wolverine-ai 3.5.0 → 3.6.1

package/src/core/runner.js

@@ -590,11 +590,37 @@ class WolverineRunner {
     this._healStatus = { active: true, route: routePath, error: errorDetails?.message?.slice(0, 200), phase: "diagnosing", startedAt: Date.now() };
     this.logger.info("heal.error_monitor", `Healing caught 500 on ${routePath}`, { route: routePath });
 
-    // Build a synthetic stderr from the error details
+    // Build synthetic stderr that matches the error parser's expected format
+    // If IPC didn't include a file, try to resolve from the route path or stack
+    let file = errorDetails.file;
+    let line = errorDetails.line || 1;
+    if (!file && errorDetails.stack) {
+      // Try to find user-land file in stack (not node_modules, not node:)
+      const frames = (errorDetails.stack || "").split("\n");
+      for (const frame of frames) {
+        const m = frame.match(/\(([^)]+):(\d+):(\d+)\)/) || frame.match(/at\s+([^\s(]+):(\d+):(\d+)/);
+        if (m && !m[1].includes("node_modules") && !m[1].includes("node:")) {
+          file = m[1]; line = parseInt(m[2], 10); break;
+        }
+      }
+    }
+    if (!file && routePath) {
+      // Last resort: map route path to likely file (e.g., /breakable → server/routes/breakable.js)
+      const routeName = routePath.split("/").filter(Boolean).pop();
+      if (routeName) {
+        const path = require("path");
+        const guess = path.join(this.cwd, "server", "routes", routeName + ".js");
+        if (require("fs").existsSync(guess)) { file = guess; line = 1; }
+      }
+    }
+
+    const msg = errorDetails.message || "Unknown error";
+    const hasErrorPrefix = /^\w*Error:/.test(msg);
     const stderr = [
-      errorDetails.message || "Unknown error",
+      file ? `${file}:${line}` : "",
+      hasErrorPrefix ? msg : `Error: ${msg}`,
       errorDetails.stack || "",
-      errorDetails.file ? `    at ${errorDetails.file}:${errorDetails.line || 0}` : "",
+      file ? `    at ${file}:${line}:1` : "",
     ].filter(Boolean).join("\n");
 
     try {

package/src/dashboard/server.js

@@ -336,7 +336,7 @@ class DashboardServer {
         systemPrompt: "Route a command. Respond with two words: ROUTE SIZE.\nROUTE: SIMPLE (general knowledge/explanation, no live data needed), TOOLS (needs live server data, file contents, or endpoint calls), AGENT (create/modify/fix code).\nSIZE: SMALL, MEDIUM, LARGE.\nExamples: 'what is wolverine' → SIMPLE SMALL. 'what time is it' → TOOLS SMALL. 'show me index.js' → TOOLS SMALL. 'add endpoint' → AGENT SMALL. 'build auth' → AGENT LARGE.",
         userPrompt: command,
         maxTokens: 10,
-        category: "classify",
+        category: "classifier",
       });
 
       const raw = (result.content || "").trim().toUpperCase();
@@ -424,7 +424,7 @@ ${indexContent}
 Existing route files:
 ${existingRoutes || "(none)"}`,
       maxTokens: 2048,
-      category: "develop",
+      category: "tool",
     });
 
     const raw = (result.content || "").trim().replace(/^```(?:json)?\n?/, "").replace(/\n?```$/, "");

package/src/logger/pricing.js

@@ -53,6 +53,14 @@ const DEFAULT_PRICING = {
   "claude-3-sonnet":      { input: 3.00,  output: 15.00, cache_write: 3.75,  cache_read: 0.30 },
   "claude-3-haiku":       { input: 0.25,  output: 1.25,  cache_write: 0.3125, cache_read: 0.025 },
 
+  // ── Wolverine Self-Hosted (Gemma 4 via api.wolverinenode.xyz) ──
+  // Priced between Anthropic and OpenAI — cheaper than both
+  "wolverine-test-1":     { input: 0.10,  output: 0.40 },
+  "wolverine-gemma-26b":  { input: 0.25,  output: 1.00 },
+  "wolverine-gemma-8b":   { input: 0.10,  output: 0.40 },
+  "wolverine-coding":     { input: 0.10,  output: 0.40 },
+  "wolverine-reasoning":  { input: 0.25,  output: 1.00 },
+
   // ── Fallback ──
   "_default":             { input: 1.00,  output: 4.00 },
 };

package/src/logger/token-tracker.js

@@ -33,6 +33,8 @@ class TokenTracker {
     this._byModel = {};
     // Per-category totals
     this._byCategory = {};
+    // Per-model-per-category cross-reference (model::category → stats)
+    this._byModelCategory = {};
     // Per-tool totals
     this._byTool = {};
     // Timeline: recent entries for charts (in-memory)
@@ -87,7 +89,7 @@ class TokenTracker {
     };
 
