Architecture & Cost Model

One API call.
Unlimited generation.
99% less cost.

SDG AI Studio uses Claude exactly once — to define your domain schema. Every subsequent generation, variation, and CSV output runs entirely on your local machine. No GPU. No API. No per-row cost.

1
LLM API call
per domain
100+
CSV variations
generated locally
0
GPU required
locally or remotely
99%
reduction in
API spend
System Flow

Every interaction, mapped

From your first prompt to 100 automated CSV variations — here's exactly what happens at each step, and where the cost savings come from.

LLM API call
Local — zero API cost
sdg-studio / system-sequence · v6
8 steps · 5 participants
API LOCAL Dashboard Client SDG Client app Local repo Client-side SDG Server Local server LLM Server-side 1 2 3 4 5 6 7 8 Send prompt Return response Generate domain schema Generate SPF file Send SPF file Generate output CSV Send CSV file Request 100 variations loop [×100] Auto-generate edge case SPF Auto-submit SPF variation Generate CSV variation Deliver CSV variation
STEP 01
Send prompt
LLM API
STEP 02
Return response
LLM API
STEP 03
Generate domain schema
Local
STEP 04
Generate SPF file
Local
STEP 05
Send SPF to server
Local
STEP 06
Generate output CSV
Local
STEP 07
Deliver CSV to dashboard
Local
STEP 08
100× auto batch loop
Local
Economics

Why 99% of API spend disappears

Traditional synthetic data tools hit the LLM for every row generated. SDG AI Studio uses the API once — to think — then executes everything locally.

Traditional approach
Every generation → API
N records = N API tokens
Request schema for domain GPU / API
Generate 10,000 rows GPU / API
Generate variation #1 of 100 GPU / API
Generate variation #2 of 100 GPU / API
Generate variation #3–100 GPU / API
Total API calls 102+
SDG AI Studio
Schema once → generate forever
N records = 1 API call
Request schema for domain API — once only
Generate 10,000 rows Local — free
Generate variation #1 of 100 Local — free
Generate variation #2 of 100 Local — free
Generate variations #3–100 Local — free
Total API calls 1
0
API calls with a traditional approach for 100 variations
99%
reduction in API spend — achieved by moving generation local
0
API call needed with SDG AI Studio, regardless of output volume
Local-First Architecture

No GPU. No cloud. No marginal cost.

Once Claude has defined your schema, the entire generation pipeline runs on your hardware — no inference calls, no per-token billing, no network dependency.

⚡
No GPU Required
Schema-driven generation is pure data transformation — filling field templates with distribution-sampled values. A CPU handles millions of rows in seconds with no accelerator hardware.
0 GPUrequired locally or via API
🔒
Data Never Leaves
After the initial schema call, all generation happens inside your environment. Your domain logic, field distributions, and output data stay entirely on-premises.
100%local execution after schema setup
∞
Unlimited Regeneration
Once the SPF schema file exists, you can regenerate, vary, and scale indefinitely. Run 100 variations or 10 million rows — the API cost stays at exactly 1 call.
∞regenerations at zero marginal cost
🔄
Schema Reuse
SPF files are plain, portable schema definitions. Store them in your local repo, version them in git, share them across teams — no API dependency to reuse them.
1×schema definition, reused forever
📊
Predictable Cost
Your bill is always one prompt-sized API call per domain. Whether you generate 100 rows or 10 million, whether you iterate once or 100 times — the cost is identical.
Flatcost regardless of output volume
🚀
Near-Instant Generation
Without round-trip API latency on every row, local generation runs at memory speed. 10,000 rows in under a second. 100 CSV variations in a single automated batch.
<1sfor 10,000 rows, no API wait

Ready to cut your
data generation costs?

One prompt. One schema. Unlimited local generation — with no GPU, no cloud dependency, and no per-row API cost.

Buy Now → ← Back to home