workflow_Implementation_Guide

This document serves as the platform architecture blueprint. AI agents should reference it for implementation patterns.

Knowledge Graph Schema v3.1

29 node types across 6 layers + 18 edge types. Master: 331 nodes, 561 edges. Sourced from MELTS_KG_MASTER_v3.1.json.

Node Types (29)

research Truth

▾

Canonical clinical/regulatory/competitive fact

Required: id, type, statement, confidence_level, domain

ID Pattern: ^(F\\d{2}-T\\d{3}|TRUTH-\\d{3})$

research Evidence

▾

Primary evidence source (peer-reviewed study, guideline)

Required: id, type, evidence_id, study_type, findings, source_document, domain

ID Pattern: ^(F\\d{2}-(E|EV)\\d{3,4}|EVID-.+)$

research Competitor

▾

Competitor brand profile (market facts only)

Required: id, type, company_name, domain_focus

ID Pattern: ^COMP-.+$

operational Rule

▾

Clinical or regulatory decision rule

Required: id, type, rule_id, layer, trigger, action, rationale, domain

ID Pattern: ^(F\\d{2}-R(ULE)?-?\\w+|R-ED-.+)$

operational IntakeQuestion

▾

Intake form question for patient assessment

Required: id, type, question_id, text, answer_type, phase, domain

ID Pattern: ^Q-ED-\\d{3}$

operational RegulatoryConstraint

▾

Regulatory requirement or constraint

Required: id, type, constraint_id, jurisdiction, citation, requirement, domain

ID Pattern: ^(F\\d{2}-RC\\d{3}|RC_.+)$

operational Accreditation

▾

Pharmacy/platform accreditation or certification

Required: id, type, accreditation_id, name, issuer, domain

ID Pattern: ^ACCRED_.+$

clinical ValidatedScore

▾

Validated clinical assessment instrument

Required: id, type, score_type, name, range, interpretation_bands, domain

ID Pattern: ^SCORE-.+$

clinical RiskFactor

▾

Patient risk factor category

Required: id, type, risk_category, risk_description, severity, domain

ID Pattern: ^RISK-.+$

clinical MedicationExposure

▾

Medication class for contraindication checking

Required: id, type, medication_class, contraindication_severity, domain

ID Pattern: ^MED-.+$

product ProductStack

▾

Product stack (e.g. Edge, Velocity, Connection)

Required: id, type, stack_id, public_name, tagline, domain_focus, launch_phase

ID Pattern: ^STACK-S\\d_.+$

product StackTier

▾

Specific tier within a product stack

Required: id, type, tier_id, stack_id, tier_name, ingredients

ID Pattern: ^TIER-S\\d-.+$

product ActiveIngredient

▾

Active pharmaceutical ingredient

Required: id, type, generic_name, drug_class, mechanism_of_action, primary_cyp_pathway

ID Pattern: ^API-.+$

product FormFactor

▾

Dosage form (Mini-Melt, NovaFilm, Troche)

Required: id, type, name, description, max_api_load_mg

ID Pattern: ^FORM-.+$

safety SafetyGate

▾

Safety screening gate in intake process

Required: id, type, gate_id, name, gate_type, scope, trigger_question

ID Pattern: ^GATE-G\\d+.*$

safety GateDecision

▾

Decision outcome from safety gate evaluation

Required: id, type, gate_id, condition, action, severity

ID Pattern: ^GDEC-.+$

safety DrugInteraction

▾

Drug-drug interaction rule

Required: id, type, api_id, interacting_class, severity, action

ID Pattern: ^DI-.+$

safety EscalationPath

▾

Provider escalation protocol

Required: id, type, tier, trigger_condition, response_time, action

ID Pattern: ^ESC-.+$

algorithm ScoringDimension

▾

Scoring dimension in recommendation algorithm

Required: id, type, dim_id, name, weight, description, key_inputs

ID Pattern: ^DIM-D\\d$

algorithm PatientArchetype

▾

Patient archetype with expected recommendation

Required: id, type, name, profile_summary, expected_stack, expected_tier, confidence

ID Pattern: ^ARCH-\\d{2}$

algorithm Pathway

▾

Clinical pathway triggered by intake signals

Required: id, type, name, trigger_conditions, primary_stacks

ID Pattern: ^PATH-.+$

algorithm Comorbidity

▾

Comorbidity affecting stack selection/dosing

Required: id, type, name, category, prevalence_in_ed, stack_impact

ID Pattern: ^COMORB-.+$

algorithm AlgorithmScorer

▾

Scoring engine (Apo 13pt, PT-141 12pt, OT 16pt)