     // Accumulate by model
-    if (!this._byModel[model]) this._byModel[model] = { input: 0, output: 0, total: 0, calls: 0, cost: 0, successes: 0, failures: 0, totalLatencyMs: 0, minLatencyMs: Infinity, maxLatencyMs: 0, cacheCreation: 0, cacheRead: 0, cacheSavings: 0 };
+    if (!this._byModel[model]) this._byModel[model] = { input: 0, output: 0, total: 0, calls: 0, cost: 0, successes: 0, failures: 0, totalLatencyMs: 0, totalLatencyTokens: 0, timedCalls: 0, minLatencyMs: Infinity, maxLatencyMs: 0, cacheCreation: 0, cacheRead: 0, cacheSavings: 0 };
     const m = this._byModel[model];
     m.input += entry.input;
     m.output += entry.output;
@@ -100,6 +102,8 @@ class TokenTracker {
     if (entry.success) m.successes++; else m.failures++;
     if (latencyMs > 0) {
       m.totalLatencyMs += latencyMs;
+      m.totalLatencyTokens += total;
+      m.timedCalls++;
       if (latencyMs < m.minLatencyMs) m.minLatencyMs = latencyMs;
       if (latencyMs > m.maxLatencyMs) m.maxLatencyMs = latencyMs;
     }
@@ -112,6 +116,18 @@ class TokenTracker {
     this._byCategory[category].calls++;
     this._byCategory[category].cost += cost.total;
 
+    // Accumulate by model+category cross-reference
+    const mcKey = `${model}::${category}`;
+    if (!this._byModelCategory[mcKey]) this._byModelCategory[mcKey] = { model, category, input: 0, output: 0, total: 0, calls: 0, cost: 0, successes: 0, failures: 0, totalLatencyMs: 0 };
+    const mc = this._byModelCategory[mcKey];
+    mc.input += entry.input;
+    mc.output += entry.output;
+    mc.total += total;
+    mc.calls++;
+    mc.cost += cost.total;
+    if (entry.success) mc.successes++; else mc.failures++;
+    if (latencyMs > 0) mc.totalLatencyMs += latencyMs;
+
     // Accumulate by tool
     if (tool) {
       const toolKey = tool.split(" ")[0];
@@ -158,6 +174,7 @@ class TokenTracker {
       },
       byModel: this._formatModelStats(),
       byCategory: this._byCategory,
+      byModelCategory: this._formatModelCategoryStats(),
       byTool: this._byTool,
       // Recent in-memory timeline
       timeline: this._timeline.slice(-100).map(e => ({
@@ -188,19 +205,42 @@ class TokenTracker {
         cacheCreation: m.cacheCreation || 0,
         cacheRead: m.cacheRead || 0,
         cacheSavings: Math.round((m.cacheSavings || 0) * 1000000) / 1000000,
-        successes: m.successes || m.calls,
+        successes: m.successes != null ? m.successes : m.calls - (m.failures || 0),
         failures: m.failures || 0,
-        successRate: m.calls > 0 ? Math.round(((m.successes || m.calls) / m.calls) * 100) : 0,
-        avgLatencyMs: m.calls > 0 && m.totalLatencyMs ? Math.round(m.totalLatencyMs / m.calls) : 0,
+        successRate: m.calls > 0 ? parseFloat((((m.calls - (m.failures || 0)) / m.calls) * 100).toFixed(2)) : 0,
+        // Latency normalized by token count
+        avgLatencyMs: (m.timedCalls || 0) > 0 ? Math.round(m.totalLatencyMs / m.timedCalls) : 0,
+        msPerKToken: (m.totalLatencyTokens || 0) > 0 ? Math.round((m.totalLatencyMs / m.totalLatencyTokens) * 1000) : 0,
+        tokensPerSecond: m.totalLatencyMs > 0 ? Math.round((m.totalLatencyTokens || m.total) / (m.totalLatencyMs / 1000) * 10) / 10 : 0,
+        outputTokPerSecond: m.totalLatencyMs > 0 && m.output > 0 ? Math.round((m.output / (m.totalLatencyMs / 1000)) * 10) / 10 : 0,
+        timedCalls: m.timedCalls || 0,
         minLatencyMs: m.minLatencyMs === Infinity ? 0 : (m.minLatencyMs || 0),
         maxLatencyMs: m.maxLatencyMs || 0,
-        tokensPerSecond: m.totalLatencyMs > 0 ? Math.round((m.total / (m.totalLatencyMs / 1000)) * 10) / 10 : 0,
         costPerCall: m.calls > 0 ? Math.round((m.cost / m.calls) * 1000000) / 1000000 : 0,
       };
     }
     return result;
   }
 
