@nathapp/nax 0.54.6 → 0.54.8

package/dist/nax.js

@@ -17982,7 +17982,8 @@ var init_schemas3 = __esm(() => {
   SemanticReviewConfigSchema = exports_external.object({
     modelTier: ModelTierSchema.default("balanced"),
     rules: exports_external.array(exports_external.string()).default([]),
-    timeoutMs: exports_external.number().int().positive().default(600000)
+    timeoutMs: exports_external.number().int().positive().default(600000),
+    excludePatterns: exports_external.array(exports_external.string()).default([":!test/", ":!tests/", ":!*_test.go", ":!*.test.ts", ":!*.spec.ts", ":!**/__tests__/"])
   });
   ReviewConfigSchema = exports_external.object({
     enabled: exports_external.boolean(),
@@ -18278,7 +18279,8 @@ var init_defaults = __esm(() => {
       semantic: {
         modelTier: "balanced",
         rules: [],
-        timeoutMs: 600000
+        timeoutMs: 600000,
+        excludePatterns: [":!test/", ":!tests/", ":!*_test.go", ":!*.test.ts", ":!*.spec.ts", ":!**/__tests__/"]
       }
     },
     plan: {
@@ -19927,14 +19929,13 @@ function extractQuestion(output) {
   const text = output.trim();
   if (!text)
     return null;
-  const sentences = text.split(/(?<=[.!?])\s+/);
-  const questionSentences = sentences.filter((s) => s.trim().endsWith("?"));
-  if (questionSentences.length > 0) {
-    const q = questionSentences[questionSentences.length - 1].trim();
-    if (q.length > 10)
-      return q;
-  }
-  const lower = text.toLowerCase();
+  const lines = text.split(`
+`).filter((l) => l.trim().length > 0);
+  const lastLine = lines.at(-1)?.trim() ?? "";
+  if (lastLine.endsWith("?") && lastLine.length > 10) {
+    return lastLine;
+  }
+  const lower = lastLine.toLowerCase();
   const markers = [
     "please confirm",
     "please specify",
@@ -19946,7 +19947,7 @@ function extractQuestion(output) {
   ];
   for (const marker of markers) {
     if (lower.includes(marker)) {
-      return text.slice(-200).trim();
+      return lastLine;
     }
   }
   return null;
@@ -20133,7 +20134,19 @@ class AcpAgentAdapter {
     };
   }
   async complete(prompt, _options) {
-    const model = _options?.model ?? "default";
+    let model = _options?.model;
+    if (!model && _options?.modelTier && _options?.config?.models) {
+      const tier = _options.modelTier;
+      const { resolveModel: resolveModel2 } = await Promise.resolve().then(() => (init_schema(), exports_schema));
+      const models = _options.config.models;
+      const entry = models[tier] ?? models.balanced;
+      if (entry) {
+        try {
+          model = resolveModel2(entry).model;
+        } catch {}
+      }
+    }
+    model ??= "default";
     const timeoutMs = _options?.timeoutMs ?? 120000;
     const permissionMode = resolvePermissions(_options?.config, "complete").mode;
     const workdir = _options?.workdir;
@@ -22352,7 +22365,7 @@ var package_default;
 var init_package = __esm(() => {
   package_default = {
     name: "@nathapp/nax",
-    version: "0.54.6",
+    version: "0.54.8",
     description: "AI Coding Agent Orchestrator \u2014 loops until done",
     type: "module",
     bin: {
@@ -22429,8 +22442,8 @@ var init_version = __esm(() => {
   NAX_VERSION = package_default.version;
   NAX_COMMIT = (() => {
     try {
-      if (/^[0-9a-f]{6,10}$/.test("955ab31"))
-        return "955ab31";
+      if (/^[0-9a-f]{6,10}$/.test("6f97ec3"))
+        return "6f97ec3";
     } catch {}
     try {
       const result = Bun.spawnSync(["git", "rev-parse", "--short", "HEAD"], {
@@ -22478,7 +22491,7 @@ function buildInteractionBridge(chain, context, timeoutMs = DEFAULT_INTERACTION_
 }
 var QUESTION_PATTERNS, DEFAULT_INTERACTION_TIMEOUT_MS = 120000;
 var init_bridge_builder = __esm(() => {
-  QUESTION_PATTERNS = [/\?/, /\bwhich\b/i, /\bshould i\b/i, /\bunclear\b/i, /\bplease clarify\b/i];
+  QUESTION_PATTERNS = [/\?\s*$/, /\bwhich\b/i, /\bshould i\b/i, /\bunclear\b/i, /\bplease clarify\b/i];
 });
 
