Infrastructure

Compute Infrastructure

The computational backbone powering 10,000 Faculty agents, real-time dialogue, RAG pipelines, voice synthesis, and the entire Inquiry Institute platform.

10,000

Faculty Agents

30+

Edge Functions

300+

TTS Voices

Compute Categories

Estimated Monthly Compute Costs

$4,500–$16,150

Varies based on usage, Faculty activity, and course enrollment

🧠

LLM Inference

~$2,000–8,000/mo

Large language model inference for Faculty conversations, content generation, and interactive dialogue with 10,000 AI personas.

OpenRouter Gateway

OpenRouter

Multi-model routing to Claude, GPT-4, Llama, Mixtral, and other providers

•Dynamic model selection per Faculty
•Automatic fallback on rate limits
•Usage tracking per Faculty agent
•Cost optimization via model routing

Context Window Management

Managing long conversations with historical Faculty personas

•128K+ token contexts for deep dialogue
•Sliding window for extended sessions
•RAG augmentation for corpus access

📚

RAG & Embeddings

~$500–1,500/mo

Retrieval-Augmented Generation pipelines that give Faculty agents access to their historical corpora and institutional knowledge.

OpenAI Embeddings

OpenAI

Text embeddings for semantic search across Faculty corpora

•text-embedding-3-small for efficiency
•1536-dimensional vectors
•Batch processing for corpus ingestion

pgvector (Supabase)

Supabase

Vector similarity search integrated with PostgreSQL

•HNSW indexing for fast retrieval
•Hybrid search (vector + full-text)
•Per-Faculty corpus isolation

LangChain Pipelines

Self-hosted

RAG orchestration and chain management

•Document loaders for PDFs, ePub, HTML
•Chunking strategies per corpus type
•Citation tracking and source attribution

🎙️

Audio & Speech

~$1,000–3,000/mo

Text-to-speech and speech-to-text for voice interactions with Faculty agents, including the Speaking Busts feature.

Microsoft Edge TTS

Microsoft

300+ neural voices for Faculty speech synthesis

•Free tier with high quality
•40+ languages supported
•SSML for prosody control
•Real-time streaming via WebSocket

Free

Azure Speech Services

Microsoft Azure

Premium TTS with viseme support for Speaking Busts

•22 viseme IDs for lip-sync
•Neural voice cloning (future)
•Custom voice training
•Real-time synthesis <200ms

ElevenLabs

High-fidelity voice synthesis for premium Faculty voices

•Voice cloning from samples
•Emotional expression control
•Multi-language support

Whisper STT

OpenAI / Self-hosted

Speech-to-text for voice input to Faculty

•Multi-language transcription
•Real-time streaming mode
•Speaker diarization

🗿

Audio2Face Animation

~$200–500/mo

Real-time facial animation for Speaking Busts, driven by audio and viseme data.

Viseme Processing

Azure viseme events to mouth shape mapping

•22 Azure visemes → 9 mouth shapes
•60ms easing between frames
•Canvas2D compositing

Bust Animation Pipeline

Real-time 2.5D portrait animation

•Micro-motion (2px amplitude)
•Random blinks (3.5–7s intervals)
•Sprite-based mouth atlas
•WebGL voxel rendering

RunPod GPU (Future)

RunPod

Serverless GPU for advanced animation models

•LivePortrait model inference
•Real-time video generation
•A10G / A100 instances

🎨

Image & Video Generation

~$500–2,000/mo

AI image generation for Faculty portraits, article illustrations, and iNQ publications.

Replicate

Serverless AI model hosting for image generation

•Stable Diffusion XL
•FLUX models
•ControlNet for style transfer
•Pay-per-inference pricing

HeyGen Avatars

HeyGen

Video avatar generation for Faculty presentations

•Talking head videos
•Custom avatar creation
•Lip-sync from audio

ComfyUI Workflows

Self-hosted / RunPod

Custom image generation pipelines for iNQ

•William Blake style LoRA
•Mucha Art Nouveau LoRA
•Batch illustration generation

💾

Database & Storage

~$100–500/mo

PostgreSQL database, file storage, and real-time subscriptions for the platform.

Supabase Database

Supabase

PostgreSQL with extensions for the entire platform

•Row Level Security (RLS)
•pgvector for embeddings
•Real-time subscriptions
•8GB database (Pro plan)

Supabase Storage

Supabase

File storage for Faculty assets and user uploads

•Faculty portraits and avatars
•Audio files and transcripts
•PDF publications
•100GB included

AWS S3

AWS

Bulk storage for Faculty corpora and archives

•Historical text corpora
•Training data archives
•Backup storage

⚡

Edge Functions

~$50–200/mo

30+ serverless functions handling Faculty agents, webhooks, and real-time operations.

