Skip to content

Latest commit

 

History

History
227 lines (152 loc) · 4.94 KB

GUI Usage Guide.md

File metadata and controls

227 lines (152 loc) · 4.94 KB

Quant Kernel GUI — Usage Guide

This GUI is a configuration-driven market scanner and renderer. You declare what to fetch, compose indicators & strategies, and the system executes + renders incrementally.

The UI is divided into four logical regions:

[ API Selection ]   [ Fetch Config ]   [ Run ]
[ Indicators ]      [ Strategies ]     [ Charts ]

Quant Kernel Screenshot

1. API Selection Panel

This section defines where tickers come from and how many are processed.

Fields

Field Meaning
api (QKApi) Data provider (yfinance, upstox, dhan)
exchange Exchange namespace (e.g. NSE)
start_index Starting index into ticker list
end_index Ending index (exclusive)
enable_filter Enable signal-based filtering
filter_last_n Lookback window (last N candles)

How ticker ranges work

Tickers are loaded from your YAML source once, then sliced:

tickers[start_index : end_index]

Examples:

  • 0 → 10 → first 10 tickers
  • 10 → 20 → next batch
  • Enables batch scanning without restarting the app

Signal Filter (Optional)

When enable_filter is checked:

  • For each ticker:

    • Look at last N candles
    • If any strategy emits BUY or SELL → chart is rendered
    • If only HOLD or no signals → ticker is skipped

This:

  • avoids rendering noise
  • conserves memory
  • scales to large ticker sets

2. Fetch Config Panel

Controls how market data is fetched.

Fields

Field Meaning
mode historical or intraday
interval Candle interval (days / minutes)
intraday_interval Required only for intraday
from_date Start date (YYYY-MM-DD)
to_date End date
unit Time unit (days, etc.)

Notes

  • Dates are validated but not auto-corrected
  • Changing fetch config invalidates all downstream computation
  • GUI does not cache results — every run is explicit

3. Indicators Panel

Indicators are pure feature generators.

How to use

  1. Select an indicator from Indicator Config
  2. Fill in parameters
  3. Click Add
  4. Repeat to add multiple indicators

Key behavior

  • Multiple instances allowed
  • Same indicator + same params → deduplicated internally
  • Indicators do not generate signals
  • Indicators are computed once per run

4. Strategies Panel

Strategies convert indicators → signals.

How to use

  1. Select a strategy
  2. Configure parameters
  3. Click Add
  4. Multiple strategies can coexist

Behavior

  • Each strategy instance gets a unique signal column
  • BUY / SELL / HOLD values are written into the DataFrame
  • Strategies declare which indicators they need
  • Strategy execution happens after all indicators

5. Run Pipeline Button

When you press RUN PIPELINE:

Execution flow

UI State
  ↓
Ticker Resolution (range-based)
  ↓
For each ticker:
  → Fetch Data
  → Run Indicators
  → Run Strategies
  → Apply Signal Filter (optional)
  → Append Chart

Important details

  • Execution runs in a background thread
  • UI remains responsive
  • Charts appear incrementally
  • Failed tickers do not block others

6. Market Charts View (Right Panel)

This is a scrollable multi-chart renderer.

Behavior

  • One ticker → one chart → one Matplotlib figure
  • Charts are stacked vertically
  • Scrollbar appears automatically
  • No shared axes between tickers

What is rendered

  • Candlesticks (OHLC)

  • Indicator overlays (MA, VWAP, etc.)

  • Strategy signals:

    • ▲ BUY
    • ▼ SELL

  • Secondary Y-axis for percentage-based indicators

Clearing behavior

  • Each run clears previous charts
  • Charts are destroyed, not hidden
  • Prevents figure leaks and memory growth

7. Typical Usage Patterns

Scan in batches

Run 1: start=0, end=25
Run 2: start=25, end=50
Run 3: start=50, end=75

Signal-only scan

  • Enable filter
  • Set filter_last_n = 3
  • Render only actionable tickers

Strategy comparison

  • Add multiple strategies
  • Compare signals visually per ticker

What the GUI Does NOT Do

  • ❌ No automatic trading
  • ❌ No hidden state
  • ❌ No auto-refresh
  • ❌ No implicit caching

Every run is explicit and reproducible.

Mental Model (Important)

  • UI = configuration
  • Controller = orchestration
  • DataFrame = truth
  • Renderer = visualization only

Once you understand that, everything else clicks.