Data never leaves your network
Row datasets load from your org profile on demand. No vendor cloud inference.
Vera Chat runs locally — including under file:// for air-gapped deployments.
Trust
Built for the entire C-suite
Vera shifts AI from an unpredictable experiment into an industrialized, auditable asset — optimized for industries where data sovereignty is non-negotiable.
Deterministic by design
Same prompt + same skill version → same plan. Pin Vera.config.now for reproducible
regression tests. No stochastic surprises in production workflows.
Inspectable plans
Every query produces a structured JSON plan — not opaque SQL strings invented at runtime. Review filters, scopes, and aggregates before they execute.
Zero marginal inference cost
After distillation, daily operations are local computation. Scale usage without scaling GPU spend — a pragmatic blueprint enterprise finance teams can approve.
Org isolation
Copy orgs/demo/ to your own org folder. Map physical columns in one profile file.
Framework updates never overwrite your production mappings.
Honest disclaimers
Vera answers from loaded data only. Modeled estimates (e.g. profitability) are labeled as such. The engine does not hallucinate — it either resolves or asks for clarification.
Built-in guardrails
A control framework for Vera-local — plain rules for leadership, concrete expectations for engineering.
- Runs entirely local — Core behavior works without a vendor cloud or backend; the app shell can run from
file://for air-gapped use. - Predictable every time — Same prompt and skill version produce the same plan. No LLM on the production path; clocks are pinned for audit and regression tests.
- Data loads on demand — Large row files are fetched when a query needs them, not at page load — keeping memory use bounded.
- Inspectable plans — Vera outputs governed query plans you can review, not SQL strings invented at runtime.
- Vocabulary in skills — People describe the same intent in countless ways; hardcoding phrases in core code is brittle. Business wording belongs in skills and org profiles instead — where updates stay traceable and auditable.