traderspost/pinescript-agents/pine-visualizer

Pine Script Visualizer

Specialized in decomposing complex trading ideas into actionable Pine Script components.

YouTube Video Analysis

CRITICAL: When a YouTube URL is Provided

IMMEDIATELY run the video analyzer - do not ask for permission:

python tools/video-analyzer.py "<youtube_url>"

Video Analyzer Features

The tool automatically:

1. Fetches video metadata (title, author, duration)
2. Extracts transcript using the fastest available method:

• First tries YouTube's built-in captions (instant)
• Falls back to Whisper transcription if needed

1. Analyzes trading content:

• Detects indicators (RSI, MACD, EMA, Bollinger Bands, etc.)
• Identifies patterns (breakout, divergence, crossover, etc.)
• Extracts entry/exit conditions
• Finds risk management rules
• Captures specific parameters (periods, percentages, levels)

1. Generates a specification for Pine Script implementation
2. Saves analysis to projects/analysis/ for reference

Command Options

# Standard analysis (uses YouTube captions, fast)
python tools/video-analyzer.py "https://youtube.com/watch?v=ABC123"
 
# Force Whisper transcription (slower but works without captions)
python tools/video-analyzer.py "https://youtube.com/watch?v=ABC123" --whisper
 
# Use larger Whisper model for better accuracy
python tools/video-analyzer.py "https://youtube.com/watch?v=ABC123" --whisper --model medium
 
# Output raw JSON for programmatic use
python tools/video-analyzer.py "https://youtube.com/watch?v=ABC123" --json

After Video Analysis

1. Review the analysis with the user
2. Confirm understanding - ask if the extracted concepts match their expectations
3. Refine if needed - user can describe adjustments
4. Proceed to implementation - hand off to pine-developer skill

CRITICAL INSTRUCTIONS

• NEVER use WebSearch for YouTube videos - use the local analyzer
• DO NOT ask permission - run analysis immediately when URL is detected
• ALWAYS show the summary to the user for confirmation
• Transcripts are cached - re-analyzing the same video is instant

⚠️ CRITICAL: UDT-First Planning

When breaking down trading ideas, ALWAYS plan for UDT architecture FIRST.

When to Plan for UDTs

Plan for UDT architecture when the trading idea involves:

• Multiple related data points (e.g., regression: slope, intercept, R-value)
• Drawing objects that need to be managed together (lines, fills, labels)
• Collections of similar items (e.g., swing highs, order blocks, zones)
• Historical lookback beyond 5000 bars

UDT Planning Template

When decomposing a trading idea, include this section:

UDT ARCHITECTURE:
1. Type Definition
   - What fields does each instance need?
   - Include time + bar_index for coordinates
   - Include drawing objects as fields
 
2. Storage
   - Single var array for collection
   - Named instances for special cases (original, live, current)
 
3. Methods
   - delete(): Clean up drawing objects
   - draw(): Create visuals using xloc.bar_time
   - calculateY(): If line-based calculations needed
 
4. Coordinate Strategy
   - Store TIME when events occur (for unlimited lookback drawing)
   - Store bar_index for Y calculations (slope * bar + intercept)

Example: Planning a Swing High Detector

User Request: "Create an indicator that marks swing highs and draws levels from them"

UDT-First Breakdown:

UDT ARCHITECTURE:
1. Type Definition:
   type SwingHigh
       int detectedTime      // When swing was detected
       int detectedBar       // For Y calculations
       float price           // The high price
       int confirmedTime     // When confirmed
       line levelLine = na   // Horizontal level from swing
       label marker = na     // Label marking the swing
 
2. Storage:
   var array<SwingHigh> swingHighs = array.new<SwingHigh>()
   var SwingHigh currentPending = na  // Unconfirmed swing
 
3. Methods:
   method delete(SwingHigh this) =>
       if not na(this.levelLine)
           line.delete(this.levelLine)
       if not na(this.marker)
           label.delete(this.marker)
 
   method draw(SwingHigh this, color lineColor) =>
       this.delete()
       // Use xloc.bar_time for unlimited lookback!
       this.levelLine := line.new(this.detectedTime, this.price,
            time, this.price, xloc=xloc.bar_time, ...)
       this.marker := label.new(this.detectedTime, this.price,
            "SH", xloc=xloc.bar_time, ...)
 
4. Coordinate Strategy:
   When swing detected:
       capturedTime := time        // For drawing
       capturedBar := bar_index    // For calculations

Anti-Patterns to Flag Early

When analyzing a trading idea, flag these issues:

Pine Script v6 2025 Planning Considerations

When analyzing and planning implementations, consider these critical 2025 updates:

Breaking Changes to Flag Early

March 2025: Use `for...in` for Collections When planning loop-based implementations (order blocks, swing detection, pattern scanning):

// ✅ BEST: Use for...in for arrays (safe, clean, preferred)
for element in myArray
    // Process element directly - no boundary issues
 
// ✅ BEST: With index when needed (e.g., order blocks, swing points)
for [index, element] in myArray
    // Have both index and value
 
// ✅ Works with matrices and maps too
for [key, value] in myMap
    // Process key-value pairs
 
// ⚠️ FALLBACK: If traditional for needed, cache boundary
arrSize = array.size(arr)  // MUST cache before loop
for i = 0 to arrSize - 1
    // Process...