Required: id, type, name, max_points, threshold_prescribe, routes_to

ID Pattern: ^ALGO-.+$

algorithm QuizArchetype

▾

Spark Quiz archetype with clinical routing

Required: id, type, name, emoji, threshold, clinical_routing

ID Pattern: ^ARCH-QUIZ-.+$

clinical IntakeInstrument

▾

Clinical assessment instrument (Spark Score)

Required: id, type, name, dimensions, output, archetypes

ID Pattern: ^QUIZ-.+$

clinical ClinicalEvidence

▾

Published clinical study with PK/efficacy data

Required: id, type, title, authors, journal, year, key_finding, evidence_grade

ID Pattern: ^EVIDENCE-.+$

product DigitalTherapy

▾

Digital therapy product (Melts.la Mind)

Required: id, type, name, modules, tiers, regulatory_status

ID Pattern: ^PRODUCT-MIND$

operational AcquisitionChannel

▾

GTM acquisition channel

Required: id, type, name, target_subs, cac_range

ID Pattern: ^CHANNEL-.+$

operational RegulatoryEvent

▾

Regulatory enforcement event (FDA actions)

Required: id, type, name, date, description, impact

ID Pattern: ^REGULATORY-.+$

safety DoseAdjustment

▾

Dose ceiling or adjustment due to interaction/comorbidity

Required: id, type, condition, adjustment, max_dose

ID Pattern: ^DADJ-.+$

operational IntakeSection

▾

Intake form section grouping questions

Required: id, type, section_id, name, order, branching

ID Pattern: ^SEC-.+$

Edge Types (44)

Edge Type	Direction	Description
`SUPPORTED_BY`	Truth → Evidence	Truth is supported by evidence
`DERIVED_FROM`	Rule → Truth \| Evidence	Rule is derived from truth or evidence
`REQUIRES_INPUT`	Rule \| SafetyGate → IntakeQuestion	Requires intake question input
`CONSTRAINS`	RegulatoryConstraint → Rule \| IntakeQuestion \| ProductStack	Regulatory constrains
`CONTRAINDICATES`	DrugInteraction → MedicationExposure	Contraindicates medication class
`ASSESSES`	ValidatedScore → RiskFactor	Score assesses risk factor
`TRIGGERS_GATE`	IntakeQuestion \| RiskFactor → SafetyGate	Triggers a safety gate
`CONTAINS_INGREDIENT`	ProductStack → ActiveIngredient	Stack contains ingredient
`HAS_TIER`	ProductStack → StackTier	Stack has tier
`USES_FORMAT`	StackTier → FormFactor	Tier uses dosage form
`INTERACTS_WITH`	ActiveIngredient → DrugInteraction	API has drug interaction
`GATES_STACK`	SafetyGate → ProductStack	Gate evaluates stack eligibility
`EXCLUDES_STACK`	GateDecision → ProductStack	Gate excludes a stack
`CAPS_DOSE`	GateDecision \| DoseAdjustment → StackTier	Caps dose to max tier
`REQUIRES_ESCALATION`	GateDecision → EscalationPath	Requires provider escalation
`SCORED_AGAINST`	ProductStack → ScoringDimension	Stack scored against dimension
`RECOMMENDED_FOR`	ProductStack → PatientArchetype	Stack recommended for archetype
`ROUTES_TO`	Pathway \| Scorer \| Archetype → ProductStack	Pathway/scorer/archetype routes to stack
`DELIVERED_AS`	ProductStack → FormFactor	Stack delivered in dosage form
`CONTAINS_API`	ProductStack → ActiveIngredient	Stack contains API with dose
`BUNDLES`	CouplesStack → ProductStack	Couples bundle includes component
`SYNERGY`	ActiveIngredient → ActiveIngredient	API synergy relationship
`FEEDS_SCORER`	IntakeInstrument → AlgorithmScorer	Quiz feeds scorer via dimension
`FEEDS_DIMENSION`	IntakeInstrument → ScoringDimension	Quiz feeds scoring dimension
`EVIDENCE_FOR`	ClinicalEvidence → ActiveIngredient	Study provides evidence for API
`COMPETES_WITH`	Competitor → ProductStack	Competitor product competes with stack
`BLOCKS`	RegulatoryEvent → ProductStack	Regulatory event blocks product
`EXEMPT_FROM`	ProductStack → SafetyGate	Stack exempt from safety gate
`REQUIRED_FOR`	FormFactor → ActiveIngredient	Form factor required for API
`INCOMPATIBLE_WITH`	FormFactor → ActiveIngredient	Form factor incompatible with API
`PREFERRED_FOR`	FormFactor → ActiveIngredient	Preferred form for API (higher BA)
`PRIMARY_PRODUCT`	AcquisitionChannel → ProductStack	Channel's primary product
`SECONDARY_PRODUCT`	AcquisitionChannel → ProductStack	Channel's secondary product
`DRIVES_TRAFFIC_TO`	AcquisitionChannel → IntakeInstrument	Channel drives traffic to quiz/intake
`BOOSTS_AFFINITY`	ScoringDimension → ProductStack	Dimension score boosts stack affinity
`MODIFIES_DOSE`	Comorbidity → StackTier	Comorbidity modifies dosing
`DETECTS_MEDICATION`	IntakeQuestion → MedicationExposure	Question screens for medication
`JUSTIFIES_STACK`	Truth → ProductStack	Truth justifies stack inclusion
`JUSTIFIES_INGREDIENT`	Truth → ActiveIngredient	Truth justifies ingredient use
`ENFORCES_RULE`	SafetyGate → Rule	Gate enforces clinical rule
`JUSTIFIES_FORMAT`	Truth → FormFactor	Truth justifies form factor choice
`CONTAINS_QUESTION`	IntakeSection → IntakeQuestion	Section contains question
`INSTRUMENT_MEASURES`	ValidatedScore → IntakeQuestion	Instrument measures via question
`CONTRIBUTES_TO_SCORE`	IntakeQuestion → ScoringDimension	Question contributes to score

