Skip to content
AQREVI

Version_1.0_·_effective_29_September_2026

AI in AQREVI

The core rule: the engine computes, the model explains. The numbers in the journal, reports, tiers, feed and copier are computed by code; a model only retells them and can be wrong in its words, but not in the numbers themselves.

$ /legal/ai

AI in AQREVI

1. What goes to a model#

When you ask the coach or an agent, or run an analysis, the model receives your question and the context the engine assembled: statistics, trade excerpts, dossier rules, a note's text — exactly what the answer needs.

Version 1.0 · effective 29 September 2026

When you ask the coach or an agent, or run an analysis, the model receives your question and the context the engine assembled: statistics, trade excerpts, dossier rules, a note's text — exactly what the answer needs. The precise content of every request is visible in the Trajectory / Trace tab.

2. Who answers#

Without your own key, the platform model answers through OpenRouter; we ask it not to collect or retain request data (data_collection: deny).

Without your own key, the platform model answers through OpenRouter; we ask it not to collect or retain request data (data_collection: deny). With your own key (OpenAI, Anthropic, Google, xAI, DeepSeek, Moonshot, Perplexity and others) the request goes to that provider under your own agreement with them; some providers are outside the EU and the US — the choice is yours. A local model (Ollama) in the local version runs on your computer and sends requests nowhere.

3. What we do not do#

We do not train models on your data and do not sell it.

We do not train models on your data and do not sell it. A model makes no decisions with legal or financial effect: it does not open trades, does not change rules or the dossier without your approval, does not manage money. Agents are off by default, and anything irreversible needs an explicit permission.

4. A trace of every call#

Every model call leaves a trace in your database: what was sent, who answered, how many tokens and what it cost.

Every model call leaves a trace in your database: what was sent, who answered, how many tokens and what it cost. Traces can be switched off in settings — then no new ones are written; what was recorded is deleted only by you.

5. Model errors#

Models make mistakes, including in a confident tone.

Models make mistakes, including in a confident tone. A model's answer is neither advice nor fact; check it against the engine numbers it cites. Numbers that are not in the computed context are flagged as unverified.

AQREVI. A web service, your journal in its own database.

AQREVI — AI in AQREVI