GenCost — Generative Content Economics Engine

Meter, route, and optimize gen-AI API costs in real-time.

Built for the AI Tinkerers x ElevenLabs Generative Media Hackathon — NYC, April 11, 2026.

Track 3: Economics & Attribution | Bonus: Best Cost Reducer

---

What It Does

GenCost is infrastructure that sits between your application and gen-AI APIs. Every API call — voice generation, LLM completion, image synthesis — flows through a metering proxy that logs cost, latency, and lineage in real-time.

Not a content tool. A content economics engine.

Feature	What It Does
Metering Proxy	Intercepts all gen-AI API calls, calculates real cost per call
Smart Router	Routes requests to cheapest model meeting quality threshold
Content Pipeline	End-to-end generation: LLM → TTS → SFX → Image → Video, all metered
A/B Compare	Runs the same prompt at Draft / Standard / Premium tiers side-by-side
Attribution Tracker	Cost fingerprint per asset, lineage chains, remix tracking
Content Provenance	Voice-clone licenses, consent labels, asset provenance records
Waste Detection	Flags generated-but-unused assets, quantifies wasted spend
Teleport	Google Street View → fal.ai video + ElevenLabs audio guide (fully metered)
Terminal Chat	Siri-style in-app console that drives the router/pipeline from natural language
Dashboard	Real-time dark-themed UI with cost tickers, charts, phone-app panels, auth

Architecture

┌──────────────────────────────────────────────────────┐
│  localhost:3000  ← Dashboard (Next.js 15 + Supabase) │
│       ↑ WebSocket (real-time cost events)            │
│  localhost:8000  ← FastAPI Backend                   │
│       │                                              │
│       ├── Auth (Supabase JWT verification)           │
│       ├── Metering Proxy (intercepts all API calls)  │
│       ├── Smart Router (cost-optimized model picks)  │
│       ├── Content Pipeline (LLM → TTS → SFX → img)   │
│       ├── Teleport Engine (street view → video+VO)   │
│       ├── Attribution Engine (lineage + waste)       │
│       ├── Provenance + Voice Licensing               │
│       │                                              │
│       ├──→ api.elevenlabs.io   (TTS, SFX, Music)    │
│       ├──→ openrouter.ai       (LLM routing)        │
│       ├──→ api.deepseek.com    (DeepSeek direct)    │
│       ├──→ api.x.ai            (Grok direct)        │
│       ├──→ fal.ai              (image + video gen)  │
│       └──→ maps.googleapis.com (Street View frames) │
│                                                      │
│  SQLite ← all cost data     WebSocket ← live events │
└──────────────────────────────────────────────────────┘

API Keys Required

Provider	Get Key At	Cost	What It Powers
ElevenLabs	elevenlabs.io (free Creator tier at hackathon)	Free	TTS, SFX, Music, Teleport audio guide
OpenRouter	openrouter.ai	~$5 credits	LLM routing (DeepSeek, Gemini, Claude, Llama)
DeepSeek	platform.deepseek.com	~$2 credits	Direct cheap LLM
Grok/x.ai	console.x.ai	~$2 credits	Grok 3 Mini
fal.ai	fal.ai	~$5 credits	FLUX image + video generation
Google Maps	console.cloud.google.com (Street View Static API)	Free tier	Teleport street-view frames
Supabase	supabase.com	Free tier	Auth + per-user settings / API keys

Total demo budget: ~$15-20 for real API calls.

Quick Start

1. Clone and configure

git clone https://github.com/ron2k1/gencost.git
cd gencost
cp .env.example .env
# Fill in your API keys in .env (at minimum: ElevenLabs, OpenRouter, fal.ai, Supabase)

Windows one-click setup (recommended)

After cloning and filling in .env, run:

install-shortcut.bat

This creates a backend\.venv, installs backend + frontend dependencies, writes dashboard\.env.local, and drops a GenCost shortcut on your Desktop. Double-clicking it runs start-gencost.bat, which launches both the FastAPI backend (port 8000) and the Next.js dashboard (port 3000) in their own visible windows, waits for /api/health, and opens your browser. The scripts derive the repo location from their own path, so they work no matter where you cloned the repo.

On first launch, start-gencost.bat will also auto-heal a missing backend\.venv or dashboard\node_modules, so teammates who skip install-shortcut.bat still get a working app.

If you prefer the manual path (macOS/Linux, or just want more control), follow steps 2–4 below.