Diagrams

System architecture, ingestion workflow, data flow, and product entity model — rendered from Mermaid.

System Architecture

2026-01-05 · Technical Architect

graph TB subgraph SL["SOURCE LAYER"] SR["SOURCE_ROOT
Workspace Root"] EG["ED_GEM_ROOT
Structured Synthesis Vault"] SR --> F01_SR["F01-F14 Folders
Raw Research"] EG --> F01_EG["F01-F19 Folders
Canonical Deliverables"] EG --> F00["F00_CANONICAL_CORE
Governance Layer"] end subgraph IE_SUB["INGESTION ENGINE"] IE["MELTS_INGESTION_ENGINE"] --> SCAN["Dual-Root Scanner"] SCAN --> PARSE["Deliverables Parser"] PARSE --> NORM["ID Normalizer"] NORM --> FILTER["Ultra-Lean Filter"] FILTER --> DEDUPE["Deduplication Engine"] DEDUPE --> WRITE["Output Writer"] end subgraph KG_SUB["KNOWLEDGE GRAPH STORE"] KG["knowledge_graph_export.json"] KG --> NODES["Nodes: Truth, Evidence, Rule,
Competitor, Accreditation"] KG --> EDGES["Edges: supported_by, enforces,
constrains, includes, refines"] end subgraph WF_SUB["WORKFLOW ENGINE"] WF["ABACUS WORKFLOWS"] WF --> INTAKE["Intake Workflow"] WF --> TRIAGE["Triage Workflow"] WF --> DECISION["Decision Workflow"] WF --> FULFILLMENT["Fulfillment Workflow"] end F01_EG --> IE F00 --> IE IE --> KG KG --> WF style F00 fill:#FFE6CC,stroke:#D79B00 style KG fill:#D5E8D4,stroke:#82B366 style WF fill:#E1D5E7,stroke:#9673A6

Canonical Ingestion Workflow

2026-01-05 · Data Engineers

flowchart TD START(["START: New Folder Ingestion"]) --> INIT["Initialize Run Context
run_id, timestamp, schema"] INIT --> SCAN{"Dual-Root Scan
SOURCE_ROOT + ED_GEM_ROOT"} SCAN -->|Both exist| USE_EG["Use ED_GEM_ROOT
Preferred Authority"] SCAN -->|Only SOURCE| USE_SR["Use SOURCE_ROOT
Fallback"] SCAN -->|Neither| ERR_NF["ERROR: Not Found
STOP"] USE_EG --> CHECK_DEL{"Check _DELIVERABLES/"} USE_SR --> CHECK_DEL CHECK_DEL -->|EXISTS| PARSE["Parse Truths, Rules,
Evidence, Competitors"] CHECK_DEL -->|MISSING| ERR_ND["ERROR: No Deliverables
STOP"] PARSE --> NORMALIZE["Normalize IDs
Schema Compliance"] NORMALIZE --> VALIDATE{"Schema
Validation"} VALIDATE -->|PASS| FILTER["Ultra-Lean Filter
Remove Noise"] VALIDATE -->|FAIL| ERR_SC["ERROR: Schema Violation"] FILTER --> DEDUPE["Deduplication
vs Master Registers"] DEDUPE --> BUILD["Build Knowledge Graph
Nodes + Edges"] BUILD --> OUTPUT["Write Outputs
KG + Registers + Context"] OUTPUT --> SYNC["Sync Master Registers"] SYNC --> REPORT["Generate Report"] REPORT --> DONE{"More folders?"} DONE -->|NO| FINISH(["END: Complete"]) DONE -->|YES| AWAIT["AWAIT: Continue Command"] AWAIT --> START style START fill:#D5E8D4,stroke:#82B366 style FINISH fill:#D5E8D4,stroke:#82B366 style ERR_NF fill:#F8CECC,stroke:#B85450 style ERR_ND fill:#F8CECC,stroke:#B85450 style ERR_SC fill:#F8CECC,stroke:#B85450 style FILTER fill:#E1D5E7,stroke:#9673A6 style DEDUPE fill:#DAE8FC,stroke:#6C8EBF

