Truthound 3.x¶

Truthound is a layered data quality system. The center of that system is Truthound Core: a validation-first, Polars-first kernel built around TruthoundContext, ValidationRunResult, deterministic auto-suites, and a planner/runtime execution boundary. Around that core sit two first-party layers: Truthound Orchestration for host-native execution in schedulers and workflow systems, an additive truthound.ai namespace for reviewable AI proposal and analysis workflows, and Truthound Dashboard for operating Truthound through an installation-managed control-plane UI.

This portal keeps the core and orchestration surfaces fully documented, adds the public truthound.ai contract, and keeps the dashboard surface visible at the boundary level without reproducing the full console implementation here.

Truthound By Layer¶

Layer	Owns	Start here when you need to...
Truthound Core	Validation kernel, zero-config workspace, result model, reporters, Data Docs, checkpoint runtime, profiling, benchmarked execution path	run your first validation, use the Python API or CLI, understand `ValidationRunResult`, or evaluate the core contract
Truthound AI	Optional review-layer APIs for proposal generation, run analysis, approval logs, and controlled apply flows	understand the AI boundary, compile prompt-driven suite proposals, or analyze run evidence without mutating core state by default
Truthound Orchestration	First-party execution integration layer for Airflow, Dagster, Prefect, dbt, Mage, and Kestra	run Truthound inside a scheduler, asset graph, flow system, or warehouse-native orchestration surface
Truthound Dashboard	Installation-managed operational console for sessions, RBAC, sources, artifacts, incidents, AI review, and observability	operate Truthound deployments, manage teams and ownership, or review proposals and run analysis through a web UI

Choose Your Entry Point¶

I want to...	Start here
Run my first validation with almost no setup	Core Getting Started
Learn the core workflow end to end	Core Tutorials
Use Truthound from scripts or services	Core Python API
Use Truthound from a terminal or CI job	Core CLI Reference
Learn the optional AI proposal and analysis contract	Truthound AI
Design scheduler-native execution	Truthound Orchestration
Operate a control-plane UI for Truthound	Truthound Dashboard
Understand how the layers fit together	Concepts & Architecture

Why The Core Comes First¶

Truthound Core is the most rigorously validated layer in the product line. It is where the primary runtime contracts, benchmark evidence, and release-grade behavior are fixed.

ValidationRunResult is the canonical runtime output
deterministic auto-suite selection replaces "run everything" defaults
planner/runtime boundaries keep execution exact-by-default and maintainable
TruthoundContext owns the zero-config .truthound/ workspace
benchmark claims are intentionally bounded to comparable core workloads

Verified Core Benchmark Snapshot¶

The latest fixed-runner benchmark verification shows:

Truthound Core finished ahead of Great Expectations on all eight comparable release-grade workloads
local speedups ranged from 1.51x to 11.70x
SQLite pushdown speedups ranged from 3.69x to 7.58x
local peak RSS stayed between 35.88% and 48.16% of Great Expectations
correctness parity was preserved across the full comparable workload set

Read the evidence in Latest Verified Benchmark Summary.

How This Portal Is Organized¶

Core¶

Use Core when you need the kernel itself:

Orchestration¶

Use Orchestration when Truthound should feel native inside Airflow, Dagster, Prefect, dbt, Mage, or Kestra:

AI¶

Use AI when you need a reviewable, artifact-driven AI layer on top of the core validation contract:

Dashboard¶

Use Dashboard when you need the operational control-plane:

Dashboard Overview
installation-managed operational console guidance
deployment-aligned review and approval surface
AI review, proposal, and run-analysis workflows built on truthound.ai