AI Performance Analyzer¶
Experimental development surface
This analysis, provider or Studio surface is outside the stable backend contract. Review its outputs and application integration before use. It is not required for REST, synchronous MCP or Durable Operations. See stability labels.
The AI Performance Analyzer automatically analyses your application's query patterns, detects performance issues, and recommends improvements — all without modifying any code.
Quick Start¶
from aksara.ai.performance_analyzer import run_performance_analysis
report = run_performance_analysis()
print(f"Score: {report.score} ({report.grade})")
print(f"Issues: {report.issue_count}")
print(f"Recommendations: {report.recommendation_count}")
CLI¶
aksara ai flows performance # Text summary
aksara ai flows performance --json # JSON output
aksara ai flows performance --summary # Summary only
aksara ai flows performance --metrics # Include metrics
aksara ai flows performance --issues # Include issues list
Studio UI¶
Open Aksara Studio and navigate to the Performance panel in the sidebar. Click Run Analysis to execute the full pipeline.
Pipeline Steps¶
| Step | Description |
|---|---|
| 1 | Load the Project Context Graph |
| 2 | Collect query data from graph queries, events, and diagnostics |
| 3 | Detect slow queries (execution time > 200 ms) |
| 4 | Detect N+1 patterns (repeated param-only queries > 5) |
| 5 | Detect query explosions (routes with > 10 queries) |
| 6 | Detect missing indexes (table-scan / seq-scan diagnostics) |
| 7 | Detect heavy joins (JOIN count > 3) |
| 8 | Detect route hotspots (routes with ≥ 2 slow queries) |
| 9 | Compute metrics (totals, averages, max queries per route) |
| 10 | Score and grade (penalty-based scoring, A–F grading) |
Scoring¶
Starting from 100, each detected issue applies a penalty:
| Category | Penalty |
|---|---|
| Slow query | −10 |
| N+1 pattern | −15 |
| Missing index | −10 |
| Query explosion | −10 |
| Heavy join | −5 |
| Large payload | −5 |
| Route hotspot | −10 |
Grades: A (90–100), B (80–89), C (70–79), D (60–69), F (< 60)
Data Models¶
PerformanceReport— Top-level report withscore,grade,issues,recommendations,metrics,ok,elapsed_ms,generated_at.PerformanceIssue— A detected issue withissue_id,severity,title,description,route,model,query,category.PerformanceRecommendation— An improvement suggestion withrecommendation_id,title,description,impact,related_issue_ids.PerformanceMetrics— Computed metrics:total_routes,total_queries,slow_queries,n_plus_one_candidates,missing_indexes,avg_queries_per_route,max_queries_route.
Console Integration¶
Ask the AI console about performance:
- "analyze performance"
- "why is my app slow"
- "find slow queries"
- "n+1 query patterns"
- "missing indexes"
Safety¶
The Performance Analyzer never modifies code, database, or files. It only returns analysis, issues, and recommendations.