Platform Data Flow

2026-01-08 · System Overview

flowchart LR A["Patient UI
(Melts.la)"] --> B["Intake Orchestrator
(branching + scoring)"] B --> C["Policy Engine
Hard Stops + Decision"] C -->|Eligible options| D["AI Support
(RAG bounded)"] D --> E["Clinician Review
(PC)"] E -->|Approved Rx| F["Pharmacy Fulfillment
(503A)"] F --> G["Shipping / Tracking"] E --> H["Patient Education"] B --> I["Audit Log
(append-only)"] C --> I D --> I E --> I

Product Architecture — Entity Model

v0.1 · Product Schema

classDiagram class Catalog { +schema_version +catalog_id } class Jurisdiction { +country +state +care_model } class PlatformConstraint { +platform_key +route +max_api_load_mg } class Ingredient { +ingredient_id +class +synonyms } class MenuProduct { +product_id +display_name +category } class Tier { +tier_id +label } class Dose { +ingredient_id +dose_mg } class AdjunctModule { +module_id +display_name } Catalog "1" o-- "1" Jurisdiction Catalog "1" o-- "*" PlatformConstraint Catalog "1" o-- "*" Ingredient Catalog "1" o-- "*" MenuProduct MenuProduct "1" o-- "*" Tier Tier "1" o-- "*" Dose Dose "*" --> "1" Ingredient Catalog "1" o-- "*" AdjunctModule

Intake Decision Flow — Safety Gates

2026-03-20 · Clinical Workflow

flowchart TD START(["START: Patient Intake"]) --> SELF{"G1: Self-Attestation"} SELF -->|No| STOP_G1["HARD STOP: Must be for self"] SELF -->|Yes| AGE{"G2: Age Gate 18+"} AGE -->|No| STOP_G2["HARD STOP: Under 18"] AGE -->|Yes| CV{"G3: CV Screen"} CV -->|Flag| STOP_G3["HARD STOP: CV Risk"] CV -->|Clear| HYPO{"G4: Hypotension"} HYPO -->|Flag| STOP_G4["HARD STOP: Hypotension"] HYPO -->|Clear| MED{"G5: Medication Screen"} MED -->|Interaction| CAUTION["CAUTION: Dose Limit"] MED -->|Clear| PSYCH{"G6: Psychosocial Screen"} PSYCH --> SCORE["6-Dim Score"] SCORE --> REC{"Recommendation Engine"} REC -->|PDE5i Only| S1["S1 Edge / S5 Daily"] REC -->|Desire| S2["S2 Velocity / S4 Spark"] REC -->|Connection| S3["S3 Connection"] REC -->|Multi-Mech| S8["S8 Quantum"] S1 --> DOC["Clinician Review"] S2 --> DOC S3 --> DOC S8 --> DOC CAUTION --> DOC DOC -->|Approved| RX["Pharmacy Fulfillment"] style START fill:#D5E8D4,stroke:#82B366 style STOP_G1 fill:#F8CECC,stroke:#B85450 style STOP_G2 fill:#F8CECC,stroke:#B85450 style STOP_G3 fill:#F8CECC,stroke:#B85450 style STOP_G4 fill:#F8CECC,stroke:#B85450 style CAUTION fill:#FFF2CC,stroke:#D6B656 style RX fill:#D5E8D4,stroke:#82B366 style DOC fill:#E1D5E7,stroke:#9673A6 style SCORE fill:#DAE8FC,stroke:#6C8EBF

Product Catalog

21+ product stacks across 7 categories — ED, Couples, TRT, Wellness, Hair, Digital, GLP-1 (muted).

