TrueVectors SuiteAgent Behavioral Reliability
TrueVectors Hub
The TrueVectors Hub keeps every verdict — proved, corroborated, or model judgment — in a durable system of record.
A system of record for agent behavior
Every audit the TrueVectors Auditor runs is persisted to the Hub — what the agent claimed, what was proved, what your team confirmed or dismissed. Not a throwaway check, but a behavioral audit trail that survives model upgrades, stays on your machine, and proves diligence when someone asks.
Queryable behavioral history
The Hub is a durable, queryable history of how your AI agents behave over time — runs, receipts, evidence tags, and the invariants that were checked. When a model is swapped or a team rotates, the audit trail stays.
Dashboard
The Hub dashboard is where audit runs live: an evidence-tier toggle on the run detail page, defaulted to Proved, with both counts shown. Review receipts, export PDFs, and trace every verdict back to what was checked.
See the Hub at work
The Hub is the durable record across every repo you audit — a fleet view, per-repo run history, and the persisted proof behind each verdict. Raw counts only, never a score, and your artifacts never leave your machine.
A fleet coloured by its latest honest verdict
Every audited repo, grouped broken / unverified / clean with its violation count and how recently it was checked. Descriptive counts only — no composite "reliability score" — so the dashboard reflects what was proved, not a number someone could game.
Every run, kept with its evidence tier
Open a repo to its run history and a plain summary — how many rules were checked, how many violations, and whether any AI was used. Findings are bucketed by evidence tier, and each outcome verdict is marked proved or violated against the real artifact.
The proof behind each verdict, persisted
Expand any verdict to the evidence that was recorded — the invariant id, its kind, the status, and the offending row count. Not a throwaway pass/fail, but a durable, inspectable trail that survives model upgrades and team changes.
Proved tier first, in the record
The Hub surfaces the same proved tier discipline as the Auditor. On the run detail page, proved-tier findings lead; unproved llm findings are never hidden — they're one toggle away, and their count is always visible. What you see in the dashboard is what was persisted: no retroactive rewriting, no confidence theater.
Local-first, private by default
For authorized evaluations, the Hub records metrics-only receipts from the Auditor running in your CI/VPC. TrueVectors uses zero product telemetry: no usage ping, analytics, or phone-home. Audit artifacts, file paths, and findings stay in your environment except for explicitly enabled LLM calls.
Build your audit trail
Public distribution is closed. Request an authorized CI-first evaluation to review metrics-only verdict records in the Hub.
Every verdict stamped with where it came from. Trust through proof.