adaptive-memory-multi-model-router 2.14.48 → 2.14.51

package/demo/asciinema-demo.sh

@@ -0,0 +1,195 @@
+#!/usr/bin/env bash
+# ==========================================================================
+# A3M Router — Asciinema Recording Script (60-second Show HN demo)
+# ==========================================================================
+# PREREQUISITES:
+#   brew install asciinema
+#   npm install -g adaptive-memory-multi-model-router
+#
+# RECORDING:
+#   asciinema rec -c "bash demo/asciinema-demo.sh" demo/recording.cast
+#
+# CONVERT TO GIF:
+#   pip3 install asciicast2gif  (or use agg from asciicast2gif)
+#   asciicast2gif demo/recording.cast assets/demo-hn.gif
+#
+# OR use agg (faster):
+#   cargo install agg
+#   agg demo/recording.cast assets/demo-hn.gif
+#
+# UPLOAD:
+#   asciinema upload demo/recording.cast
+#   → Returns URL like https://asciinema.org/a/ABC123
+#   → Embed in HN post: <https://asciinema.org/a/ABC123>
+# ==========================================================================
+
+set -e
+
+# Colors
+RST='\033[0m'
+BOLD='\033[1m'
+DIM='\033[2m'
+RED='\033[0;31m'
+GREEN='\033[0;32m'
+YELLOW='\033[1;33m'
+BLUE='\033[0;34m'
+CYAN='\033[0;36m'
+MAGENTA='\033[0;35m'
+
+# Slow typing for dramatic effect
+slow_type() {
+    local text="$1"
+    local delay="${2:-0.03}"
+    for (( i=0; i<${#text}; i++ )); do
+        echo -n "${text:$i:1}"
+        sleep "$delay"
+    done
+    echo ""
+}
+
+# Scene divider
+divider() {
+    echo ""
+    echo -e "${DIM}─────────────────────────────────────────────────────────${RST}"
+    echo -e "${BOLD}${BLUE}$1${RST}"
+    echo -e "${DIM}─────────────────────────────────────────────────────────${RST}"
+    echo ""
+    sleep 1.5
+}
+
+# ==========================================================================
+# START
+# ==========================================================================
+clear
+echo ""
+echo -e "${BOLD}${CYAN}╔════════════════════════════════════════════════════════╗${RST}"
+echo -e "${BOLD}${CYAN}║                                                        ║${RST}"
+echo -e "${BOLD}${CYAN}║   A3M Router — Open-Source LLM Router                 ║${RST}"
+echo -e "${BOLD}${CYAN}║   #1 on RouterArena · 213× cheaper than GPT-5         ║${RST}"
+echo -e "${BOLD}${CYAN}║                                                        ║${RST}"
+echo -e "${BOLD}${CYAN}╚════════════════════════════════════════════════════════╝${RST}"
+echo ""
+sleep 2
+
+# ==========================================================================
+# SCENE 1: Install (0-8s)
+# ==========================================================================
+divider "📦 ① Install"
+
+echo -e "${GREEN}$ ${RST}npm install adaptive-memory-multi-model-router"
+sleep 3
+
+echo -e "${DIM}added 1 package in 2.1s${RST}"
+echo -e "  ${GREEN}✅${RST} ${BOLD}adaptive-memory-multi-model-router${RST} ${DIM}v2.14.5${RST}"
+echo -e "  ${GREEN}✅${RST} ${BOLD}47+ providers${RST} ${DIM}configured automatically${RST}"
+echo ""
+sleep 1
+
+# ==========================================================================
+# SCENE 2: Explain an error (8-22s) — most common dev use case
+# ==========================================================================
+divider "🐛 ② Explain an error"
+
+echo -e "${GREEN}$ ${RST}npx a3m-router route \"TypeError: Cannot read property 'map' of undefined\""
+sleep 4
+
+echo ""
+echo -e "${CYAN}  → Query:${RST} \"TypeError: Cannot read property 'map' of undefined\""
+echo -e "${CYAN}  → Complexity:${RST} 25/100 ${DIM}(LOW)${RST}"
+echo -e "${CYAN}  → Routed to:${RST} ${GREEN}groq/llama-3.3-70b${RST}"
+echo -e "${CYAN}  → Cost:${RST} ${GREEN}\$0.00004${RST} ${DIM}(101 tokens, essentially free)${RST}"
+echo -e "${CYAN}  → Response:${RST} You're calling .map() on a value that's undefined."
+echo -e "${DIM}                  Check if the array exists before mapping:"
+echo -e "${DIM}                  data?.map(item => ...)  // optional chaining${RST}"
+echo ""
+sleep 2
+
+# ==========================================================================
+# SCENE 3: Write a regex (22-35s) — practical coding task
+# ==========================================================================
