Preferences
For the complete documentation index, see llms.txt. Markdown versions of documentation pages are available by appending
.mdto page URLs; this page is available as Markdown.
AI Chart Analysis — Preferences
Core Persistent Preferences
The preference surface is deliberately small — only a handful of settings are exposed directly, so configuration never becomes a chore before you can start analyzing:
Default analysis detail level — Minimal, Standard, or Detailed
Maximum number of automatic drawings
Lines versus zones — whether structures render as single lines or shaded zones by default
Wick-based versus close-based levels — whether support, resistance, and range boundaries default to candle wicks or closing prices
Everything else is either inferred from how you use the feature, or set per-analysis with a temporary instruction.
Detail Levels
Minimal
Primary structure only
Three or four drawings at most
A one-line summary
Standard (recommended default)
Primary structure
Supporting context
Relevant measurements
The most important warning, if any
Detailed
Secondary structures
Additional indicators
Alternative interpretations
More measurements
An expanded explanation
You can drop back to a lighter view at any time:
This removes secondary information while keeping the primary analysis, and it's handled instantly — no round trip to the AI planner is needed (see Chart Behavior).
Inferred Preferences
Beyond the four core settings, Analysis Mode learns your stylistic preferences from your corrections rather than asking you to configure them up front. For example:
Repeatedly removing an indicator (say, "remove RSI" three times) → the planner stops including it by default and tells you: "I'll stop adding RSI by default. Undo?"
Repeatedly switching a Fibonacci anchor to an earlier swing → your swing-sensitivity preference is adjusted.
Repeatedly saying "simplify" → your default detail level is lowered, after you confirm.
This learning follows a few firm rules:
An inference only happens after at least three consistent corrections — a single edit is never over-interpreted as a standing preference.
Every inferred change is surfaced to you with a one-tap undo.
All inferred preferences are visible and editable in a single "Learned preferences" list — nothing is learned silently.
You can freeze learning entirely if you'd rather set everything explicitly.
Session Context
Every time you send a prompt or a refinement, the AI planner receives the full context needed to interpret it correctly:
Current symbol, asset class, and timeframe
Visible time range and chart scale
Current session (regular / pre-market / post-market)
Your selected candles or region, as UI-captured coordinates
Existing user drawings and existing AI drawings, by object ID
Active indicators
Your active analysis context — the object IDs, types, states, and key values from the most recent engine result
Previous commands in the same analysis session
This is what lets short, contextual instructions like "use the earlier swing low" or "remove Fibonacci" work without you having to restate the whole analysis.
Temporary Instructions
You can override any preference for a single analysis without changing your defaults:
Temporary instructions apply only to the analysis you're running — they're never saved as a preference unless you explicitly say so.
Last updated