@qatonic_innovations/qaios 0.5.0 → 0.5.2

package/dist/index.js

@@ -4300,6 +4300,8 @@ function tempForTier(tier) {
   }
 }
 var DEFAULT_LLM_TIMEOUT_MS = 24e4;
+var DEFAULT_MAX_OUTPUT_TOKENS = 16384;
+var MAX_OUTPUT_TOKENS_CEILING = 21e3;
 function llmTimeoutMs() {
   const raw = process.env["QAIOS_LLM_TIMEOUT_MS"];
   if (raw === void 0) return DEFAULT_LLM_TIMEOUT_MS;
@@ -4360,6 +4362,7 @@ var SkillRunner = class {
     const maxAttempts = maxRetries + 1;
     let attempt = 0;
     let lastValidationError = null;
+    let currentMaxTokens = DEFAULT_MAX_OUTPUT_TOKENS;
     while (attempt < maxAttempts) {
       attempt++;
       const userPromptBase = skill.userPrompt(effectiveCtx, input);
@@ -4377,9 +4380,9 @@ var SkillRunner = class {
           // but Sonnet 4.6 routinely hits stopReason=max_tokens at
           // that limit — surfacing as a Zod validation error from
           // truncated tool input. Sonnet supports up to 128k output;
-          // 16k is plenty of headroom for v0.1 skill outputs without
-          // burning a runaway loop's worth of cost.
-          maxTokens: 16384,
+          // 16k is the default headroom, escalated on truncation-retry
+          // (see currentMaxTokens) for genuinely large suites.
+          maxTokens: currentMaxTokens,
           temperature: tempForTier(skill.modelTier)
         };
         const signal = buildCallSignal(ctx.cancelSignal);
@@ -4432,14 +4435,21 @@ var SkillRunner = class {
       });
       if (response.stopReason === "max_tokens") {
         if (attempt < maxAttempts) {
+          const escalated = Math.min(currentMaxTokens * 2, MAX_OUTPUT_TOKENS_CEILING);
           ctx.auditLogger.append({
             workflowId: ctx.workflowId,
             phase: skill.phase,
             skillId: skill.id,
             event: "skill.retry",
-            payload: { attempt, reason: "output_truncated_max_tokens" },
+            payload: {
+              attempt,
+              reason: "output_truncated_max_tokens",
+              maxTokensWas: currentMaxTokens,
+              maxTokensNext: escalated
+            },
             actor: "agent"
           });
+          currentMaxTokens = escalated;
           continue;
         }
         ctx.auditLogger.append({

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@qatonic_innovations/qaios",
-  "version": "0.5.0",
+  "version": "0.5.2",
   "type": "module",
   "description": "AI QA engineer in your terminal — designs, writes, runs, heals, and explores tests for web UI and APIs with audit-grade traceability.",
   "license": "MIT",