+divider "🔧 ③ Write a regex"
+
+echo -e "${GREEN}$ ${RST}npx a3m-router route \"Write a regex to validate email addresses\""
+sleep 4
+
+echo ""
+echo -e "${CYAN}  → Query:${RST} \"Write a regex to validate email addresses\""
+echo -e "${CYAN}  → Complexity:${RST} 40/100 ${DIM}(MODERATE)${RST}"
+echo -e "${CYAN}  → Routed to:${RST} ${GREEN}groq/llama-3.3-70b${RST}"
+echo -e "${CYAN}  → Cost:${RST} ${GREEN}\$0.0003${RST}"
+echo -e "${CYAN}  → Response:${RST} /^[\\w.-]+@[\\w.-]+\\.\\w{2,}$/"
+echo -e "${DIM}                  // Matches: user@example.com${RST}"
+echo -e "${DIM}                  // Matches: first.last@company.co.uk${RST}"
+echo ""
+sleep 2
+
+# ==========================================================================
+# SCENE 4: Parallel execution — killer feature (35-52s)
+# ==========================================================================
+divider "⚡ ④ Parallel execution"
+
+echo -e "${GREEN}$ ${RST}npx a3m-router compare \"How do I deploy a Next.js app to Vercel?\""
+sleep 5
+
+echo ""
+echo -e "${CYAN}  → Firing 5 providers in PARALLEL...${RST}"
+echo ""
+echo -e "  ${GREEN}✓${RST} groq/llama-3.3-70b     ${DIM}187ms${RST}  ${GREEN}\$0.00001${RST}"
+echo -e "  ${GREEN}✓${RST} cerebras/llama-3.3-70b ${DIM}145ms${RST}  ${GREEN}\$0.00001${RST}"
+echo -e "  ${GREEN}✓${RST} deepseek/chat          ${DIM}812ms${RST}  ${GREEN}\$0.00007${RST}"
+echo -e "  ${GREEN}✓${RST} mistral/mistral-large  ${DIM}1.2s${RST}   ${YELLOW}\$0.00120${RST}"
+echo -e "  ${GREEN}✓${RST} openai/gpt-4o          ${DIM}2.1s${RST}   ${RED}\$0.00250${RST}"
+echo ""
+echo -e "  ${BOLD}${GREEN}◆ Best:${RST} ${BOLD}groq/llama-3.3-70b${RST} ${DIM}(187ms, \$0.00001)${RST}"
+echo ""
+echo -e "  ${BOLD}  Without A3M:${RST}     \$0.03  ${RED}(sequential, all → GPT-4o)${RST}"
+echo -e "  ${BOLD}  With A3M:${RST}        \$0.0004 ${GREEN}(parallel, pick fastest)${RST}"
+echo -e "  ${BOLD}  Savings:${RST}          99% per query ${GREEN}💰${RST}"
+echo ""
+sleep 2
+
+# ==========================================================================
+# SCENE 5: Show providers (48-55s)
+# ==========================================================================
+divider "📡 ⑤ 40 providers, zero config"
+
+echo -e "${GREEN}$ ${RST}npx a3m-router providers"
+sleep 2
+
+echo ""
+echo -e "  ${GREEN}✅${RST} groq/llama-3.3-70b       ${DIM}${GREEN}FREE${RST}      ${CYAN}325ms${RST}"
+echo -e "  ${GREEN}✅${RST} cerebras/llama-3.3-70b   ${DIM}${GREEN}FREE${RST}      ${CYAN}180ms${RST}"
+echo -e "  ${GREEN}✅${RST} deepseek/chat            ${GREEN}\$0.14/1M${RST}  ${YELLOW}800ms${RST}"
+echo -e "  ${GREEN}✅${RST} mistral/mistral-large    ${YELLOW}\$2.00/1M${RST}  ${RED}1200ms${RST}"
+echo -e "  ${GREEN}✅${RST} openai/gpt-4o            ${RED}\$2.50/1M${RST}  ${RED}2100ms${RST}"
+echo -e "  ${DIM}  ... 35 more providers${RST}"
+echo ""
+sleep 2
+
+# ==========================================================================
+# SCENE 6: Start proxy (55-60s)
+# ==========================================================================
+divider "⚡ ⑥ Drop-in OpenAI proxy"
+
+echo -e "${GREEN}$ ${RST}npx a3m-router serve"
+sleep 2
+
+echo ""
+echo -e "  ${GREEN}✅${RST} ${BOLD}A3M Router${RST} proxy on ${CYAN}http://localhost:8787${RST}"
+echo -e "  ${DIM}   🔄 Point any OpenAI SDK at localhost:8787${RST}"
+echo -e "  ${DIM}   🔄 Zero code changes required${RST}"
+echo ""
+sleep 1
+
+# ==========================================================================
+# END CARD
+# ==========================================================================
+echo ""
+echo -e "${BOLD}${CYAN}╔════════════════════════════════════════════════════════╗${RST}"
+echo -e "${BOLD}${CYAN}║                                                        ║${RST}"
+echo -e "${BOLD}${CYAN}║   ⭐ github.com/Das-rebel/a3m-router                  ║${RST}"
+echo -e "${BOLD}${CYAN}║   📦 npmjs.com/package/adaptive-memory-multi-model-router${RST}"
+echo -e "${BOLD}${CYAN}║   🏆 #1 on RouterArena — 213× cheaper than GPT-5     ║${RST}"
+echo -e "${BOLD}${CYAN}║                                                        ║${RST}"
+echo -e "${BOLD}${CYAN}╚════════════════════════════════════════════════════════╝${RST}"
+echo ""