ED Stacks (S1–S8)

Stack	Name	APIs	Tiers	Form	Price/Dose	Phase
S1	Edge	Vardenafil + Tadalafil	LITE / PRO / MAX	Mini-Melt, ODT, NovaFilm	$8–14	🟢 Launch
S2	Velocity	Vardenafil + Tadalafil + PT-141	LITE / PRO / MAX	ODT	$10–16	🟢 Launch
S3	Connection	Vardenafil + Tadalafil + Oxytocin	LITE / PRO / MAX	ODT	$10–16	🟢 Launch
S4	Spark	PT-141 + Oxytocin (NO PDE5i)	GLOW / SPARK / WILD	Mini-Melt	$10–16	🟢 Launch
S5	Daily	Tadalafil + L-Citrulline	LITE / PRO	Capsule	$69–99/mo	🟢 Launch
S6	Vitality+	Enclomiphene + Tadalafil	LITE / PRO	Capsule + Mini-Melt	$149–179/mo	🟡 Phase 2
S7	Vitality	Tadalafil + L-Citrulline + GHK-Cu	—	Capsule	$199/mo	🟡 Phase 3
S8	Quantum	Vardenafil + Tadalafil + Apomorphine + OT	Q1 / Q2	Troche (mandatory)	$18–24	🟡 Phase 2

Tier Breakdown — S1-S3 (PDE5i Stacks)

Tier	Audience	Vardenafil	Tadalafil	Adjunct	Price/Dose
LITE	First-time, mild ED, dose-cautious	5mg	5mg	Low-dose if applicable	$8–10
PRO	Standard therapeutic — majority of patients	10mg	10mg	Standard dose	$10–14
MAX	Severe ED, PDE5i partial/non-responders	15mg	15mg	Maximum dose	$14–16

Tier Breakdown — S4 Spark (No PDE5i)

Tier	Audience	PT-141	Oxytocin	Price/Dose
GLOW	Desire-curious, first-time	Low	Low	$10–12
SPARK	Standard desire pathway	Standard	Standard	$12–14
WILD	Maximum desire + bonding	High	High	$14–16

Tier Breakdown — S8 Quantum

Tier	Audience	Formulation	Price/Dose
Q1	Standard quadra-mech	Var 10 + Tad 10 + Apo 2mg + OT low	$18–20
Q2	Maximum quadra-mech	Var 15 + Tad 15 + Apo 3mg + OT standard	$20–24

Couples Bundles (C1–C3)

Bundle	His	Hers	Shared	Price
C1 Together	S1 Edge	Scream Cream	—	$99–149/mo
C2 Desire Duo	S4 Spark	Scream Cream	—	$119–169/mo
C3 Connection+	S3 Connection	S4 Spark	OT Nasal	$179–249/mo

TRT (T1–T4)

Stack	API	Price	Labs Required
T1 Optimize	Enclomiphene	$99/mo	Yes
T2 Power	Enclomiphene + Tadalafil	$149/mo	Yes
T3 Complete	Testosterone Cypionate	$199/mo	Yes + DEA
T4 Elite	T-Cyp + S1 Edge	$329/mo	Yes + DEA

Phased Categories

Status	Categories
🟢 Launch	ED (S1-S8), Couples (C1-C3), Digital (Mind)
🟡 Phase 2	TRT (T1-T4), NAD+ (W1), Methylene Blue (W2), Hair (H1)
🟡 Phase 3	GHK-Cu (W3), PE
🔴 Muted	GLP-1 (FDA enforcement active Mar 2026)

Algorithm & Scoring

6 weighted dimensions, 3 adjunct scorers, 9 safety gates, 8 patient archetypes → personalized stack ranking.

Recommendation Pipeline

Intake answers → Safety Gates (G1–G9) → 6 Weighted Dimensions → 3 Adjunct Scorers → Archetype Matching → Stack Ranking → Clinician Output JSON

6 Scoring Dimensions

ID	Dimension	Weight	Key Inputs
D1	Vascular Severity	0.25	SHIM, EHS, ED duration, cardiometabolic
D2	Psychogenic Profile	0.20	Situational pattern, morning erections, anxiety
D3	Desire & Libido	0.15	Sexual thought frequency, initiative pattern
D4	Safety Compatibility	0.20	Gate pass/fail, interaction severity
D5	Patient Preference	0.10	Equalizer slider values, goal selection
D6	Treatment History	0.10	Prior PDE5i use, response, side effects

3 Adjunct Scorers

Scorer	Max Points	Threshold	Routes To
Apomorphine Scorer	13	≥5	S8 Quantum
PT-141 Scorer	12	≥6	S2 Velocity, S4 Spark
Oxytocin Scorer	16	≥10	S3 Connection, S4 Spark

Safety Gates (G1–G9)

Gate	Scope	Description
G1	Universal	Cardiovascular hard-stop (nitrates, MI, stroke)
G2	Universal	Blood pressure & hemodynamic screen
G3	Universal	Medication interaction matrix
G4	Universal	Hepatic & renal clearance
G5	Universal	Age verification (18-75)
G6	Universal	Priapism risk (sickle cell, leukemia)
G7	S2, S4, S8	PT-141 nausea tolerance
G8	S8	5-HT3 antagonist × Apomorphine
G9	W2	SSRI × Methylene Blue (serotonin syndrome)

Clinical Evidence

Published bioavailability data and efficacy studies for all novel APIs: PT-141, Oxytocin, and Apomorphine.

