AI assistant (chart, scripting, backtest)

An AI assistant in trading software is a language-model panel that reads the application's current state, such as the mounted chart, a script or a backtest report, and answers questions or proposes typed actions about it. It works from the user's own model key and explains what is on screen; it does not predict prices or send orders.

Senzoukria · Glossary · Updated September 2026

What it reads and what it returns

The assistant is given context assembled by the application: the instrument and session, the visible bars and their computed values, the current settings, or the source of a script and its last error. It returns text, and in some screens a list of proposed changes. Its usefulness depends on the quality of that context, which is why the desktop shows it: the live context can be opened and read before a question is asked.

  • Chart: explains the flow, delta, heatmap and gamma state it was given; proposes settings as typed actions.
  • Scripting: writes or edits an indicator or strategy, runs it, and reads errors back.
  • Backtest and report: reads the statistics produced by a test and comments on them.
  • General agent: reads GEX, Option Flow, Account and News state.

Boundaries

  • It does not arm the autopilot and cannot send an order; the general agent reads context only.
  • It states observations from the fields it received; a hidden intent it describes is a hypothesis, not a measurement.
  • The conversation, and any attached screenshot, goes to the model provider and is billed to the user's key.
  • A script it produces is a draft to test, not evidence that the rule works.

In Senzoukria

The main menu has an Assistant entry. Its empty state reads: I read your live tape, flow, delta, heatmap, gamma; I don't predict and I don't give financial advice. It changes settings through typed actions that the user confirms with a click. The Scripting screen has its own Assistant panel with Insert into editor, Attach a screenshot and the running and reading states shown while it works. The Replay screen has a Backtest assistant that reads the backtest statistics, and the report screen has an Assistant as well.

The Assistant engine picker in Settings offers Claude with the user's own key, Groq with a key from its console, or a local Ollama model. Keys are stored in the operating-system keychain and never sent to Senzoukria's servers; the help text says the key powers the assistant on every screen: chart, scripting, backtest, report.

Common mistakes

  • Asking for a prediction and treating a fluent answer as one.
  • Accepting a proposed configuration without reading the typed actions first.
  • Attaching a screenshot that contains account details to a conversation billed to an external provider.
  • Confusing the general agent with the strategy autopilot; they are different components with different permissions.

This page in other languages

Frequently asked questions

Which model does the assistant use?
The engine is chosen in Settings: Claude with the user's own Anthropic key, Groq with a key from its console, or a local Ollama model pulled on the machine. The key is stored in the operating-system keychain and used from that machine only. There is no shared Senzoukria key and no proxy through its servers.
Can the assistant place trades or arm the autopilot?
No. The general agent reads application context and answers; it cannot arm the strategy autopilot, which requires explicit arming, account permissions and configured limits. Chart settings proposed by the assistant are typed actions that take effect only after the user confirms them with a click.
How is the assistant used in a backtest?
On the Replay screen, the Backtest assistant reads the statistics produced by a completed test and comments on them. The strategy-testing guide suggests giving it the observation and its limits, asking it to separate evidence from explanation, and checking any proposed rule against the data before running a test.

Keep reading

Strategy researchTest a trading strategy, from observation to challenge rulesDefine an observation, question it with the copilot, write a rule and explore costs, drawdown and fictional challenge limits in an interactive example.Quant · Start hereQuant trading for beginners: from idea to backtestDefine a testable rule, choose suitable data and evaluate later periods after costs.Order flow · HubOrder flow trading: the complete guideThe hub. What order flow is, what it shows — aggression, delta, imbalance, absorption, liquidity — and the tools that read it, each linked to a full guide.