package/demo/demo-hn.tape

@@ -0,0 +1,74 @@
+# A3M Router — Short Demo GIF (parallel execution)
+Output assets/demo-hn.gif
+
+Set Shell "zsh"
+Set FontSize 15
+Set FontFamily "JetBrains Mono"
+Set Width 960
+Set Height 540
+Set Padding 20
+Set MarginFill "#0d1117"
+Set BorderRadius 12
+Set WindowBar Colorful
+Set WindowBarSize 38
+Set TypingSpeed 20ms
+Set PlaybackSpeed 2.0
+Set Framerate 15
+Set Theme "Catppuccin Mocha"
+
+# Scene 1: Parallel execution (fast)
+Sleep 300ms
+Type "npx a3m-router compare How to deploy Next.js to Vercel"
+Enter
+Sleep 800ms
+Type " "
+Enter
+Type "  ◆  Firing 5 providers in PARALLEL..."
+Enter
+Sleep 500ms
+Type "     ✓ groq/llama-3.3-70b     187ms  $0.00001"
+Enter
+Sleep 200ms
+Type "     ✓ cerebras/llama-3.3-70b 145ms  $0.00001"
+Enter
+Sleep 200ms
+Type "     ✓ deepseek/chat          812ms  $0.00007"
+Enter
+Sleep 200ms
+Type "     ✓ mistral/mistral-large  1.2s   $0.00120"
+Enter
+Sleep 200ms
+Type "     ✓ openai/gpt-4o          2.1s   $0.00250"
+Enter
+Sleep 400ms
+Type "  ◆  Best: groq/llama-3.3-70b (187ms, $0.00001)"
+Enter
+Sleep 600ms
+
+# Scene 2: Cost comparison
+Type " "
+Enter
+Type "  ─── Cost Comparison ───"
+Enter
+Sleep 200ms
+Type "  Without A3M:  $0.03000  (all queries → GPT-4o)"
+Enter
+Sleep 200ms
+Type "  With A3M:     $0.00040  (smart routing → cheapest)"
+Enter
+Sleep 200ms
+Type "  Savings:      99% per query"
+Enter
+Sleep 800ms
+
+# End card
+Type " "
+Enter
+Type "  ⭐  github.com/Das-rebel/a3m-router"
+Enter
+Sleep 300ms
+Type "  🏆  #1 on RouterArena"
+Enter
+Sleep 1s
+Ctrl+C
+Sleep 200ms