Bioavailability — Route Comparison

API	SL Melt BA	Nasal BA	SC BA	Key Reference
PT-141	⚠️ Poor (no PK data)	❌ Abandoned	✅ 100% (FDA)	Diamond 2004, Kingsberg 2019
Oxytocin	~4.5%	~11%	100%	Martins 2024 (N=37, crossover)
Apomorphine	✅ ~17%	N/A	100%	Heaton 2001, Hauser 2020
Vardenafil	✅ Excellent	N/A	N/A	Standard SL delivery
Tadalafil	✅ Excellent	N/A	N/A	Standard SL/oral

Key Published Studies

Study	Year	API	Key Finding	Grade
Diamond LE et al. J Urol	2004	PT-141	SC 0.3-10mg: erections in 17/20 including PDE5i non-responders	II
Kingsberg SA et al. Obstet Gynecol	2019	PT-141	Phase III RECONNECT: SC 1.75mg improved FSFI-D for HSDD	I
Martins DA et al. Psychoneuroendocrinol	2024	Oxytocin	SL BA 4.45% vs nasal 11.07% but same functional effects	II
Behnia B et al. JCEM	2014	Oxytocin	IN OT improved orgasm parameters in males	II
Heaton J. World J Urol	2001	Apomorphine	SL 2-3mg: 71% successful intercourse (meta-analysis 5 RCTs)	I
Hauser RA et al. Lancet Neurol	2020	Apomorphine	SL film ~17% BA, 83% ON state in 45 min	I

GTM Strategy

4-channel Anti-Hims playbook targeting 300 LA subscribers in 120 days with local-first, premium positioning.

LA Launch — 300 Subscribers in 120 Days

Channel	Target	CAC	Primary Product
B2B Referrals (barbershops, trainers, medspas)	50	$36–72	S1 Edge, T1 Optimize
Hyper-Local Search & SEO	100	$100–300	S1 Edge
Micro-Influencer Seeding	75	$100–150	S4 Spark
Spark Quiz Viral Loop	75	~$0	C1 Together, C2 Desire Duo

Unit Economics

Metric	Target
Break-even	~180 subs ($20K MRR)
Target MRR at 300	$30–38K
LTV:CAC	≥3:1
Payback period	≤60 days

"Anti-Hims" Positioning

Same-day local delivery, vertically integrated compounding, premium/concierge. Zero Meta/TikTok bidding wars.

Competitive Intelligence

8-competitor analysis with pricing, vulnerability mapping, and regulatory alert tracking.

Competitive Landscape

Competitor	Products	Price/Dose	Our Advantage	Their Vulnerability
Hims (47% share)	Generic PDE5i, Hard Mints	$4–8	Adjuncts, AI algo, SL format	Generic commodity
Ro (Roman) (15%)	Tablets, Sparks SL, Gummies	$3–12	Multi-mechanism stacks	No adjuncts beyond PDE5i
Rugiet	Rugiet Ready (Sil+Tad+Apo)	$10–24	OT bonding, AI intake	No OT, no personalization
BlueChew	PDE5i chewables	$2–3	Everything (adjuncts, forms, AI)	Zero differentiation
Strut (GoodRx)	4-in-1 (Tad+Var+OT+Apo)	$12–18	AI-guided personalization	Regulatory risk
Maximus	Enc+Tad combo	$70–400/mo	Lower price, SL format	Expensive, no innovation
HelloCake	Sil+Tad combo	$8–12	Couples vertical, AI	FDA Warning Letter 2025
MangoRx	Tad+Sil+L-arg+OT RDT	N/A	Full stack diversity	$467K revenue, declining

Regulatory Alerts

🚨 Mar 3, 2026: FDA issued 30 warning letters to GLP-1 telehealth companies including SkinnyRx. All compounded semaglutide/tirzepatide forms affected. GLP-1 lanes remain MUTED.

Interactive Tools

Standalone interactive pages — click to open in a new tab.

🌳

Operational Workflows

24 production workflows in .gemini/workflows/ — executed via slash commands.

💡 Run any workflow with /command-name in the AI assistant.