 // src/interaction/chain.ts
@@ -24853,9 +24866,13 @@ var init_language_commands = __esm(() => {
 
 // src/review/semantic.ts
 var {spawn: spawn2 } = globalThis.Bun;
-async function collectDiff(workdir, storyGitRef) {
+async function collectDiff(workdir, storyGitRef, excludePatterns) {
+  const cmd = ["git", "diff", "--unified=3", `${storyGitRef}..HEAD`];
+  if (excludePatterns.length > 0) {
+    cmd.push("--", ".", ...excludePatterns);
+  }
   const proc = _semanticDeps.spawn({
-    cmd: ["git", "diff", "--unified=3", `${storyGitRef}..HEAD`],
+    cmd,
     cwd: workdir,
     stdout: "pipe",
     stderr: "pipe"
@@ -24870,26 +24887,44 @@ async function collectDiff(workdir, storyGitRef) {
   }
   return stdout;
 }
-function truncateDiff(diff) {
+async function collectDiffStat(workdir, storyGitRef) {
+  const proc = _semanticDeps.spawn({
+    cmd: ["git", "diff", "--stat", `${storyGitRef}..HEAD`],
+    cwd: workdir,
+    stdout: "pipe",
+    stderr: "pipe"
+  });
+  const [exitCode, stdout] = await Promise.all([
+    proc.exited,
+    new Response(proc.stdout).text(),
+    new Response(proc.stderr).text()
+  ]);
+  return exitCode === 0 ? stdout.trim() : "";
+}
+function truncateDiff(diff, stat) {
   if (diff.length <= DIFF_CAP_BYTES) {
     return diff;
   }
   const truncated = diff.slice(0, DIFF_CAP_BYTES);
-  const fileCount = (truncated.match(/^diff --git/gm) ?? []).length;
-  return `${truncated}
-... (truncated, showing first ${fileCount} files)`;
+  const visibleFiles = (truncated.match(/^diff --git/gm) ?? []).length;
+  const totalFiles = (diff.match(/^diff --git/gm) ?? []).length;
+  const statPreamble = stat ? `## File Summary (all changed files)
+${stat}
+
+## Diff (truncated \u2014 ${visibleFiles}/${totalFiles} files shown)
+` : "";
+  return `${statPreamble}${truncated}
+... (truncated at ${DIFF_CAP_BYTES} bytes, showing ${visibleFiles}/${totalFiles} files)`;
 }
 function buildPrompt(story, semanticConfig, diff) {
   const acList = story.acceptanceCriteria.map((ac, i) => `${i + 1}. ${ac}`).join(`
-`);
-  const defaultRulesText = DEFAULT_RULES.map((r, i) => `${i + 1}. ${r}`).join(`
 `);
   const customRulesSection = semanticConfig.rules.length > 0 ? `
-## Custom Rules
+## Additional Review Rules
 ${semanticConfig.rules.map((r, i) => `${i + 1}. ${r}`).join(`
 `)}
 ` : "";
-  return `You are a code reviewer. Review the following git diff against the story requirements and rules.
+  return `You are a semantic code reviewer. Your job is to verify that the implementation satisfies the story's acceptance criteria (ACs). You are NOT a linter or style checker \u2014 lint, typecheck, and convention checks are handled separately.
 
 ## Story: ${story.title}
 
@@ -24898,27 +24933,29 @@ ${story.description}
 
 ### Acceptance Criteria
 ${acList}
-
-## Review Rules
-
-### Default Rules
-${defaultRulesText}
 ${customRulesSection}
-## Git Diff
+## Git Diff (production code only \u2014 test files excluded)
 
 \`\`\`diff
 ${diff}\`\`\`
 
 ## Instructions
 
-Respond with JSON only. No markdown fences around the JSON response itself.
-Format:
+For each acceptance criterion, verify the diff implements it correctly. Flag issues only when:
+1. An AC is not implemented or partially implemented
+2. The implementation contradicts what the AC specifies
+3. New code has dead paths that will never execute (stubs, noops, unreachable branches)
+4. New code is not wired into callers/exports (written but never used)
+
+Do NOT flag: style issues, naming conventions, import ordering, file length, or anything lint handles.
+
+Respond in JSON format:
 {
   "passed": boolean,
   "findings": [
     {
       "severity": "error" | "warn" | "info",
-      "file": "path/to/file.ts",
+      "file": "path/to/file",
       "line": 42,
       "issue": "description of the issue",
       "suggestion": "how to fix it"
@@ -24926,7 +24963,7 @@ Format:
   ]
 }
 
