Truthound 3.1 Release Notes¶

Highlights¶

Truthound 3.1 adds a new additive AI review surface while keeping the core validation kernel and zero-config workflow intact.

Key changes:

truthound.ai is now the canonical optional AI namespace
truthound.has_ai_support() and truthound.get_ai_support_status() expose a root-level feature probe
suggest_suite(...) produces persisted, reviewable proposal artifacts
explain_run(...) produces canonical run analysis artifacts
approval, rejection, and apply now have typed lifecycle facades
live smoke runners exist for both proposal and run-analysis OpenAI paths
public docs now describe the dashboard through a boundary-level overview instead of mirroring a full dashboard manual

The stable AI additions are:

root probe: has_ai_support(), get_ai_support_status()
review APIs: suggest_suite(...), explain_run(...)
lifecycle APIs: approve_proposal(...), reject_proposal(...), apply_proposal(...)
review helpers: list_proposals(...), show_proposal(...), list_analyses(...), show_analysis(...)

These remain additive. Importing truthound without AI dependencies continues to work.

Truthound 3.1 also tightens the public documentation boundary:

orchestration remains fully documented as a public first-party layer
dashboard remains visible in the docs portal through a concise overview page
the public docs no longer depend on a mirrored dashboard manual
a new AI section explains the technical boundary, artifact schema, redaction policy, provider contract, and release gates

Install the additive AI surface only when you need it:

pip install truthound[ai]

For the rest of the core workflow, existing pip install truthound behavior is unchanged.