@blockrun/clawrouter 0.8.23 → 0.8.25

package/README.md

@@ -30,6 +30,7 @@ _The world's first hackathon run entirely by AI agents, powered by USDC_
 
 ## Why ClawRouter?
 
+- **4 routing profiles** — auto (balanced), eco (95.9-100% savings), premium (best quality), free (zero cost)
 - **100% local routing** — 15-dimension weighted scoring runs on your machine in <1ms
 - **Zero external calls** — no API calls for routing decisions, ever
 - **30+ models** — OpenAI, Anthropic, Google, DeepSeek, xAI, Moonshot through one wallet
@@ -73,13 +74,21 @@ Done! Smart routing (`blockrun/auto`) is now your default model.
 
 **For advanced users:** See the [complete manual installation guide](docs/windows-installation.md) with step-by-step PowerShell instructions.
 
-### Tips
+### Routing Profiles
 
-- **Use `/model blockrun/auto`** in any conversation to switch on the fly
-- **Free tier?** Use `/model free` — routes to gpt-oss-120b at $0
-- **Model aliases:** `/model sonnet`, `/model grok`, `/model deepseek`, `/model kimi`
-- **Want a specific model?** Use `blockrun/openai/gpt-4o` or `blockrun/anthropic/claude-sonnet-4`
-- **Already have a funded wallet?** `export BLOCKRUN_WALLET_KEY=0x...`
+Choose your routing strategy with `/model <profile>`:
+
+| Profile | Strategy | Savings | Use Case |
+|---------|----------|---------|----------|
+| `/model auto` | Balanced (default) | 74-100% | Best overall balance |
+| `/model eco` | Cost optimized | 95.9-100% | Maximum savings |
+| `/model premium` | Quality focused | 0% | Best quality (Opus 4.5) |
+| `/model free` | Free tier only | 100% | Zero cost |
+
+**Other shortcuts:**
+- **Model aliases:** `/model sonnet`, `/model grok`, `/model gpt5`, `/model o3`
+- **Specific models:** `blockrun/openai/gpt-4o` or `blockrun/anthropic/claude-sonnet-4`
+- **Bring your wallet:** `export BLOCKRUN_WALLET_KEY=0x...`
 
 ---
 
@@ -116,6 +125,27 @@ No external classifier calls. Ambiguous queries default to the MEDIUM tier (Grok
 
 **Deep dive:** [15-dimension scoring weights](docs/configuration.md#scoring-weights) | [Architecture](docs/architecture.md)
 
+### Routing Profiles (NEW in v0.8.21)
+
+ClawRouter now offers 4 routing profiles to match different priorities:
+
+| Profile | Strategy | Savings vs Opus 4.5 | When to Use |
+|---------|----------|---------------------|-------------|
+| **auto** (default) | Balanced quality + cost | 74-100% | General use, best overall |
+| **eco** | Maximum cost savings | 95.9-100% | Budget-conscious, high volume |
+| **premium** | Best quality only | 0% | Mission-critical tasks |
+| **free** | Free tier only | 100% | Testing, empty wallet |
+
+Switch profiles anytime: `/model eco`, `/model premium`, `/model auto`
+
+**Example:**
+```
+/model eco                    # Switch to cost-optimized routing
+"Write a React component"     # Routes to DeepSeek ($0.28/$0.42)
+                              # vs Auto → Grok ($0.20/$1.50)
+                              # 98.3% savings vs Opus 4.5
+```
+
 ### Tier → Model Mapping
 
