Skip to main content
Home Architecture
100% Rust ARM64 Native Axum + Actix

19 Rust Microservices. Zero Node.js.

Every service explained — what it does, how it’s built, and why it matters.

From the API gateway to the publishing worker, V100 is a single Rust workspace compiled into 19 independent binaries. No interpreters. No garbage collectors. No cold starts. Just native ARM64 machine code on Graviton.

System Overview

19 services grouped by function. The gateway routes every request. Services communicate via internal HTTP and Redis pub/sub.

v100-gateway :4000

JWT auth, rate limiting, circuit breaker, cache

video-processing
:3002
video-analysis
:3020
media-processor
:3013
recording-proc
:3012
ai-orchestration
:3001
rag-engine
:3011
meeting
:3005
meeting-signaling
:3015
realtime
:3014
stream-proxy
:3016
transcription
:3003
avatar
:3004
scheduling
:3006
integration
:3009
social-publisher
:3025
publishing-worker
bg
billing
:8085
billing-engine
:8086
Video Pipeline
AI & Intelligence
Real-Time & Comms

Core Infrastructure

The foundation: gateway, billing, and credit management

:4000

v100-gateway

API gateway and front door for all 18 downstream services. Every request hits this first.

  • JWT and API key authentication with per-tenant isolation
  • Tiered cache: L1 in-process + L2 Redis with request coalescing
  • Rate limiting by billing tier, circuit breaker for downstream
  • Gzip and Brotli response compression
Axum 0.7 Redis Tower
:8085

v100-billing

Credit-based billing service. Provisions and deducts credits in real time with full audit trail.

  • Credit provisioning and real-time deduction
  • Stripe Checkout integration for credit top-ups
  • Full transaction history with filterable audit log
Axum 0.7 PostgreSQL Stripe
:8086

v100-billing-engine

Advanced billing engine for subscriptions, metered usage, and revenue sharing across partners.

  • Monthly/annual subscriptions with overage calculation
  • Metered usage tracking with custom rate cards
  • Revenue sharing splits and pay-per-event charges
  • Cron-based invoice generation with Stripe sync
Axum 0.7 PostgreSQL Stripe Cron

Video Processing Pipeline

Upload, transcode, analyze, and extract — all in Rust with FFmpeg

:3002

v100-video-processing

Core video service. Handles the full lifecycle from chunked upload to platform export.

  • Chunked S3 upload with resume support
  • Project management: video_projects → video_highlights → video_clips
  • FFmpeg transcode, thumbnail generation, auto-trim, smart-crop
  • Clip generation with export to YouTube, TikTok, Instagram
Axum 0.7 PostgreSQL S3 FFmpeg
:3020

v100-video-analysis

AI video intelligence engine. Detects scenes, faces, speech segments, and scores clips for viral potential.

  • Scene detection via FFmpeg scdet, face detection (YCbCr + flood-fill + NMS + IoU tracking)
  • Audio analysis: speech/silence segmentation
  • AI classification via Claude API with 6-factor viral scoring
  • 16-worker processor pool with keyframe extraction
Axum 0.7 FFmpeg 6.1 Claude API
:3013

v100-media-processor

Low-level media operations. Zero-copy assembly, caption rendering, and audio normalization.

  • Zero-copy chunk assembly via Linux copy_file_range
  • Caption generation in SRT, VTT, and JSON formats
  • Language detection via whatlang, audio resampling/normalization
  • FFmpeg transcode orchestration with per-platform specs
Axum 0.7 FFmpeg whatlang
:3012

v100-recording-processor

Meeting recording pipeline. Streams uploads to S3 and transcodes to MP4 for instant playback.

  • S3 multipart upload with streaming (500MB limit)
  • FFmpeg WebM-to-MP4 transcode pipeline
  • Thumbnail generation at 5s mark, presigned download URLs
  • Real-time Redis event publishing for progress tracking
Axum 0.7 S3 FFmpeg Redis

AI & Intelligence

LLM orchestration, agent management, and semantic search with RAG

:3001

v100-ai-orchestration

The AI brain. Manages agents, knowledge bases, and multi-model chat completion. Powers the Training Sandbox.

  • Agent CRUD with PostgreSQL-backed persistence
  • Agent testing with real LLM completion (Anthropic Claude / Google Gemini)
  • Knowledge base management with document ingestion
  • OpenAI-compatible streaming chat endpoint, rate-limited per user
Axum 0.7 PostgreSQL Claude API Gemini Redis
:3011

v100-rag-engine

Semantic search and retrieval-augmented generation. Turns your documents into AI-ready context.

  • Document ingestion with tiktoken BPE text chunking
  • Embedding generation via OpenAI text-embedding-3-small
  • Cosine similarity search in pgvector
  • Full RAG pipeline: retrieve top-k chunks → Claude synthesis
Axum 0.7 pgvector OpenAI Claude API

Communication & Real-Time

Meetings, WebRTC signaling, live state streaming, and media proxying

:3005

v100-meeting

Meeting management service. Handles the full meeting lifecycle from creation to recording.