+  /**
+   * Format model+category cross-reference for analytics.
+   * Returns array of { model, category, calls, cost, tokens, successRate, avgLatencyMs }
+   */
+  _formatModelCategoryStats() {
+    return Object.values(this._byModelCategory).map(mc => ({
+      model: mc.model,
+      category: mc.category,
+      calls: mc.calls,
+      cost: Math.round(mc.cost * 1000000) / 1000000,
+      tokens: mc.total,
+      input: mc.input,
+      output: mc.output,
+      successRate: mc.calls > 0 ? parseFloat((((mc.calls - (mc.failures || 0)) / mc.calls) * 100).toFixed(2)) : 100,
+      avgLatencyMs: mc.calls > 0 && mc.totalLatencyMs > 0 ? Math.round(mc.totalLatencyMs / mc.calls) : 0,
+      tokensPerSecond: mc.totalLatencyMs > 0 ? Math.round((mc.total / (mc.totalLatencyMs / 1000)) * 10) / 10 : 0,
+    }));
+  }
+
   /**
    * Load full history from JSONL file. For dashboard charts across sessions.
    * @param {number} limit — max entries to return (default: 500)
@@ -253,6 +293,7 @@ class TokenTracker {
       lastSaved: Date.now(),
       byModel: this._byModel,
       byCategory: this._byCategory,
+      byModelCategory: this._byModelCategory,
       byTool: this._byTool,
       totalTokens: this._totalTokens,
       totalCalls: this._totalCalls,
@@ -275,6 +316,7 @@ class TokenTracker {
         const data = JSON.parse(fs.readFileSync(this.usagePath, "utf-8"));
         this._byModel = data.byModel || {};
         this._byCategory = data.byCategory || {};
+        this._byModelCategory = data.byModelCategory || {};
         this._byTool = data.byTool || {};
         this._totalTokens = data.totalTokens || 0;
         this._totalCalls = data.totalCalls || 0;

package/src/monitor/perf-monitor.js

@@ -236,7 +236,7 @@ Provide a brief analysis and actionable suggestions. Focus on:
 
 Keep your response under 300 words. Be specific and actionable.`,
         maxTokens: 512,
-        category: "security",
+        category: "audit",
       });
 
       const analysis = result.content;

package/src/notifications/notifier.js

@@ -172,7 +172,7 @@ class Notifier {
       systemPrompt: "You summarize server errors for developers. Write 1-2 short sentences. Be direct and actionable. Do not include any secrets, passwords, or API key values — only refer to them by name (e.g. 'the OPENAI_API_KEY').",
       userPrompt: `Summarize this error for a developer notification:\n\nCategory: ${classification.category}\nError: ${safeError}\n\nStack (first 300 chars): ${safeStack.slice(0, 300)}`,
       maxTokens: 100,
-      category: "security",
+      category: "audit",
     });
 
     // Double-sanitize the AI response (in case the AI echoes something)

package/src/platform/telemetry.js

@@ -66,7 +66,8 @@ function collectHeartbeat(subsystems) {
       totalCalls: tokenTracker?._totalCalls || usage?.session?.totalCalls || 0,
       totalCacheSavings: _sumCacheSavings(usage?.byModel || {}),
       byCategory: usage?.byCategory || {},
-      byModel: usage?.byModel || {},  // includes: latency, successRate, tokensPerSec, cacheSavings per model
+      byModel: usage?.byModel || {},
+      byModelCategory: usage?.byModelCategory || [],
       byTool: usage?.byTool || {},
       byProvider: _aggregateByProvider(usage?.byModel || {}),
     },

package/src/security/injection-detector.js

@@ -95,7 +95,7 @@ Respond with ONLY valid JSON:
     systemPrompt: "You are a security analyst. Respond with ONLY valid JSON.",
     userPrompt,
     maxTokens: 128,
-    category: "security",
+    category: "audit",
   });
 
   const content = result.content;