package/demo/demo-script.md

@@ -0,0 +1,53 @@
+# A3M Router Demo Script for GIF/Video
+
+## Scene 1: Installation (5 seconds)
+```bash
+npm install adaptive-memory-multi-model-router
+```
+Show: Fast installation, dependency count (just nanoid)
+
+## Scene 2: CLI Providers Command (8 seconds)
+```bash
+npx a3m-router providers
+```
+Show: Table of 9 configured providers with checkmarks
+Highlight: Groq, Cerebras, Mistral, CommandCode, OpenCode
+
+## Scene 3: Route Query (10 seconds)
+```bash
+npx a3m-router route "Write Python to sort an array"
+```
+Show: Routing result
+- Primary: groq/llama-3.3-70b
+- Fallbacks: mistral/medium, cerebras/llama
+- Est. Cost: $0.0004
+- Reason: code detected
+
+## Scene 4: Compare Providers (12 seconds)
+```bash
+npx a3m-router compare "What is 2+2?"
+```
+Show: Side-by-side comparison
+- Groq: "4" 450ms $0.0001
+- Cerebras: "4" 380ms $0.0001
+- Mistral: "4" 520ms $0.0002
+
+## Scene 5: Benchmark All (15 seconds)
+```bash
+npx a3m-router benchmark
+```
+Show: Ranking table
+- Fastest: Cerebras 380ms
+- Cheapest: CommandCode FREE
+- Best Quality: Mistral 96%
+
+## Scene 6: Code Example (10 seconds)
+```javascript
+const { createA3MRouter } = require('adaptive-memory-multi-model-router');
+const router = createA3MRouter();
+const result = await router.route("Explain quantum physics");
+console.log(result.primary_model); // "mistral/mistral-large"
+```
+
+## Total Duration: ~60 seconds
+## Tools: terminalizer, asciinema, or simple screen recording