2. Backend

cd backend
pip install -r requirements.txt
python seed_data.py          # Pre-seed 55 pipelines for demo
uvicorn app.main:app --port 8000 --reload

3. Dashboard

cd dashboard
cp .env.example .env.local   # Fill in Supabase + Google Maps keys
npm install
npm run dev                  # Opens localhost:3000

4. Demo

1. Open localhost:3000 — see pre-seeded spend data
2. Click "Generate" with a prompt — watch real API costs tick up live
3. Adjust quality tier — see router switch to cheaper/premium models
4. Check waste detection — see undistributed asset costs

API Endpoints

Dashboard & metering

Method	Endpoint	Description
GET	/api/stats	Aggregated dashboard statistics
GET	/api/calls	Recent API call logs with cost data
GET	/api/calls/{pipeline_id}	Calls scoped to a single pipeline
GET	/api/health	Health check + API key status
WS	/ws	Real-time cost events via WebSocket

Pipelines & assets

Method	Endpoint	Description
POST	/api/pipeline	Trigger content generation (auth required)
GET	/api/pipelines	List pipeline runs with total costs
GET	/api/pipelines/{id}	Full pipeline detail (calls + assets)
GET	/api/assets/{id}/file	Serve a generated asset file
POST	/api/compare	Run the same prompt at draft / standard / premium in parallel (auth)

Attribution, waste & provenance

Method	Endpoint	Description
GET	/api/attribution/{pipeline_id}	Cost attribution breakdown
GET	/api/lineage/{asset_id}	Parent/child asset lineage tree
GET	/api/waste	Waste report (generated but unused)
POST	/api/distribute/{asset_id}	Mark an asset as distributed (clears from waste)
GET	/api/provenance/{asset_id}	Consent / provenance label for an asset
GET	/api/licenses	List licensed voice clones
GET	/api/licenses/{voice_id}	Single voice-clone license detail

Router, models & budget

Method	Endpoint	Description
GET	/api/router/table	Full routing table with model pricing
GET	/api/router/decide	Explain which model the router would pick
GET	/api/models	All known provider models with prices
GET	/api/budget	Current budget + quality tier
POST	/api/budget	Update budget controls (auth)

User settings (Supabase-backed)

Method	Endpoint	Description
GET	/api/settings/keys	List user's stored API keys (redacted)
POST	/api/settings/keys	Add / update a user's provider key
DELETE	/api/settings/keys/{provider}	Remove a user's provider key
POST	/api/generate-icon	Generate an icon via fal.ai

Terminal & Teleport

Method	Endpoint	Description
POST	/api/terminal/chat	Siri-style natural-language console (auth)
POST	/api/teleport	Street-view → video + audio guide teleport run (auth)
GET	/api/teleport/check	Street-view availability for a location
GET	/api/teleport/streetview-url	Signed Street View Static image URL

Tech Stack

Layer	Technology
Backend	Python 3.11+, FastAPI, httpx, SQLite, Pydantic, PyJWT, MoviePy
Frontend	Next.js 15, React 19, Tailwind v4, Framer Motion, Recharts, Lottie
Auth	Supabase (SSR cookies + JWT verification in FastAPI)
APIs	ElevenLabs, OpenRouter, DeepSeek, Grok, fal.ai, Google Street View
Real-time	Native WebSocket (FastAPI)
Database	SQLite (zero-config, hackathon-speed)

Modules

Module	Files
Metering Proxy	backend/app/proxy/
Smart Router	backend/app/router/
Content Pipeline	backend/app/pipeline/
Attribution	backend/app/attribution/
Teleport (Street View → video + VO)	backend/app/teleport/
WebSocket hub	backend/app/ws/
Auth / Supabase JWT	backend/app/auth.py
Dashboard	dashboard/src/
Infra / Demo	scripts/, backend/seed_data.py, start-gencost.bat

Pricing Tables (Real, April 2026)

ElevenLabs

Model	Cost
Flash v2.5	$0.18/1K chars
Multilingual v2	$0.30/1K chars
SFX v2	$0.04/generation

LLMs (via OpenRouter)

Model	Input/1M	Output/1M
DeepSeek V3	$0.14	$0.28
Gemini 2.5 Flash	$0.15	$0.60
Grok 3 Mini	$0.30	$0.50
Claude Sonnet 4	$3.00	$15.00

Images (fal.ai)

Model	Cost
FLUX Schnell	$0.003/image
FLUX Dev	$0.025/image
FLUX Pro 1.1	$0.05/image

License

MIT