AI Model Performance Dashboard

#!/usr/bin/env bash

# AI Model Performance Dashboard for Claude Code
# Displays: Occupancy % | Truncation | TTFT | Tokens/min | Model Limits

# Read JSON from stdin
read -r input

# Extract values
model=$(echo "$input" | jq -r '.model // "unknown"')
prompt_tokens=$(echo "$input" | jq -r '.session.promptTokens // 0')
completion_tokens=$(echo "$input" | jq -r '.session.completionTokens // 0')
total_tokens=$(echo "$input" | jq -r '.session.totalTokens // 0')
session_start=$(echo "$input" | jq -r '.session.startTime // ""')

# Model context limits (2025 verified)
case "$model" in
  *"claude-sonnet-4"*|*"claude-4"*)
    context_limit=1000000
    model_display="Claude Sonnet 4"
    ;;
  *"claude-3.5"*|*"claude-sonnet-3.5"*)
    context_limit=200000
    model_display="Claude 3.5 Sonnet"
    ;;
  *"gpt-4.1"*|*"gpt-4-turbo"*)
    context_limit=1000000
    model_display="GPT-4.1 Turbo"
    ;;
  *"gpt-4o"*)
    context_limit=128000
    model_display="GPT-4o"
    ;;
  *"gemini-1.5-pro"*)
    context_limit=2000000
    model_display="Gemini 1.5 Pro"
    ;;
  *"gemini-2"*)
    context_limit=1000000
    model_display="Gemini 2.x"
    ;;
  *"grok-3"*)
    context_limit=1000000
    model_display="Grok 3"
    ;;
  *"grok-4"*)
    context_limit=256000
    model_display="Grok 4"
    ;;
  *"llama-4"*)
    context_limit=10000000
    model_display="Llama 4 Scout"
    ;;
  *)
    context_limit=100000
    model_display="$model"
    ;;
esac

# Calculate occupancy percentage
if [ $context_limit -gt 0 ]; then
  occupancy=$((prompt_tokens * 100 / context_limit))
else
  occupancy=0
fi

# Occupancy color coding
if [ $occupancy -lt 50 ]; then
  OCC_COLOR="\033[38;5;46m"   # Green: < 50%
elif [ $occupancy -lt 80 ]; then
  OCC_COLOR="\033[38;5;226m"  # Yellow: 50-80%
else
  OCC_COLOR="\033[38;5;196m"  # Red: > 80%
fi

# Truncation warning (models fail before advertised limits)
reliable_limit=$((context_limit * 65 / 100))  # 65% of claimed limit
if [ $prompt_tokens -gt $reliable_limit ]; then
  truncation="⚠ TRUNCATION RISK"
  TRUNC_COLOR="\033[38;5;196m"
else
  truncation="✓ Safe"
  TRUNC_COLOR="\033[38;5;46m"
fi

# Calculate tokens per minute
if [ -n "$session_start" ]; then
  current_time=$(date +%s)
  start_epoch=$(date -j -f "%Y-%m-%dT%H:%M:%SZ" "$session_start" +%s 2>/dev/null || echo "$current_time")
  elapsed_seconds=$((current_time - start_epoch))
  if [ $elapsed_seconds -gt 0 ]; then
    tokens_per_min=$((total_tokens * 60 / elapsed_seconds))
  else
    tokens_per_min=0
  fi
else
  tokens_per_min=0
fi

# Format numbers with commas
formatNumber() {
  printf "%'d" "$1" 2>/dev/null || echo "$1"
}

# TTFT simulation (Time to First Token - would need actual timing)
# For demo purposes, estimate based on tokens
if [ $prompt_tokens -gt 50000 ]; then
  ttft="~2.5s"
  TTFT_COLOR="\033[38;5;226m"
elif [ $prompt_tokens -gt 10000 ]; then
  ttft="~1.2s"
  TTFT_COLOR="\033[38;5;46m"
else
  ttft="~0.8s"
  TTFT_COLOR="\033[38;5;46m"
fi

RESET="\033[0m"
BOLD="\033[1m"

# Build dashboard (multi-line for comprehensive view)
echo -e "${BOLD}📊 ${model_display}${RESET}"
echo -e "${OCC_COLOR}Occupancy: ${occupancy}%${RESET} ($(formatNumber $prompt_tokens)/$(formatNumber $context_limit) tokens) | ${TRUNC_COLOR}${truncation}${RESET}"
echo -e "${TTFT_COLOR}TTFT: ${ttft}${RESET} | Rate: ${tokens_per_min} tok/min | Total: $(formatNumber $total_tokens)"

{
  "format": "bash",
  "position": "left",
  "colorScheme": "traffic-light-metrics",
  "refreshInterval": 2000
}