Supabase Edge Functions

Supabase

Deno-based serverless functions at the edge

•30+ deployed functions
•faculty, ask-faculty, inquire
•matrix-webhook, social-media-post
•issue-microcredential, llm-gateway

🌐

Application Hosting

~$100–300/mo

Web hosting, CDN, and container orchestration for the platform.

GitHub Pages

GitHub

Static hosting for the Next.js frontend

•Automatic deploys via GitHub Actions
•Global CDN
•Custom domain (inquiry.institute)

Free

Fly.io Containers

Fly.io

Container hosting for Matrix, WorkAdventure, and services

•Matrix Synapse homeserver
•Element Web client
•WorkAdventure campus
•Global edge deployment

Vercel (Backup)

Vercel

Alternative hosting for preview deployments

•Preview URLs for PRs
•Edge functions
•Analytics

💬

Communication Infrastructure

~$50–150/mo

Matrix chat server, authentication, and real-time communication.

Matrix Synapse

Self-hosted on Fly.io

Decentralized chat for Faculty Club and salons

•Federation-ready homeserver
•End-to-end encryption
•Bot integration for Faculty
•Persistent room history

Matrix & Supabase Auth

Fly.io / Supabase

Identity and chat

•Matrix Synapse (password auth)
•Supabase Auth for app and bots
•Faculty bots via Application Service

Infrastructure Architecture

┌─────────────────────────────────────────────────────────────────────────┐
│                           INQUIRY INSTITUTE                              │
│                        Compute Infrastructure                            │
├─────────────────────────────────────────────────────────────────────────┤
│                                                                          │
│  ┌──────────────┐    ┌──────────────┐    ┌──────────────┐              │
│  │   Frontend   │    │   Supabase   │    │   Fly.io     │              │
│  │  (GitHub     │◄──►│  (Database,  │◄──►│  (Matrix,    │              │
│  │   Pages)     │    │  Auth, Edge) │    │  WorkAdv.)   │              │
│  └──────────────┘    └──────────────┘    └──────────────┘              │
│         │                   │                   │                       │
│         ▼                   ▼                   ▼                       │
│  ┌──────────────────────────────────────────────────────┐              │
│  │                    Edge Functions                     │              │
│  │  faculty │ ask-faculty │ llm-gateway │ matrix-webhook │              │
│  └──────────────────────────────────────────────────────┘              │
│         │                   │                   │                       │
│         ▼                   ▼                   ▼                       │
│  ┌──────────────┐    ┌──────────────┐    ┌──────────────┐              │
│  │  OpenRouter  │    │   pgvector   │    │  Azure TTS   │              │
│  │  (LLM API)   │    │   (RAG)      │    │  (Speech)    │              │
│  └──────────────┘    └──────────────┘    └──────────────┘              │
│         │                   │                   │                       │
│         ▼                   ▼                   ▼                       │
│  ┌──────────────────────────────────────────────────────┐              │
│  │          10,000 Faculty Agents with RAG Corpora         │              │
│  └──────────────────────────────────────────────────────┘              │
│                                                                          │
└─────────────────────────────────────────────────────────────────────────┘

Monthly Cost Breakdown

Category	Primary Provider	Low Est.	High Est.
🧠LLM Inference	OpenRouter	$2,000	$8,000
📚RAG & Embeddings	OpenAI	$500	$1,500
🎙️Audio & Speech	Microsoft	$1,000	$3,000
🗿Audio2Face Animation	Various	$200	$500
🎨Image & Video Generation	Replicate	$500	$2,000
💾Database & Storage	Supabase	$100	$500
⚡Edge Functions	Supabase	$50	$200
🌐Application Hosting	GitHub	$100	$300
💬Communication Infrastructure	Self-hosted on Fly.io	$50	$150
Total Monthly Estimate		$4,500	$16,150

Estimates based on moderate usage. Actual costs vary with traffic and Faculty activity.

Scaling Considerations

Cost Optimizations

• Response caching for common queries
• Model routing (smaller models for simple tasks)
• Edge TTS (free tier) for most voices
• Batch processing for corpus updates
• Reserved capacity discounts

Scaling Triggers

• Course enrollment increases
• Simultaneous salon sessions
• Journal publication cycles
• Marketing campaigns
• Video generation requests

Support Our Infrastructure

Help fund the computational resources that power Faculty agents, enable real-time dialogue, and keep the Institute running.

Support the Institute View Full Tech Stack