wolverine-ai 3.7.5 → 3.7.7

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "wolverine-ai",
-  "version": "3.7.5",
+  "version": "3.7.7",
   "description": "Self-healing Node.js server framework powered by AI. Catches crashes, diagnoses errors, generates fixes, verifies, and restarts — automatically.",
   "main": "src/index.js",
   "bin": {

package/src/agent/agent-engine.js

@@ -331,6 +331,7 @@ class AgentEngine {
     this.logger = options.logger;
     this.cwd = options.cwd || process.cwd();
     this.mcp = options.mcp || null; // McpRegistry for external tools
+    this.category = options.category || "tool"; // Analytics category for token tracking
 
     // Budget constraints (claw-code: QueryEngineConfig)
     this.maxTurns = options.maxTurns || 15;
@@ -422,6 +423,7 @@ class AgentEngine {
             messages: this.messages,
             tools: allTools,
             maxTokens: 4096,
+            category: this.category,
           }),
           new Promise((_, reject) => setTimeout(() => reject(new Error(`AI call timed out after ${AI_CALL_TIMEOUT_MS / 1000}s`)), AI_CALL_TIMEOUT_MS)),
         ]);

package/src/agent/sub-agents.js

@@ -97,6 +97,8 @@ async function spawnAgent(type, task, options = {}) {
     });
   }
 
+  // Map sub-agent type to analytics category: triage agents = classifier, fix/db = coding
+  const categoryMap = { explore: "classifier", plan: "classifier", verify: "classifier", research: "research", fix: "coding", database: "coding", security: "audit" };
   const agent = new AgentEngine({
     sandbox: options.sandbox,
     logger: options.logger,
@@ -104,6 +106,7 @@ async function spawnAgent(type, task, options = {}) {
     mcp: options.mcp,
     maxTurns: config.maxTurns,
     maxTokens: config.maxTokens,
+    category: categoryMap[type] || "tool",
   });
 
   // Override the system prompt for this agent type

package/src/brain/embedder.js

@@ -1,4 +1,4 @@
-const { getClient, aiCall, detectProvider } = require("../core/ai-client");
+const { getClient, aiCall, detectProvider, _trackEmbedding } = require("../core/ai-client");
 const { getModel } = require("../core/models");
 
 /**
@@ -45,12 +45,14 @@ async function embed(text) {
   const openai = getClient("openai");
   const model = getModel("embedding");
 
+  const startMs = Date.now();
   const response = await openai.embeddings.create({
     model,
     input: text,
   });
 
   const embedding = response.data[0].embedding;
+  _trackEmbedding(model, response.usage, Date.now() - startMs, true);
   _cacheSet(text, embedding);
   return embedding;
 }
@@ -83,10 +85,12 @@ async function embedBatch(texts) {
   const openai = getClient("openai");
   const model = getModel("embedding");
 
+  const startMs = Date.now();
   const response = await openai.embeddings.create({
     model,
     input: uncached,
   });
+  _trackEmbedding(model, response.usage, Date.now() - startMs, true);
 
   // Sort by index to maintain order
   const sorted = response.data.sort((a, b) => a.index - b.index);

package/src/core/ai-client.js

@@ -32,6 +32,12 @@ function _track(model, category, usage, tool, latencyMs, success) {
   _tracker.record(model, category, input, output, tool, latencyMs, success, cacheCreation, cacheRead);
 }
 
+function _trackEmbedding(model, usage, latencyMs, success) {
+  if (!_tracker) return;
+  const input = usage?.prompt_tokens || usage?.total_tokens || 0;
+  _tracker.record(model, "embedding", input, 0, null, latencyMs, success, 0, 0);
+}
+
 // ── Client Management ──
 
 function getClient(provider) {
@@ -42,13 +48,14 @@ function getClient(provider) {
 
 function _getWolverineClient() {
   if (!_wolverineClient) {
-    // Wolverine inference: direct to GPU or via proxy
-    // WOLVERINE_GPU_KEY = internal key for direct GPU access (llama.cpp --api-key)
-    // WOLVERINE_API_KEY = user key for billed proxy access (api.wolverinenode.xyz)
+    // Wolverine inference: always route through billing proxy when API key is set.
+    // WOLVERINE_API_KEY = billed API key (credits deducted per call)
+    // WOLVERINE_GPU_KEY = direct GPU access (no billing, admin/internal only)
+    // Priority: API_KEY (billed) > GPU_KEY (direct) — billing is the default path
+    const apiKey = process.env.WOLVERINE_API_KEY || process.env.WOLVERINE_GPU_KEY || "none";
     const baseURL = process.env.WOLVERINE_INFERENCE_URL
       ? process.env.WOLVERINE_INFERENCE_URL + "/v1"
       : "https://api.wolverinenode.xyz/v1";
-    const apiKey = process.env.WOLVERINE_GPU_KEY || process.env.WOLVERINE_API_KEY || "none";
     _wolverineClient = new OpenAI({ apiKey, baseURL });
   }
   return _wolverineClient;
@@ -618,4 +625,4 @@ Include both if needed, or just one.`;
   }
 }
 
-module.exports = { requestRepair, getClient, tokenParam, aiCall, aiCallWithHistory, isResponsesModel, isAnthropicModel, setTokenTracker, getTrackerSnapshot, detectProvider };
+module.exports = { requestRepair, getClient, tokenParam, aiCall, aiCallWithHistory, isResponsesModel, isAnthropicModel, setTokenTracker, getTrackerSnapshot, detectProvider, _trackEmbedding };