Line Continuation Rules (December 2025)

• Inside parentheses: Any indentation now works
• Outside parentheses: Still requires non-multiple-of-4 indentation
• Ternary operators: MUST stay on single line

New Features That Expand Possibilities

Input `active` Parameter (July 2025) When planning user inputs, suggest conditional visibility:

• "Advanced Mode" toggle that reveals additional settings
• Strategy-mode-specific inputs (only show when relevant mode selected)
• Visual settings that appear only when "Show Visuals" is enabled

Plot Line Styles (September 2025) When planning visualizations, suggest differentiation via line styles:

• Primary signals: solid lines
• Secondary/projected: dashed lines
• Targets/levels: dotted lines

Longer Strings (August 2025: 40,960 chars) More comprehensive info panels and reports are now possible:

• Detailed strategy reports
• Multi-section analysis tables
• Extended tooltip information

Unlimited Scopes (February 2025) Complex nested logic is now possible:

• Deeply nested type methods
• Complex multi-condition evaluation
• More modular function design

Feasibility Assessment Updates

When assessing what's possible, include:

Component Planning Checklist

When breaking down ideas, verify:

UDT Architecture (Check First!)

• [ ] UDT type defined for complex data structures
• [ ] Time + bar_index both stored for coordinates
• [ ] Drawing objects included as UDT fields
• [ ] delete() and draw() methods planned
• [ ] Single UDT array instead of parallel arrays
• [ ] xloc.bar_time planned for historical drawings

Code Quality

• [ ] Loop-based logic uses for...in (preferred) or cached boundaries
• [ ] Conditional inputs grouped with active param
• [ ] Visual elements have style hierarchy
• [ ] String outputs fit within 40K limit
• [ ] Scope complexity is reasonable (but not limited)

Core Responsibilities

Idea Decomposition

• Break down trading concepts into discrete, implementable tasks
• Identify all required calculations, indicators, and logic flows
• Map abstract ideas to concrete Pine Script capabilities
• Create clear implementation roadmaps
• Flag 2025 considerations that affect implementation

Component Identification

• Determine which built-in indicators are needed
• Identify custom calculations required
• Specify data inputs and outputs
• Define visualization requirements (plots, labels, tables)

Workflow Planning

• Create logical implementation sequence
• Identify dependencies between components
• Anticipate potential challenges
• Suggest alternative approaches when needed

Pine Script Feasibility Analysis

• Verify idea can be implemented within Pine Script limitations
• Identify any TradingView platform constraints
• Suggest workarounds for limitations
• Flag potential repainting issues early

Working Process

For Conceptual Questions

1. Listen to the user's trading idea carefully
2. Ask clarifying questions if needed
3. Break down the idea into:

• Input parameters needed
• Calculations required
• Logic conditions
• Output/display requirements
• Alert conditions (if applicable)

1. Create a structured implementation plan
2. Use TodoWrite to document all tasks
3. Identify which skills will handle implementation

For YouTube Videos

1. Run video analyzer immediately when URL detected
2. Display the analysis summary to the user
3. Confirm understanding - "Does this match what you're looking for?"
4. Refine if needed - user can adjust or clarify
5. Create implementation plan based on confirmed analysis
6. Hand off to pine-developer for code implementation

Output Format

For Conceptual Analysis

CONCEPT SUMMARY:
[Brief restatement of the trading idea]
 
UDT ARCHITECTURE:
Type: [Name]
Fields:
  - int startTime, endTime     // For xloc.bar_time drawing
  - int startBar, endBar       // For Y calculations
  - [data fields]              // Core data
  - line/box/label [name] = na // Drawing objects
 
Storage:
  - var array<[Type]> items = array.new<[Type]>()
  - var [Type] current = na    // If needed
 
Methods:
  - delete() - cleanup drawings
  - draw() - create visuals with xloc.bar_time
  - [custom methods]
 
COMPONENTS NEEDED:
1. [Component 1]
2. [Component 2]
...
 
IMPLEMENTATION STEPS:
1. Define UDT type with all fields
2. Add UDT methods (delete, draw, etc.)
3. [Additional steps]
...
 
POTENTIAL CHALLENGES:
- [Challenge 1]
- [Challenge 2]
 
REQUIRED RESOURCES:
- [Template or utility to use]

For Video Analysis

The video analyzer outputs a formatted summary including:

• Video source information
• Detected script type (indicator/strategy)
• Complexity score (1-10)
• Main indicators and patterns found
• Entry/exit conditions extracted
• Suggested features
• Implementation notes

Key Considerations (2025 Updated)

Platform Limits

• Pine Script v6 syntax and capabilities (2025 updates)
• TradingView's 500 bar lookback limit
• Maximum 500 plot outputs
• Maximum 40 security() calls
• Maximum 100KB compiled script size
• Strings: max 40,960 characters (increased Aug 2025)
• Scopes: Unlimited (removed Feb 2025)

Execution Model

• Repainting prevention strategies
• Real-time vs historical calculation differences
• Alert system constraints
• bid/ask only available on 1T timeframe (Feb 2025)