Core Operations

⚙️ /curate-kb

Master KB curation loop — Scan → Curate → Codify → Apply → Score → Audit → Report

.gemini/workflows/curate-kb.md

⚙️ /prune-kb

Remove stale, duplicate, and low-quality KB entries

.gemini/workflows/prune-kb.md

⚙️ /clean-docs

Clean and standardize documentation files

.gemini/workflows/clean-docs.md

⚙️ /finish-sprint

Complete remaining sprint tasks and update roadmap

.gemini/workflows/finish-sprint.md

Intelligence & Research

🔍 /scan-la-events

Scan LA event venues and wellness events for partnership opportunities

.gemini/workflows/scan-la-events.md

🔍 /scan-medical-journals

Scan PubMed and medical journals for new ED research

.gemini/workflows/scan-medical-journals.md

🔍 /scrape-competitors

Scrape competitor websites for pricing, product, and UX intelligence

.gemini/workflows/scrape-competitors.md

🔍 /auto-scrape-competitors

Automated scheduled competitor scraping

.gemini/workflows/auto-scrape-competitors.md

🔍 /research-conversion

Deep research on conversion psychology and optimization

.gemini/workflows/research-conversion.md

Content & Marketing

📝 /generate-content

AI-powered content production for blog, social, and email

.gemini/workflows/generate-content.md

📝 /build-landing-page

Create new marketing landing pages

.gemini/workflows/build-landing-page.md

Bot & Personas

🤖 /build-telegram-bot

Build and iterate on the Cynthia Telegram bot

.gemini/workflows/build-telegram-bot.md

🤖 /upgrade-cynthia-bot

Upgrade bot with new features and refinements

.gemini/workflows/upgrade-cynthia-bot.md

🤖 /grade-bot-conversations

Grade and score bot conversation quality

.gemini/workflows/grade-bot-conversations.md

🤖 /create-persona

Create hyper-detailed clinical patient personas

.gemini/workflows/create-persona.md

🤖 /audit-personas

Audit persona accuracy and clinical fidelity

.gemini/workflows/audit-personas.md

Partners & Local

📍 /build-la-local

Build the LA Local partnership database

.gemini/workflows/build-la-local.md

📍 /update-la-local

Update and enrich LA Local database entries

.gemini/workflows/update-la-local.md

📍 /create-local-intel-tool

Build local intelligence collection tools

.gemini/workflows/create-local-intel-tool.md

📍 /manage-partners

Partner lifecycle management workflow

.gemini/workflows/manage-partners.md

📍 /nurture-leads

Lead conversion and nurturing workflow

.gemini/workflows/nurture-leads.md

📍 /upgrade-la-crm

Upgrade the LA CRM system

.gemini/workflows/upgrade-la-crm.md

Platform Build

🔧 /build-compound-library

Build interactive compound monograph pages

.gemini/workflows/build-compound-library.md

🔧 /build-outcome-tracking

Build patient outcome tracking system

.gemini/workflows/build-outcome-tracking.md

🔧 /build-provider-dashboard

Build provider-facing clinical dashboard

.gemini/workflows/build-provider-dashboard.md

🔧 /build-docs-page

Build or update interactive docs site pages

.gemini/workflows/build-docs-page.md

Knowledge Base

Full directory structure for both the root workspace and the ED Gem structured synthesis vault.

📦 Root Workspace

37 folders

▾ 📁 00_ED_Decision_Framework

• (6 items)

▾ 📁 01_Erectile_Dysfunction_ED

• (3 items)

▾ 📁 01_Framework_Research

• (7 items)

▾ 📁 02_Chemical

• (14 items)

▾ 📁 03_Drug_Research

• (24 items)

▾ 📁 04_Behavior_Psychology

• (4 items)

▾ 📁 05_Chemical_Related_Psychology

• (2 items)

▾ 📁 06_Competition

• (33 items)

▾ 📁 07_Products

• (21 items)

▾ 📁 08_Intake_Qualifier_PQ

• (5 items)

▾ 📁 09_Current_CUJ_Patient_Journey

• (4 items)

▾ 📁 10_Dream_VIP_Experience

• (3 items)

▾ 📁 11_Successful_Business_Models

• (4 items)

▾ 📁 12_Music_AI_Tech_Innovation

• (2 items)

▾ 📁 13_Telemedicine_Intake_Forms

• (18 items)

▾ 📁 14_Laws_and_Policies

• (12 items)

▾ 📁 15_Formulas_and_Formulation

• (4 items)

▾ 📁 16_BioHacker_Integrations

• (5 items)

▾ 📁 17_Wellness_and_Nootropics

• (6 items)

▾ 📁 18_Digital_Behavioral_Profiling

• (3 items)

▾ 📁 ABACUS_WORKFLOW_STARTER_KIT

• (15 items)

▾ 📁 AI_King_Bible

• (8 items)

▾ 📁 ED_KB_AUDITS

• (5 items)

▾ 📁 ED_Knowledge_Bases

• (53 items)

▾ 📁 MELTS_Implementation

• (9 items)

▾ 📁 MELTS_PLATFORM_DELIVERABLES

• (187 items)

▾ 📁 SYNTH

• (15 items)