 | Tier      | Primary Model           | Cost/M | Savings vs Opus |

package/dist/cli.js

@@ -1126,14 +1126,12 @@ var DEFAULT_ROUTING_CONFIG = {
   tiers: {
     SIMPLE: {
       primary: "nvidia/kimi-k2.5",
-      // Ultra-cheap $0.001/$0.001
+      // $0.55/$2.5 - best quality/price for simple tasks
       fallback: [
-        "google/gemini-2.5-flash",
         "nvidia/gpt-oss-120b",
-        "nvidia/gpt-oss-20b",
-        "deepseek/deepseek-chat",
-        "xai/grok-code-fast-1"
-        // Added for better quality fallback
+        // FREE fallback
+        "google/gemini-2.5-flash",
+        "deepseek/deepseek-chat"
       ]
     },
     MEDIUM: {
@@ -1147,9 +1145,15 @@ var DEFAULT_ROUTING_CONFIG = {
       ]
     },
     COMPLEX: {
-      primary: "google/gemini-2.5-pro",
-      fallback: ["xai/grok-4-0709", "openai/gpt-4o", "openai/gpt-5.2", "anthropic/claude-sonnet-4"]
-      // Grok first for cost efficiency, Sonnet as last resort
+      primary: "google/gemini-3-pro-preview",
+      // Latest Gemini - upgraded from 2.5
+      fallback: [
+        "google/gemini-2.5-pro",
+        "xai/grok-4-0709",
+        "openai/gpt-4o",
+        "openai/gpt-5.2",
+        "anthropic/claude-sonnet-4"
+      ]
     },
     REASONING: {
       primary: "xai/grok-4-1-fast-reasoning",
@@ -1157,7 +1161,8 @@ var DEFAULT_ROUTING_CONFIG = {
       fallback: [
         "xai/grok-4-fast-reasoning",
         "openai/o3",
-        // Strong reasoning model
+        "openai/o4-mini",
+        // Latest o-series mini
         "deepseek/deepseek-reasoner",
         "moonshot/kimi-k2.5"
       ]
@@ -1167,8 +1172,8 @@ var DEFAULT_ROUTING_CONFIG = {
   ecoTiers: {
     SIMPLE: {
       primary: "nvidia/kimi-k2.5",
-      // $0.001/$0.001
-      fallback: ["deepseek/deepseek-chat", "nvidia/gpt-oss-120b", "nvidia/gpt-oss-20b"]
+      // $0.55/$2.5
+      fallback: ["nvidia/gpt-oss-120b", "deepseek/deepseek-chat", "google/gemini-2.5-flash"]
     },
     MEDIUM: {
       primary: "deepseek/deepseek-chat",
@@ -1200,13 +1205,25 @@ var DEFAULT_ROUTING_CONFIG = {
     },
     COMPLEX: {
       primary: "anthropic/claude-opus-4.5",
-      // $15/$75
-      fallback: ["openai/gpt-5.2", "anthropic/claude-sonnet-4", "google/gemini-2.5-pro"]
+      // $5/$25 - Latest Opus
+      fallback: [
+        "openai/gpt-5.2-pro",
+        // $21/$168 - Latest GPT pro
+        "google/gemini-3-pro-preview",
+        // Latest Gemini
+        "openai/gpt-5.2",
+        "anthropic/claude-sonnet-4"
+      ]
     },
     REASONING: {
       primary: "openai/o3",
-      // $10/$40
-      fallback: ["anthropic/claude-opus-4.5", "openai/o1", "google/gemini-2.5-pro"]
+      // $2/$8 - Best value reasoning
+      fallback: [
+        "openai/o4-mini",
+        // Latest o-series
+        "anthropic/claude-opus-4.5",
+        "google/gemini-3-pro-preview"
+      ]
     }
   },
   // Agentic tier configs - models that excel at multi-step autonomous tasks
@@ -1227,13 +1244,23 @@ var DEFAULT_ROUTING_CONFIG = {
     },
     COMPLEX: {
       primary: "anthropic/claude-sonnet-4",
-      fallback: ["anthropic/claude-opus-4.5", "openai/gpt-5.2", "xai/grok-4-0709"]
-      // Opus 4.5 is 3x cheaper than Opus 4
+      fallback: [
+        "anthropic/claude-opus-4.5",
+        // Latest Opus - best agentic
+        "openai/gpt-5.2",
+        "google/gemini-3-pro-preview",
+        "xai/grok-4-0709"
+      ]
     },
     REASONING: {
       primary: "anthropic/claude-sonnet-4",
       // Strong tool use + reasoning for agentic tasks
-      fallback: ["xai/grok-4-fast-reasoning", "moonshot/kimi-k2.5", "deepseek/deepseek-reasoner"]
+      fallback: [
+        "anthropic/claude-opus-4.5",
+        "xai/grok-4-fast-reasoning",
+        "moonshot/kimi-k2.5",
