Styxis
Company
TRAPOT Protocol GatewayTruthound DepotTruthound Depot
TruthoundCore technology lineTruthound OrchestrationExecution and integration layerGolem / AWEHeadless Agentic Workflow EngineMabinogionProtocol resilience engine
Knowledge

Truthound Depot

AI Training Data Quality & Version Governance — Sniff before you merge.

AVAILABLE NOW

Try the workflow online. Run real operations in your own environment.

Browse a sample Public Depot or open a free demo workspace with non-sensitive test data. Paid supply is scoped as a customer-hosted installation, not public checkout.

Open online demotruthound.com
DATA

Dataset validation and versioning

Inspect dataset versions for PII, validation rules, schema drift, anomalies, and review impact before teams approve the change.

IMAGE

AI training image workflows

Review image sets, annotations, quality signals, and generation workflow metadata without exposing private source images as product data.

ERD/DB

ERD design and DB implementation audit

Audit ERDs against standard words, terms, and domains, then verify whether the approved design was implemented in the physical database.

0
Source data copied by default
3
Validation domains
1
Reviewable Evidence trail
Truthound Depot logo
01
Depot
Open demo workspace
02
Data Version
Import or connect
03
Validation
Check quality
04
Review Request
Request review
05
Evidence
Keep the basis
06
Publish / Restore
Publish or recover

One governance flow for datasets, images, and ERD/DB evidence

Truthound Depot gives reviewers the facts they need without turning private data, credentials, or non-public images into exported product payloads.

01

Dataset validation

Version datasets, run PII, validation, schema, drift, and anomaly checks, and compare results before a change is approved.

02

AI image review

Inspect training image quality, annotation consistency, and generation workflow metadata while private image assets stay inside the agreed boundary.

03

Private Agent execution

Run validation close to the source with local credentials and read-only access, then return reviewable result contracts instead of raw private data.

04

ERD design audit

Check whether ERD design follows approved standard words, standard terms, domain definitions, naming rules, and relationship intent.

05

DB implementation audit

Verify that the actual database implements the approved ERD with matching physical names, logical names, tables, columns, keys, and structure.

06

Evidence · Report

Record summaries, metadata, issue context, reviewer decisions, and reports so repeat review and approval can happen from the same basis.

Truthound Depot

Validation first, Evidence that teams can review later 

Truthound Depot keeps source execution inside safe boundaries and manages the dataset, image, ERD, DB, review, and report records around the result. 

01

Dataset source

Connect files, databases, object storage, or bounded source scopes while preserving the storage boundary chosen for that data.

02

Versioning

Track dataset versions, compare changes, and preserve the exact context reviewers used for approval.

03

PII and validation

Run PII, schema, rule, drift, anomaly, and quality checks before data affects a model, report, or delivery.

04

AI image workflow

Review training images, annotations, generated-image workflow metadata, and quality signals without exposing private image collections.

05

Private Agent

Execute validation near the source with local credentials, TLS, read-only access, and result-only contracts by default.

06

ERD design audit

Check ERD names, standard words, standard terms, domains, relationships, and design intent before implementation.

07

DB implementation audit

Compare approved ERD intent with actual physical names, logical names, tables, columns, keys, and structures in the database.

08

Evidence

Preserve summaries, metadata, issues, decisions, and report artifacts as supporting proof for later review.

09

Review Request

Route the validation result to the right people with enough context for approval, rejection, or rework.

10

Report

Export readable reports that explain what was checked, what failed, what passed, and what basis reviewers used.

11

Publish

Promote only work that passed review, with a traceable link back to the validation result and decision.

12

Restore

Recover to a previously reviewed dataset state when a later change creates risk.

Truthound Depot package image
From data to decision

Connect only what validation needs. Keep the source boundary intact. 

Local files, databases, image assets, and ERD/DB metadata use different storage and execution boundaries, but reviewers still get one consistent decision record. 

Prepare validation

  • Dataset versions and source scopes
  • Image and annotation assets
  • Generation workflow metadata
  • ERD standard word and term references
  • DB schema access scope

Review results

  • PII and validation findings
  • Dataset version comparison
  • Image quality and annotation checks
  • ERD design audit
  • DB implementation diff

Decide and share

  • Reviewer approval records
  • Evidence and report artifacts
  • Metadata-only sharing
  • Publish reviewed versions
  • Restore verified versions
CSV
Local files
JSONL
Stream data
Parquet
Columnar data
DB
Direct connection
S3
Object storage
HTML
Validation report
Open Source Foundation

Truthound Core and Orchestration

Truthound Core and Truthound Orchestration are available under Apache 2.0.

truthound-coretruthound-orchestration

Review the evidence before data work moves forward.

Truthound Depot is available at truthound.com as an online demo. Customer-hosted installation is scoped through pilot and agreement.

Open online demooffice@styxis-io.com