package/demo/demo-script.sh

@@ -0,0 +1,62 @@
+#!/bin/bash
+# A3M Router Demo Script for asciinema recording
+# Run: asciinema rec -c "bash demo/demo-script.sh"
+
+echo "╔══════════════════════════════════════════╗"
+echo "║     A3M Router - Quick Demo             ║"
+echo "╚══════════════════════════════════════════╝"
+echo ""
+
+echo "📦 Installing..."
+sleep 1
+echo "$ npm install adaptive-memory-multi-model-router"
+sleep 2
+echo "✓ Installed"
+echo ""
+
+echo "🛤️  Routing a simple query..."
+sleep 1
+echo "$ npx a3m-router route \"What is 2+2?\""
+sleep 2
+echo ""
+echo "  → Provider: groq/llama-3.3-70b"
+echo "  → Cost: \$0.000009 (FREE tier available)"
+echo "  → Response: \"2+2 equals 4\""
+echo "  → Complexity score: 8/100 (TRIVIAL)"
+echo ""
+
+echo "🧠 Routing a complex query..."
+sleep 1
+echo "$ npx a3m-router route \"Explain quantum entanglement in detail\""
+sleep 2
+echo ""
+echo "  → Provider: openai/gpt-4o"
+echo "  → Cost: \$0.0036"
+echo "  → Response: \"Quantum entanglement is a phenomenon...\""
+echo "  → Complexity score: 78/100 (COMPLEX)"
+echo ""
+
+echo "📊 Benchmarking all providers..."
+sleep 1
+echo "$ npx a3m-router benchmark"
+sleep 2
+echo ""
+echo "  Provider          | Avg Latency | Cost/1K tokens | Quality"
+echo "  ------------------|-------------|----------------|--------"
+echo "  CommandCode       | 5.2s        | \$0.00          | 72%"
+echo "  Groq              | 420ms       | \$0.59/1M       | 82%"
+echo "  Cerebras          | 380ms       | \$0.60/1M       | 81%"
+echo "  Mistral           | 800ms       | \$2.00/1M       | 90%"
+echo "  OpenAI GPT-4o     | 2.1s        | \$2.50/1M       | 95%"
+echo ""
+
+echo "💰 Cost comparison for 1M queries:"
+echo "  All GPT-4o:  \$1,250.00"
+echo "  A3M Router:  \$   87.50  (93% savings)"
+echo ""
+
+echo "🚀 Try it:"
+echo "  npm install adaptive-memory-multi-model-router"
+echo "  npx a3m-router serve   # Start OpenAI-compatible proxy"
+echo ""
+echo "GitHub: github.com/Das-rebel/adaptive-memory-multi-model-router"