+        "deepseek/deepseek-reasoner"
+      ]
     }
   },
   overrides: {
@@ -1342,8 +1369,7 @@ var MODEL_ALIASES = {
   "grok-code": "xai/grok-code-fast-1",
   // NVIDIA
   nvidia: "nvidia/gpt-oss-120b",
-  "gpt-120b": "nvidia/gpt-oss-120b",
-  "gpt-20b": "nvidia/gpt-oss-20b"
+  "gpt-120b": "nvidia/gpt-oss-120b"
   // Note: auto, free, eco, premium are virtual routing profiles registered in BLOCKRUN_MODELS
   // They don't need aliases since they're already top-level model IDs
 };
@@ -1448,14 +1474,7 @@ var BLOCKRUN_MODELS = [
     contextWindow: 128e3,
     maxOutput: 16384
   },
-  {
-    id: "openai/gpt-4.1-nano",
-    name: "GPT-4.1 Nano",
-    inputPrice: 0.1,
-    outputPrice: 0.4,
-    contextWindow: 128e3,
-    maxOutput: 16384
-  },
+  // gpt-4.1-nano removed - replaced by gpt-5-nano
   {
     id: "openai/gpt-4o",
     name: "GPT-4o",
@@ -1474,25 +1493,7 @@ var BLOCKRUN_MODELS = [
     contextWindow: 128e3,
     maxOutput: 16384
   },
-  // OpenAI O-series (Reasoning)
-  {
-    id: "openai/o1",
-    name: "o1",
-    inputPrice: 15,
-    outputPrice: 60,
-    contextWindow: 2e5,
-    maxOutput: 1e5,
-    reasoning: true
-  },
-  {
-    id: "openai/o1-mini",
-    name: "o1-mini",
-    inputPrice: 1.1,
-    outputPrice: 4.4,
-    contextWindow: 128e3,
-    maxOutput: 65536,
-    reasoning: true
-  },
+  // OpenAI O-series (Reasoning) - o1/o1-mini removed, replaced by o3/o4
   {
     id: "openai/o3",
     name: "o3",
@@ -1629,15 +1630,7 @@ var BLOCKRUN_MODELS = [
     maxOutput: 16384,
     reasoning: true
   },
-  {
-    id: "xai/grok-3-fast",
-    name: "Grok 3 Fast",
-    inputPrice: 5,
-    outputPrice: 25,
-    contextWindow: 131072,
-    maxOutput: 16384,
-    reasoning: true
-  },
+  // grok-3-fast removed - too expensive ($5/$25), use grok-4-fast instead
   {
     id: "xai/grok-3-mini",
     name: "Grok 3 Mini",
@@ -1700,15 +1693,7 @@ var BLOCKRUN_MODELS = [
     maxOutput: 16384,
     reasoning: true
   },
-  {
-    id: "xai/grok-2-vision",
-    name: "Grok 2 Vision",
-    inputPrice: 2,
-    outputPrice: 10,
-    contextWindow: 131072,
-    maxOutput: 16384,
-    vision: true
-  },
+  // grok-2-vision removed - old, 0 transactions
   // NVIDIA - Free/cheap models
   {
     id: "nvidia/gpt-oss-120b",
@@ -1718,19 +1703,11 @@ var BLOCKRUN_MODELS = [
     contextWindow: 128e3,
     maxOutput: 16384
   },
-  {
-    id: "nvidia/gpt-oss-20b",
-    name: "NVIDIA GPT-OSS 20B",
-    inputPrice: 0,
-    outputPrice: 0,
-    contextWindow: 128e3,
-    maxOutput: 16384
-  },
   {
     id: "nvidia/kimi-k2.5",
     name: "NVIDIA Kimi K2.5",
-    inputPrice: 1e-3,
-    outputPrice: 1e-3,
+    inputPrice: 0.55,
+    outputPrice: 2.5,
     contextWindow: 262144,
     maxOutput: 16384
   }
@@ -2372,6 +2349,16 @@ var FREE_MODEL = "nvidia/gpt-oss-120b";
 var HEARTBEAT_INTERVAL_MS = 2e3;
 var DEFAULT_REQUEST_TIMEOUT_MS = 18e4;
 var DEFAULT_PORT = 8402;
+var PROXY_PORT = (() => {
+  const envPort = process.env.BLOCKRUN_PROXY_PORT;
+  if (envPort) {
+    const parsed = parseInt(envPort, 10);
+    if (!isNaN(parsed) && parsed > 0 && parsed < 65536) {
+      return parsed;
+    }
+  }
+  return DEFAULT_PORT;
+})();
 var MAX_FALLBACK_ATTEMPTS = 3;
 var HEALTH_CHECK_TIMEOUT_MS = 2e3;
 var RATE_LIMIT_COOLDOWN_MS = 6e4;
@@ -2471,14 +2458,7 @@ function safeWrite(res, data) {
 }
 var BALANCE_CHECK_BUFFER = 1.5;
 function getProxyPort() {
-  const envPort = process.env.BLOCKRUN_PROXY_PORT;
-  if (envPort) {
-    const parsed = parseInt(envPort, 10);
-    if (!isNaN(parsed) && parsed > 0 && parsed < 65536) {
-      return parsed;
-    }
-  }
-  return DEFAULT_PORT;
+  return PROXY_PORT;
 }
 async function checkExistingProxy(port) {
   const controller = new AbortController();