▾ 📁 Uploads

• (57 items)

💎 ED Gem Vault

31 folders

▾ 📁 00_CANONICAL_CORE

📄 CORE_ABOUT_US_v1.0.md

📄 CORE_ARCHITECTURE_OVERVIEW_v1.0.md

📁 CORE_DOCS_PACK_v0.1/

📄 CORE_ONE_PAGE_RULE_SHEET_v1.0.md

📄 CORE_README_v1.0.md

📄 CORE_RULES_DATA_HYGIENE_v0.1.md

📄 CORE_VISION_AND_GOALS_v2.0.md

📄 CORE_WHITE_PAPER_DRAFT_v0.1.md

📄 MELTS_PLATFORM_SCHEMA_v1.0.json

▾ 📁 _SCHEMAS

📄 MELTS_KG_SCHEMA_v3.0.json

📄 MELTS_ULTRA_HYBRID_SCHEMA_v2.0.json

▾ 📁 _KG

📄 KG_DOMAIN_ALGORITHM.json

📄 KG_DOMAIN_CLINICAL_EVIDENCE.json

📄 KG_DOMAIN_COMORBIDITY.json

📄 KG_DOMAIN_INTAKE.json

📄 KG_DOMAIN_INTERACTIONS.json

📄 KG_DOMAIN_PRODUCTS.json

📄 KG_DOMAIN_REGULATORY.json

📄 KG_DOMAIN_SAFETY.json

▾ 📁 01_Erectile_Dysfunction_ED

• (26 items)

▾ 📁 03_Drug_Research

• (42 items)

▾ 📁 06_Competition

• (103 items)

▾ 📁 07_Products

• (59 items)

▾ 📁 08_Intake_Qualifier_PQ

• (1 item)

▾ 📁 09_Current_CUJ_Patient_Journey

• (3 items)

▾ 📁 10_Dream_VIP_Experience

• (3 items)

▾ 📁 11_Successful_Business_Models

• (3 items)

▾ 📁 12_Music_AI_Tech_Innovation

• (3 items)

▾ 📁 13_Telemedicine_Intake_Forms

• (66 items)

▾ 📁 14_Laws_and_Policies

• (18 items)

▾ 📁 15_Formulas_and_Formulation

• (3 items)

▾ 📁 16_BioHacker_Integrations

• (4 items)

▾ 📁 17_Wellness_and_Nootropics

• (3 items)

▾ 📁 18_Digital_Behavioral_Profiling

• (3 items)

▾ 📁 19_Ads_Payments_Trust

• (5 items)

Core Documents

About Us

What Melts.la is

Operating Model

Clinical care (PC / clinicians)

Fulfillment (pharmacy)

Platform responsibilities

Clinical Posture

Product Posture

Core Identity

White Paper Draft

Abstract

Clinical Foundation

Formulary Strategy (Adjunct-Forward)

Foundation: PDE5 inhibitors

PT-141 (bremelanotide) — Desire/arousal adjunct

Apomorphine — Rescue/alternative lane

Oxytocin — Anxiety/connection adjunct

AI Decision Support

Decision Matrix

Overview

Pre-Layer Hard Rules (Non-Rankable)

Weight Tiers

Top 10 Dimensions

Architecture Overview

System Components

Kill Switch Pattern

KB → Policy → Product Mapping

Vision & Goals

Mission

Core Values

Goals

Data Hygiene Rules

Content Taxonomy

Canonicalization Pipeline

Evidence Integrity (Non-Negotiable)

Two-Layer Logic Model

Schema Rule Sheet

Node Types

ID Stability Rules

Core README

What Lives in Core Docs

Conflict Resolution

Decision Matrix Addendum

Observed Competitor Patterns

"For yourself vs for someone else" hard gate (Hims)

Identity verification (ID / SSN / driver's license)

Disqualification messaging style

Recommendation

Decision Matrix Rationale

Core Thesis

Key Defenses

1. Experience Dimensions Rank High

2. AI_Explainability (#4) outranks Decision_Inference_Power (#13)

3. Novelty_Signal (#12) rose — compounding vertical integration

4. Competitive_Parity_Coverage (#25) is last

5. Operations rank lower than UX/AI

AI King Bible — AbacusAI Implementation Guide

Contents

Knowledge Graph Schema v3.1

Node Types (29)

research Truth

research Evidence

research Competitor

operational Rule

operational IntakeQuestion

operational RegulatoryConstraint

operational Accreditation

clinical ValidatedScore

clinical RiskFactor

clinical MedicationExposure

product ProductStack

product StackTier

product ActiveIngredient

product FormFactor

safety SafetyGate

safety GateDecision

safety DrugInteraction

safety EscalationPath

algorithm ScoringDimension