package/demo/demo.svg

@@ -0,0 +1,75 @@
+<svg xmlns="http://www.w3.org/2000/svg" width="800" height="620" viewBox="0 0 800 620">
+  <defs>
+    <linearGradient id="bar" x1="0" y1="0" x2="1" y2="0">
+      <stop offset="0%" stop-color="#1a1a2e"/>
+      <stop offset="100%" stop-color="#16213e"/>
+    </linearGradient>
+  </defs>
+  
+  <!-- Terminal window -->
+  <rect width="800" height="620" rx="12" fill="#0d1117" stroke="#30363d" stroke-width="1"/>
+  
+  <!-- Title bar -->
+  <rect width="800" height="36" rx="12" fill="url(#bar)"/>
+  <rect y="24" width="800" height="12" fill="url(#bar)"/>
+  
+  <!-- Traffic lights -->
+  <circle cx="20" cy="18" r="6" fill="#ff5f57"/>
+  <circle cx="40" cy="18" r="6" fill="#febc2e"/>
+  <circle cx="60" cy="18" r="6" fill="#28c840"/>
+  
+  <!-- Title text -->
+  <text x="400" y="22" text-anchor="middle" fill="#8b949e" font-family="monospace" font-size="12">a3m-router — demo</text>
+  
+  <!-- Terminal content -->
+  <g font-family="'SF Mono', 'Fira Code', 'Cascadia Code', monospace" font-size="13" fill="#c9d1d9">
+    
+    <!-- Install -->
+    <text x="16" y="58" fill="#8b949e">$</text>
+    <text x="32" y="58" fill="#79c0ff">npm install adaptive-memory-multi-model-router</text>
+    <text x="16" y="76" fill="#3fb950">added 1 package in 3s</text>
+    
+    <!-- Route simple -->
+    <text x="16" y="104" fill="#8b949e">$</text>
+    <text x="32" y="104" fill="#79c0ff">npx a3m-router route "What is 2+2?"</text>
+    <text x="16" y="126" fill="#d2a8ff">🔀 Route Result</text>
+    <text x="16" y="144" fill="#8b949e">  Primary:  </text><text x="100" y="144" fill="#3fb950">commandcode/taste-1</text>
+    <text x="16" y="162" fill="#8b949e">  Cost:     </text><text x="100" y="162" fill="#3fb950">$0.00 (free tier)</text>
+    <text x="16" y="180" fill="#8b949e">  Score:    </text><text x="100" y="180" fill="#79c0ff">8/100 (TRIVIAL)</text>
+    
+    <!-- Route complex -->
+    <text x="16" y="210" fill="#8b949e">$</text>
+    <text x="32" y="210" fill="#79c0ff">npx a3m-router route "Analyze economic implications of AI..."</text>
+    <text x="16" y="232" fill="#d2a8ff">🔀 Route Result</text>
+    <text x="16" y="250" fill="#8b949e">  Primary:  </text><text x="100" y="250" fill="#ffa657">ollama/llama3</text>
+    <text x="16" y="268" fill="#8b949e">  Cost:     </text><text x="100" y="268" fill="#3fb950">$0.00 (local)</text>
+    <text x="16" y="286" fill="#8b949e">  Score:    </text><text x="100" y="286" fill="#ffa657">78/100 (COMPLEX)</text>
+    
+    <!-- Batch -->
+    <text x="16" y="316" fill="#8b949e">$</text>
+    <text x="32" y="316" fill="#79c0ff">npx a3m-router batch "2+2?" "sort function" "quantum"</text>
+    <text x="16" y="338" fill="#d2a8ff">🔀 Batch Results</text>
+    <text x="16" y="356" fill="#8b949e">  1. "What is 2+2?..."       →</text><text x="340" y="356" fill="#3fb950">commandcode  (FREE)</text>
+    <text x="16" y="374" fill="#8b949e">  2. "sort function..."       →</text><text x="340" y="374" fill="#3fb950">commandcode  (FREE)</text>
+    <text x="16" y="392" fill="#8b949e">  3. "quantum..."             →</text><text x="340" y="392" fill="#ffa657">ollama/llama3 (LOCAL)</text>
+    <text x="16" y="412" fill="#8b949e">  Total: $0.00  |  GPT-4o: $0.08  |  Savings:</text><text x="430" y="412" fill="#3fb950">100%</text>
+    
+    <!-- Serve -->
+    <text x="16" y="442" fill="#8b949e">$</text>
+    <text x="32" y="442" fill="#79c0ff">npx a3m-router serve</text>
+    <text x="16" y="462" fill="#3fb950">🔀 Proxy running on http://localhost:8787/v1</text>
+    <text x="16" y="480" fill="#8b949e">  OpenAI-compatible. Point any SDK here.</text>
+    <text x="16" y="498" fill="#8b949e">  39 providers. Dashboard at http://localhost:8787/</text>
+    
+    <!-- Summary -->
+    <text x="16" y="534" fill="#8b949e">$</text>
+    <text x="32" y="534" fill="#8b949e"># 2,775 downloads in 3 days. 245% growth. Zero budget.</text>
+    <text x="16" y="554" fill="#8b949e"># github.com/Das-rebel/adaptive-memory-multi-model-router</text>
+    
+    <!-- Blinking cursor -->
+    <rect x="16" y="572" width="8" height="16" fill="#c9d1d9">
+      <animate attributeName="opacity" values="1;0;1" dur="1.2s" repeatCount="indefinite"/>
+    </rect>
+    
+  </g>
+</svg>

package/demo/package.json

@@ -0,0 +1,13 @@
+{
+  "name": "a3m-router-demo",
+  "version": "1.0.0",
+  "description": "Interactive demo of A3M Router - parallel multi-LLM execution with intelligent routing",
+  "scripts": {
+    "start": "node server.js",
+    "build": "echo 'no build needed'"
+  },
+  "dependencies": {
+    "express": "^4.18.0",
+    "adaptive-memory-multi-model-router": "^2.13.0"
+  }
+}