REMOVED ARTIFICIAL CONSTRAINTS:
- Eliminated 3% minimum stop loss requirement
- Eliminated 1% minimum take profit requirement
- AI can now choose ANY percentage based on market analysis
- Updated app/api/drift/trade/route.js to use exact AI percentages
- Removed Math.max() constraints that forced minimums
- AI now has 0.1%+ to 50%+ percentage freedom
- Modified AI_RISK_MANAGEMENT.md to reflect new freedom
- Removed all references to artificial 3%/1% minimums
- Added ultra-tight scalping examples (0.1%-1%)
- Updated volatility guidelines for all trading styles
PROVEN WITH REAL ORDERS:
- Transaction: 35QmCqWFzwJ1X2nm5M8rgExKEMbWTRqxCa1GryEsR595zYwBLqCzDowUYm3J2u13WMvYR2PRoS3eAMSzXfGvEVbe
- Confirmed: 0.5% SL / 0.25% TP working on Drift Protocol
- Verified: Orders visible in Drift UI with correct trigger prices
- Optimal risk management based on actual market conditions
- Support for all trading styles: scalping to position trading
- No more forced suboptimal stops due to artificial limits
- Professional-grade percentage precision
The AI can now freely optimize percentages for maximum trading effectiveness!
- Completely disable automatic periodic cleanup in development mode
- Disable post-analysis cleanup in development mode
- Add multiple checks for active sessions (progress tracker, recent processes, API activity)
- Increase graceful shutdown timeout from 3s to 5s
- Add conservative fallbacks to prevent premature process termination
This prevents the cleanup system from killing browser processes during active analysis,
resolving the issue where automation analysis was being interrupted by cleanup.
- Fixed automation service to use real SOL price (~89) instead of hardcoded 00
- Updated position size calculation to properly convert USD investment to token amount
- Enhanced trade display to show separate entry/exit prices with price difference
- Added data quality warnings for trades with missing exit data
- Updated API to use current SOL price (189.50) and improved trade result determination
- Added detection and warnings for old trades with incorrect price data
Resolves issue where trades showed 9-100 entry prices instead of real SOL price of 89
and position sizes of 2.04 SOL instead of correct ~0.53 SOL for 00 investment
Fixed position size calculation: 00 investment now shows 00 position (was 04.76)
Fixed token amount display: Now shows correct tokens (~0.996) for 00 investment (was 2.04)
Corrected API route: /api/automation/analysis-details now returns 200 instead of 405
Technical changes:
- Updated route calculation logic: tradingAmount / trade.price for correct token amounts
- Fixed displayPositionSize to show intended investment amount
- Used Docker Compose v2 for container management
- Resolved Next.js module export issues
The API now correctly displays trade details matching user investment intentions.
FIXES:
- Enhanced signalAnalysisCycleComplete with more intelligent cleanup logic
- Added active session detection to avoid killing processes during analysis
- Implemented graceful shutdown (SIGTERM) before force kill (SIGKILL)
- Only kills processes older than 2 minutes to avoid disrupting active analysis
- Added 10 second delay in runPostCycleCleanup to ensure trading decision is complete
- Improved process age filtering to prevent premature cleanup
- Cleanup now properly correlates with analysis completion + trading decision
- Reduced aggressive kills that were happening during active analysis
- Better CPU usage management through smarter process lifecycle
- Prevents cleanup from interfering with ongoing analysis work
This should significantly reduce the zombie process CPU usage issue by ensuring
cleanup only happens when analysis work is truly complete and decisions are finalized.
- Moved cleanup trigger from analysis phase to complete automation cycle
- Cleanup now runs after trading decision is made (enter/hold/exit)
- Added comprehensive post-cycle cleanup that waits for graceful shutdown
- Enhanced cleanup coordination with analysis cycle completion signals
- Force cleanup after complete cycle to ensure all zombie processes are killed
- Added cleanup triggers for all cycle outcomes (trade executed, no opportunity, error, etc.)
- Improved timing to wait for browser processes to close properly
- Better correlation between analysis completion and process cleanup
CRITICAL BUG FIX: Cleanup process was interfering with active analysis sessions
- Aggressive cleanup was running during active analysis, causing navigation failures
- Progress tracking was not properly coordinated with cleanup system
- No session state checking before process termination
1. AUTOMATION SERVICE COORDINATION:
- Added proper progress tracking to automation cycles
- Created unique session IDs for each analysis run
- Integrated with progressTracker for session state management
- Added post-analysis cleanup triggers with proper timing
2. ENHANCED CLEANUP INTELLIGENCE:
- Improved session checking with detailed logging of active sessions
- Added process age filtering in development mode (only kill >5min old processes)
- Better error handling when progress tracker import fails
- More granular cleanup control with session state awareness
3. TIMING IMPROVEMENTS:
- Post-analysis cleanup now waits for session completion
- Added proper delays between analysis phases
- Implemented graceful cleanup deferral when sessions are active
- Added delayed cleanup fallback for stuck sessions
4. DEVELOPMENT MODE SAFETY:
- Gentler SIGTERM → SIGKILL progression for development
- Only clean processes older than 5 minutes during dev
- Better logging of process age and cleanup decisions
- Safer fallback behavior when session tracking fails
This resolves the 'Failed to navigate to layout' errors by ensuring cleanup
doesn't interfere with active browser sessions during analysis.
- Replace time-based cleanup with on-demand cleanup in development mode
- Cleanup is triggered immediately after AI analysis completes
- Added runPostAnalysisCleanup() method to aggressive-cleanup service
- Cleanup triggers added to both single and batch analysis endpoints
- More efficient: cleanup happens only when needed, not on timer
- Prevents zombie processes without interfering with active analysis
- Production mode still uses periodic cleanup as backup (10 min intervals)
- Gentle cleanup in development: SIGTERM first, then SIGKILL if needed
- Create /api/batch-analysis endpoint that collects ALL screenshots first
- Then sends all screenshots to AI for comprehensive analysis
- Fixes issue where individual timeframes were analyzed immediately
- Multi-timeframe analysis now provides cross-timeframe consensus
- Update AIAnalysisPanel to use batch analysis for multiple timeframes
- Maintains backward compatibility with single timeframe analysis
- Added aggressive cleanup system that runs every 5 minutes to kill orphaned processes
- Enhanced process cleanup with better signal handling and forced cleanup
- Added startup initialization system to ensure cleanup is properly loaded
- Integrated cleanup system into app layouts for automatic initialization
- Added zombie process cleanup and temp directory cleanup
- Improved Docker container restart behavior for proper process cleanup
- Resolves issue with zombie Chrome processes accumulating