  • Meeting CRUD with participants, breakout rooms, waiting rooms
  • Recording lifecycle management (start, stop, retrieve)
  • Redis pub/sub for real-time event broadcasting
  • WebSocket support for live participant updates
Axum 0.7 PostgreSQL Redis WebSocket
:3015

v100-meeting-signaling

WebRTC signaling server. Connects peers for video/audio with minimal overhead. 2.3MB binary.

  • JWT-authenticated WebSocket connections
  • Room-based signaling for SDP offer/answer exchange
  • ICE candidate relay with Trickle ICE support
Axum 0.7 WebSocket JWT
:3014

v100-realtime

Real-time data streaming service. Pushes live room state and participant updates over WebSockets.

  • WebSocket server for live room state and participant lists
  • Active room monitoring and presence tracking
  • Push-based updates with minimal polling
Actix-Web WebSocket Redis
:3016

v100-stream-proxy

Media stream proxy for HLS delivery and avatar session routing. Health-checked Redis connectivity.

  • HLS streaming with adaptive bitrate support
  • Avatar session proxying for real-time video avatars
  • Media route management with health-checked backends
Axum 0.7 Redis HLS

Transcription & Language

Multi-provider transcription with streaming sessions and speaker diarization

:3003

v100-transcription

Multi-provider transcription service. Streaming and file-based with speaker diarization and word-level timestamps.

  • Streaming sessions: start, stop, status, and retrieve transcript
  • File and URL transcription with async processing
  • Deepgram (nova-2) and Whisper dual-backend support
  • Speaker diarization with word-level timestamps
Axum 0.7 Deepgram Whisper WebSocket

Avatar & Demo Agents

AI avatar service with multi-provider support for automated demos and presentations

:3004

v100-avatar

AI avatar service with multi-provider support. Create, start, and manage real-time avatar video sessions.

  • Multi-provider: HeyGen + D-ID with unified API
  • Avatar listing, voice listing, session management (create/start/stop/extend)
  • Credit checking before session creation
  • WebSocket streaming for real-time avatar video output
Axum 0.7 HeyGen D-ID WebSocket

Scheduling & Integration

Calendar sync, CRM connections, webhooks, and SMS via Twilio

:3006

v100-scheduling

Calendar integration service. Syncs with Google and Microsoft calendars for intelligent scheduling.

  • Google Calendar + Microsoft Graph OAuth integration
  • Event CRUD with availability checking via freebusy API
  • AI-optimized time slot suggestions for auto-scheduling
  • Public demo booking with embeddable scheduling links
Axum 0.7 Google Calendar MS Graph
:3009

v100-integration

External integrations hub. CRM connections, webhook management, SMS, and event streaming.

  • CRM connections with bi-directional sync
  • Webhook management with retry and dead-letter queue
  • Event streaming: poll, ack, subscribe, and history
  • SMS via Twilio for meeting reminders and notifications
Axum 0.7 PostgreSQL Twilio Redis

Content Distribution

Multi-platform publishing with OAuth and scheduled content delivery

:3025

v100-social-publisher

Multi-platform publishing service. OAuth-connected to all major social and video platforms.

  • OAuth for YouTube, TikTok, Instagram, LinkedIn, Twitter/X
  • Social account management with token refresh
  • Publish job tracking with retry logic and error recovery
  • Platform-specific upload APIs with format adaptation
Axum 0.7 PostgreSQL OAuth 2.0
background

v100-publishing-worker

Scheduled content worker. Automatically publishes content at optimal times per user preferences.

  • Polls scheduled_content table for due items on a cron loop
  • Auto-schedules based on timezone, spacing, and max posts/day
  • Processes publish pipeline via social-publisher service
Tokio PostgreSQL Cron
19
Rust Services
200+
API Endpoints
0
Node.js Services
ARM64
Graviton Native

Why Rust for Every Service?

Not a partial rewrite. Not "Rust for the hot path." The entire platform — from the API gateway to the cron-based publishing worker — is Rust.

Zero-Cost Abstractions

No garbage collector pauses. No JIT warmup. No V8 event loop bottlenecks. Every service starts in milliseconds and uses constant memory under load. The gateway processes requests in under 10 microseconds.

Memory Safety Without GC

Rust's ownership model eliminates use-after-free, buffer overflows, and data races at compile time. For a video platform handling user media, this isn't academic — it's the difference between a CVE and a safe deployment.

ARM64 Graviton Native

Every binary compiles to native ARM64 machine code for AWS Graviton. No emulation layers, no compatibility shims. We run on the most cost-efficient compute in AWS at full native speed.

See It Running

Try the live demo. Read the API docs. Or just start building — the free tier includes 1,000 credits.