Styxis
Company
Truthound DepotTruthound DepotTRAPOT Protocol Gateway
Knowledge
Styxis Labs

Truthound Depot

Sniff before you merge. — Evidence Gate for Dataset Changes

META

Metadata-only evidence

Keep raw rows, documents, and credentials in your own storage. Truthound Depot records what changed, why it changed, and which gate approved it.

AI

AI-assisted review

AI summarizes risk and prepares reviewer checklists for each pull request, approval, release, and rollback without taking over the final decision.

5

Quality gates

Upload, Branch, Pull Request, Release, and Rollback gates check schema, profile, references, and evidence before a change moves forward.

∞
Branches & snapshots
5
Quality gates
1-click
Release and rollback
Truthound Depot logo
01
Depot
Open workspace
02
Branch
Isolate change
03
Compare
Review diff
04
Pull Request
Request approval
05
Quality Gate
Record evidence
06
Release / Rollback
Release or recover

Data change evidence workflow

Every dataset change leaves a clear trail from Depot branch to approval, release, and rollback.

01

Depot

Create a home for each dataset change stream without moving raw data out of your existing storage.

02

Branch

Keep customer updates, test data, RAG knowledge, and eval changes isolated until review is complete.

03

Compare

Review schema, profile, reference, and fingerprint diffs so the change is visible before merge.

04

Pull Request

Ask for review with AI summaries, reviewer notes, and gate results in one place.

05

Quality Gate

Block risky changes before approval, release, or rollback when evidence does not meet the rules.

06

Release & Rollback

Approve a verified release, then return to a known good baseline when a dataset change breaks behavior.

Capabilities

Review dataset changes before they reach users 

Review AI/RAG/eval dataset changes before release. Raw data stays in your storage; Truthound Depot records the evidence trail. 

01

Depot

Create a workspace for each AI product, customer project, RAG dataset, or eval dataset.

02

Branch

Isolate dataset changes before they touch the release path.

03

Pull Request

Review summaries, diffs, gate results, and approval notes before merge.

04

Quality Gate

Check schema, profile, references, fingerprints, and policy evidence at each key step.

05

Evidence

Record what changed, who approved it, which gate ran, and why the change was allowed.

06

Approval

Keep human approval visible next to AI assistance and rule-based gate results.

07

Release

Promote a verified dataset baseline for customer delivery, operations, or AI production.

08

Rollback

Return to a known-good release when a quiet data change breaks behavior.

Truthound Depot package image
Enterprise Console

Review changes without moving raw data 

Truthound Depot sits above your files, tables, and object storage to record metadata, evidence, approvals, releases, and rollback points. 

Depot Workflow

  • Depot home
  • Dataset reference registry
  • Branch / compare surface
  • Pull request queue
  • Release history

Validation-native

  • Upload Gate
  • Branch Gate
  • Pull Request Gate
  • Release Gate
  • Rollback Gate

Product line

  • Truthound data change verification line
  • Depot public alpha SaaS surface
  • Core validation layer
  • Orchestration execution layer
  • Operations pipeline connection
∞
Depot workspaces
∞
Dataset branches
100%
Diff visibility
5
Quality gates
∞
Release tags
1-click
Rollback recovery
Open Source Foundation

Truthound Core and Orchestration

Truthound Core and Truthound Orchestration are available under Apache 2.0.

truthound-coretruthound-orchestration

Evaluating Truthound Depot?

Let’s discuss how to design verifiable workflows for AI-driven dataset changes.

Email us