Session Analyzer

Two tools for session analysis:

1. Model Performance Analysis

Analyzes frustration signals across all sessions (pi, codex, claude) correlated with time of day and model.

# Run analysis (outputs to ./model-performance-analysis/)
~/code/research/pi-skills/session-analyzer/model-performance.js

# Custom output directory
~/code/research/pi-skills/session-analyzer/model-performance.js /path/to/output

Outputs:

• report.md — full analysis with ASCII charts
• chart.html — interactive browser charts
• data.csv — raw data for external graphing
• model-stats.json — aggregated stats per model

Key metrics:

• Frustration rate by hour (CET)
• PST vs non-PST hours comparison
• Model comparison (codex vs opus vs sonnet)
• Worst days analysis

2. Pattern Discovery (Original)

Extracts transcripts and optionally spawns subagents to find automation opportunities.

# Extract transcripts for current directory
~/code/research/pi-skills/session-analyzer/analyze.js

# Extract transcripts for specific directory
~/code/research/pi-skills/session-analyzer/analyze.js /path/to/project

# Extract + analyze with subagents
~/code/research/pi-skills/session-analyzer/analyze.js --analyze

# Custom output directory
~/code/research/pi-skills/session-analyzer/analyze.js --output ./my-analysis --analyze

What It Does

1. Extract: Reads all session files for the given working directory from ~/.pi/agent/sessions/
2. Split: Chunks transcripts into ~100k char files (fits in context window)
3. Analyze (optional): Spawns pi subagents to identify:
   • AGENTS.md patterns: Coding style rules, conventions you repeat
   • Skill patterns: Multi-step workflows you do often
   • Prompt templates: Reusable prompts for common tasks

Output

Without --analyze:

session-transcripts/
├── session-transcripts-000.txt
├── session-transcripts-001.txt
└── ...

With --analyze:

session-transcripts/
├── session-transcripts-000.txt
├── session-transcripts-000.summary.txt  # Pattern analysis
├── session-transcripts-001.txt
├── session-transcripts-001.summary.txt
└── FINAL-SUMMARY.txt                    # Aggregated findings

Setup

Install dependencies (run once):

cd ~/code/research/pi-skills/session-analyzer
npm install

When to Use

• After working on a project for a while, to discover what rules/skills would help
• Periodically to find new automation opportunities
• When you notice you keep giving similar instructions

Source: ferologics/pi-skills Originally adapted from badlogic/pi-mono gist