Platform Architecture

19 microservices. Three databases. Zero fabricated scores.

RevFlow OS is a modular AI marketing platform built on Docker containers, a unified API gateway, production-grade monitoring, and a data integrity system that never presents derived data as measured facts.

19 Microservices
3 Database Engines
37 Total Containers
100% Docker Containerized

RevFlow OS Architecture Stack

Every module is independently deployable. All inter-module communication routes through RevCore. Every data point carries a provenance label.

Suite I: Lead Gen 13 modules: RevPrompt, RevScore IQ, RevRank, RevSEO, RevCite, RevHumanize, RevWins, RevImage, RevPublish, RevMetrics, RevSignal, RevIntel, RevDispatch
Suite II: Digital Landlord 2 modules: RevVest IQ (asset valuation) and RevSPY (competitor keyword intelligence and gap analysis)
Suite III: Tech Efficiency 4 modules: RevSpend IQ, RevCore (API gateway), RevAssist (AI support), RevLocal Grid (GBP + local SEO)
Suite IV: Virtual Composite RevGBP: orchestrates Modules 4, 5, and 19 for comprehensive Google Business Profile optimization

How Modules Connect

All inter-module communication follows a strict routing pattern through RevCore. Direct module-to-module calls are architecturally prohibited.

RevCore API Gateway

Central nervous system of the platform. Handles request routing, authentication enforcement via SuperTokens, per-client rate limiting, and end-to-end request tracing with unique trace IDs. Every external and internal request passes through RevCore.

Port 8950

Docker Network

All containers join a shared Docker network with DNS-based service discovery. Containers reach each other by name without exposing inter-service ports to the host. Only API-serving ports are mapped externally.

revflow-network

Container Strategy

Each module runs in its own container with isolated filesystem and process space. Tightly-coupled modules share consolidated containers to reduce overhead while maintaining logical separation. All use revflow/python-base:latest.

unless-stopped

Health Checks

Every container probes its module's /health endpoint on 30-second intervals. Three consecutive failures mark the container unhealthy. Prometheus + AlertManager triggers alerts on unhealthy states.

30s interval

Three Database Engines

Each engine serves a distinct purpose. New development uses PostgreSQL exclusively. JSON file storage is legacy and prohibited for new features.

PostgreSQL

Single source of truth for all structured data: assessments, K-scores, Trinity calculations, content catalogs, rule evaluations, keyword data, and audit logs. All 19 modules read and write to the same instance.

Primary

ChromaDB

Vector storage for RAG operations. Modules 4, 5, and 15 use ChromaDB for semantic similarity search across content passages, competitor pages, and citation contexts. 180-day rolling retention.

Vector DB

Redis

Response caching and session management. Expensive LLM probes and DataForSEO API responses are cached with configurable TTLs. SuperTokens session data provides fast validation without database round-trips.

Cache Layer

Data Retention Policy

All DataForSEO data retained minimum 90 days in PostgreSQL. Assessment results retained indefinitely for historical trending. Redis TTLs range from 15 minutes (real-time metrics) to 24 hours (API caching).


Monitoring Stack

Production-grade monitoring operates independently from application modules. If a module fails, monitoring continues reporting.

Prometheus

Scrapes metrics from every container on 15-second intervals: request latency (p50/p95/p99), error rates, DB query performance, cache hit ratios, and custom business metrics like assessments completed and probes executed.

Grafana

Visual dashboard layer with pre-configured views organized by suite and concern. Module health, request flow patterns, database performance, cache efficiency, and resource utilization trends all in one place.

AlertManager

Routes alerts from Prometheus to notification channels. Categories: module health (container down), performance degradation (latency spikes), resource exhaustion (disk/memory), and data integrity violations.


MSP v3.1 Part 12: Zero Fabrication

Every score carries a provenance label. Every output includes a credibility envelope. Silence is better than fabrication.

LIVE

Real API call this run, call_id logged. Highest confidence tier: the measurement was taken now, from the actual source.

CACHED

Prior run data with original timestamp. Valid but not current. Consumers evaluate whether data age is acceptable.

ESTIMATE

Derived from other data using a disclosed formula. Consumers can evaluate methodology and confidence independently.

UNAVAILABLE

Data source failed. Returns NULL with failure explanation. No fallback values, no substitution, no "best guess."

Critical Integrity Violations Prevented

Invented Methodology (computing scores without data), Silent Fallback (substituting defaults on failure), Forbidden Phrases (unsupported generalizations), and RevProof Unreachable (auto-passing when quality gate is down). The pipeline halts rather than fabricate.


Security Architecture

Defense-in-depth across every layer. Security is built in, not bolted on.

SuperTokens Auth

Session management, token rotation, CSRF protection, and MFA. Runs in its own container with isolated auth state. RevCore validates every session before requests reach modules.

TLS 1.3 + Immutable Config

All external traffic encrypted via TLS 1.3 through nginx. Critical config files protected with filesystem immutable flags. API keys stored in a single shared .env mounted read-only.


See the Platform in Action

Request a technical demo. See the microservices architecture, monitoring dashboards, and data integrity system working on live client data.

The GEO/AEO/EVS Trinity AI Visibility Monitoring AEO+ Optimization AI Transparency