Files
gh-shakes-tzd-contextune/commands/ctx-stats.md
2025-11-30 08:56:10 +08:00

5.3 KiB
Raw Blame History

name, description, keywords, executable
name description keywords executable
ctx:stats View Contextune detection statistics
show stats
statistics
detection stats
performance metrics
stats
metrics
show statistics
commands/slashsense-stats.py

Contextune Statistics

Display detection performance metrics and usage statistics from the observability database.


Execution

This command runs automatically via the executable script. The markdown provides documentation only.

Script: commands/slashsense-stats.py Execution: Automatic when command is triggered Data Source: ~/.claude/plugins/contextune/data/observability.db


What This Command Does

Step 1: Load Statistics

Reads detection data from the observability database:

sqlite3 ~/.claude/plugins/contextune/data/observability.db \
  "SELECT tier, COUNT(*), AVG(latency_ms), AVG(confidence)
   FROM detections GROUP BY tier"

Step 2: Generate Report

Creates formatted output using Rich library showing:

  1. Detection Performance by Tier

    • Keyword: Detection count, average latency, accuracy
    • Model2Vec: Detection count, average latency, accuracy
    • Semantic Router: Detection count, average latency, accuracy
  2. Top Detected Commands

    • Command name and frequency count
    • Shows top 5 most-used commands
  3. Confidence Distribution

    • Breakdown by confidence range (50-70%, 70-85%, 85%+)
    • Visual progress bars

Step 3: Display to User

Outputs formatted tables and panels to terminal.


Example Output

╭─────────────────────────── Contextune Statistics ───────────────────────────╮
│                                                                              │
│ Total Detections: 1,247                                                     │
│                                                                              │
│ Performance by Tier                                                         │
│ ┌───────────────┬────────────┬─────────────┬──────────┐                    │
│ │ Tier          │ Detections │ Avg Latency │ Accuracy │                    │
│ ├───────────────┼────────────┼─────────────┼──────────┤                    │
│ │ Keyword       │        892 │      0.05ms │     98%  │                    │
│ │ Model2Vec     │        245 │      0.18ms │     94%  │                    │
│ │ Semantic      │        110 │     47.30ms │     89%  │                    │
│ └───────────────┴────────────┴─────────────┴──────────┘                    │
│                                                                              │
│ Top Commands                                                                │
│ 1. /sc:analyze       324 detections                                         │
│ 2. /sc:implement     218 detections                                         │
│ 3. /sc:test          187 detections                                         │
│ 4. /sc:git           156 detections                                         │
│ 5. /sc:improve       134 detections                                         │
│                                                                              │
╰──────────────────────────────────────────────────────────────────────────────╯

Data Sources

If observability.db exists:

  • Shows actual detection data
  • Real latency measurements
  • Actual command frequencies

If observability.db doesn't exist:

  • Shows example/mock data (for demonstration)
  • Indicates data is not from actual usage

Interpreting Results

Tier Performance:

  • Keyword (Target: <0.1ms): Fastest, highest accuracy, handles 60% of queries
  • Model2Vec (Target: <1ms): Fast, good accuracy, handles 30% of queries
  • Semantic Router (Target: <100ms): Slower, handles complex/ambiguous 10%

Latency Analysis:

  • < 1ms: Excellent (no perceptible delay)
  • 1-10ms: Good (barely noticeable)
  • 10-50ms: Acceptable (slight delay)
  • 100ms: Needs optimization

Accuracy Expectations:

  • 95%+: Excellent (trust the detection)
  • 85-95%: Good (verify before auto-execute)
  • 70-85%: Fair (suggest to user)
  • < 70%: Skip (don't suggest)

Troubleshooting

"No data available":

  No detection data found. Using example statistics.
  • This is normal for new installations
  • Data accumulates as you use Contextune
  • Mock data shows what stats will look like

"Database error":

  • Check: ls ~/.claude/plugins/contextune/data/observability.db
  • Permissions: Ensure readable
  • Corruption: Delete and let it recreate on next detection

  • /ctx:usage - View token usage and cost optimization
  • /ctx:help - View all available commands
  • /ctx:configure - Configure Contextune settings