Mock Data

Schema in. Sealed dataset out.

Describe a dataset in one English sentence, get back a fully-sealed synthetic CSV in under a minute — with the contract, the seed, the BLAKE3 chain, and the same offline verifier as every other engine on the platform. Same prompt, same seed: byte-for-byte equal across clusters, regions, and years.

Latency
~38 s
Industries
47
Determinism
byte-equal
Bundle
sealed
mock / saas-customers · 200 rows · seed 42rapid-rrf
prompt

“200 SaaS customer accounts with MRR 50–5000, plan Starter/Growth/Enterprise, region AMER/EMEA/APAC, signup date in the last 18 months.”

contract K · sealed
{
  "rows":   200,
  "seed":   42,
  "fields": ["id","name","plan","mrr","region","signup"],
  "ranges": { "mrr":[50,5000] },
  "enums":  { "plan":["S","G","E"], "region":["AMER","EMEA","APAC"] }
}
rows.preview · first 6 of 200byte-equal on rerun
idnameplanmrrregionsignup
cust_a17Aurora KadeGrowth$1284.50EMEA2025-08-12
cust_b22Marcus YanEnterprise$4192.00AMER2024-11-30
cust_c08Inès BouchardStarter$79.00EMEA2026-02-04
cust_d31Hiroshi TanakaGrowth$962.10APAC2025-05-19
cust_e44Priya RamanathanEnterprise$3580.75APAC2025-01-22
cust_f17Liam O'ConnorGrowth$1107.40AMER2025-09-03
latency
38 s · 200 rows
contract sha
9c10ab…
blake3 root
a4f2…d801

Six promises Mock Data keeps every time.

These aren’t configuration toggles — they are how the engine was built. Switch none of them off, configure none of them on. The engine simply behaves this way for every prompt, every tenant, every run.

Promise · 01

Sub-minute, every time

Our rapid fabricator is built for latency, not training. Most prompts complete in 20–60 s for up to 200 k rows — including evidence sealing — so it sits comfortably inside an interactive UX or a CI pipeline.

Promise · 02

Deterministic by construction

Same prompt + same seed + same engine version = byte-equal output. We test this in CI: the regression suite re-runs 47 reference contracts and checks every cell against a stored hash. If a single byte drifts, the build fails.

Promise · 03

Any industry, on demand

Mock is LLM-driven schema fabrication, not a fixed template list. Describe a banking dataset, an EHR cohort, a smart-meter feed, a retail basket, a manufacturing line — the engine drafts a domain-aware the sealed contract with sensible defaults, and you can override any of them per prompt.

Promise · 04

Same evidence chain as the rest

Mock outputs join the same BLAKE3-chained ledger as Synthesize, Virtual SCADA, and ICS Security. The bundle that ships with a Mock dataset is structurally identical — your downstream tooling treats them the same way.

Promise · 05

Two surfaces, same engine

Use the slt CLI for shell pipelines and one-off ops, or call client.mock.create() from the typed Python SDK. The contract goes through the same compiler; the audit log records who called it and from where.

Promise · 06

Quality fail-closed

Even on the fast path we check column-level distribution sanity, range bounds, enum compliance, and nullability against the contract before the bundle is sealed. A regression aborts; you never receive a quietly-broken dataset.

How it works

From sentence to sealed bundle in six stages.

Each stage writes a typed artefact into the chain. Even the prompt itself joins the audit trail — so an auditor can trace any byte in the dataset back to the operator who asked for it.

CLI · sample
slt mock create \
  --prompt "200 SaaS customers, MRR 50-5000, plan
            Starter/Growth/Enterprise, region AMER/EMEA/APAC,
            signup date in last 18 months" \
  --rows 200 --seed 42 \
  --evidence ./customer-bundle.tar.zst

# in 38 s →
✓ contract K sealed   sha 9c10ab…
✓ 200 rows generated  byte-equal across cluster, region, year
✓ blake3 chain root   a4f2…d801
✓ bundle on disk      4.7 MB
Stage 01

Describe what you want

One natural-language sentence, or an inline JSON schema, or a YAML pulled from a referenced industry template — pick the surface that fits your tooling. The compiler accepts all three and normalises to the sealed contract.

Stage 02

the sealed contract seals before any row is born

Schema, ranges, enums, constraints, seed, engine version and operator intent are sealed into a JSON artefact. The contract hash is the unique identity of every dataset that comes from this run.

Stage 03

Generation

The fabricator pulls from the contract: sample rules, range/enum constraints, correlation rules where defined. No external model call by default; fully air-gappable.

Stage 04

Quality gates check

Distribution sanity, range bounds, enum compliance, nullability and constraint satisfaction are checked. Fail-closed — a mismatch aborts before sealing.

Stage 05

BLAKE3-chain + sign

Per-step IO hashed and chained; the final bundle is signed and ready to verify offline with the open-source evidence verifier CLI.

Stage 06

Bundle delivered

.tar.zst lands on disk, in S3, in your tenant artefact prefix, or as a stream — your choice. Re-running the contract anywhere yields the same hash.

Why it’s not just another faker.js with a UI.

Most “mock data” tools are shell scripts behind a SaaS skin. our rapid fabricator treats every output as evidence — sealed, hashed, replayable, and auditable on the same chain as the rest of the platform.

CapabilityRadMah AI MockTypical mock-data tool
Latency for 200 k rows20 – 60 s incl. evidenceMinutes to hours
Same prompt → same outputByte-equal, contract-pinnedRun-to-run drift
Industries47, all overridableHand-built per project
Evidence bundle per runBLAKE3 chain by defaultOptional, paid add-on
Air-gapped operationNo external model callLLM round-trip required
Quality fail-closedDistribution / range / enum gatesBest-effort, silent drift
Same chain as Synthesize / SCADA / ICSYes — one ledgerSeparate tools, separate audit trails

Where Mock Data earns its keep.

Four work patterns the team uses every week. None of them require touching production data; all of them produce a bundle you can hand to anyone in the business.

Sandbox / demo data

Stand up a realistic-looking demo environment for sales, customer success, or training without touching production. Re-seed in 30 seconds for a fresh run.

Load & soak testing

Generate millions of rows that pass your validation but exercise edge ranges your real data doesn't carry. Seed-pinned so test failures are reproducible.

Test fixtures in CI

Drop slt mock into your CI to generate fixture data per branch. The contract is checked into git; the bundle is not — but it rebuilds byte-equal on demand.

Rapid prototyping

Prototype an analytics dashboard, an ML model, an API contract — all against synthetic data shaped like the real thing, before legal even joins the call.

Posture you don’t configure.

All four guarantees apply by default. None of them are paid add-ons.

Tenant-isolated

Per-tenant Fernet at rest, per-tenant artefact prefix, ORM-level row filtering — same as the rest of the platform.

Air-gappable

No external LLM call by default; runs entirely inside your VPC or air-gapped enclave.

Tamper-evident

Offline verifier flags any in-place mutation of the bundle, regardless of how it travelled.

Audit-ready

Caller, scope, contract hash and bundle root recorded in the audit log on every call.

Bring a sentence. We’ll ship a sealed bundle.

30-minute working session: tell us the dataset shape you need and one downstream task you want to drive with it. We’ll cut the contract, generate the bundle, and walk you through verifying it offline.

evidence pipeline · job_9b3df1
live
K
Sealed contract
E
Engine run
Q
Quality gates
B
BLAKE3 chain
S
Sealed bundle
rows generated
10 000
qa score
0.957
chain root
a4f2…d801