-If the implementation looks correct, respond with { "passed": true, "findings": [] }.`;
+If all ACs are correctly implemented, respond with { "passed": true, "findings": [] }.`;
 }
 function parseLLMResponse(raw) {
   try {
@@ -24969,7 +25006,7 @@ function toReviewFindings(findings) {
     source: "semantic-review"
   }));
 }
-async function runSemanticReview(workdir, storyGitRef, story, semanticConfig, modelResolver) {
+async function runSemanticReview(workdir, storyGitRef, story, semanticConfig, modelResolver, naxConfig) {
   const startTime = Date.now();
   const logger = getSafeLogger();
   if (!storyGitRef) {
@@ -24982,9 +25019,15 @@ async function runSemanticReview(workdir, storyGitRef, story, semanticConfig, mo
       durationMs: Date.now() - startTime
     };
   }
-  logger?.info("review", "Running semantic check", { storyId: story.id, modelTier: semanticConfig.modelTier });
-  const rawDiff = await collectDiff(workdir, storyGitRef);
-  const diff = truncateDiff(rawDiff);
+  logger?.info("review", "Running semantic check", {
+    storyId: story.id,
+    modelTier: semanticConfig.modelTier,
+    configProvided: !!naxConfig
+  });
+  const rawDiff = await collectDiff(workdir, storyGitRef, semanticConfig.excludePatterns);
+  const needsTruncation = rawDiff.length > DIFF_CAP_BYTES;
+  const stat = needsTruncation ? await collectDiffStat(workdir, storyGitRef) : undefined;
+  const diff = truncateDiff(rawDiff, stat);
   const agent = modelResolver(semanticConfig.modelTier);
   if (!agent) {
     logger?.warn("semantic", "No agent available for semantic review \u2014 skipping", {
@@ -25005,7 +25048,9 @@ async function runSemanticReview(workdir, storyGitRef, story, semanticConfig, mo
     rawResponse = await agent.complete(prompt, {
       sessionName: `nax-semantic-${story.id}`,
       workdir,
-      timeoutMs: semanticConfig.timeoutMs
+      timeoutMs: semanticConfig.timeoutMs,
+      modelTier: semanticConfig.modelTier,
+      config: naxConfig
     });
   } catch (err) {
     logger?.warn("semantic", "LLM call failed \u2014 fail-open", { cause: String(err) });
@@ -25020,6 +25065,20 @@ async function runSemanticReview(workdir, storyGitRef, story, semanticConfig, mo
   }
   const parsed = parseLLMResponse(rawResponse);
   if (!parsed) {
+    const looksLikeFail = /"passed"\s*:\s*false/.test(rawResponse);
+    if (looksLikeFail) {
+      logger?.warn("semantic", "LLM returned truncated JSON with passed:false \u2014 treating as failure", {
+        rawResponse: rawResponse.slice(0, 200)
+      });
+      return {
+        check: "semantic",
+        success: false,
+        command: "",
+        exitCode: 1,
+        output: "semantic review: LLM response truncated but indicated failure (passed:false found in partial response)",
+        durationMs: Date.now() - startTime
+      };
+    }
     logger?.warn("semantic", "LLM returned invalid JSON \u2014 fail-open", { rawResponse: rawResponse.slice(0, 200) });
     return {
       check: "semantic",
@@ -25072,19 +25131,12 @@ ${formatFindings(parsed.findings)}`;
     durationMs
   };
 }
-var _semanticDeps, DIFF_CAP_BYTES = 12288, DEFAULT_RULES;
+var _semanticDeps, DIFF_CAP_BYTES = 51200;
 var init_semantic = __esm(() => {
   init_logger2();
   _semanticDeps = {
     spawn: spawn2
   };
-  DEFAULT_RULES = [
-    "No stubs or noops left in production code paths",
-    "No placeholder values (TODO, FIXME, hardcoded dummy data)",
-    "No unrelated changes outside the story scope",
-    "All new code is properly wired into callers and exports",
-    "No silent error swallowing (catch blocks that discard errors without logging)"
-  ];
 });
 
 // src/review/runner.ts
@@ -25231,7 +25283,7 @@ async function getUncommittedFilesImpl(workdir) {
     return [];
   }
 }
-async function runReview(config2, workdir, executionConfig, qualityCommands, storyId, storyGitRef, story, modelResolver) {
+async function runReview(config2, workdir, executionConfig, qualityCommands, storyId, storyGitRef, story, modelResolver, naxConfig) {
   const startTime = Date.now();
   const logger = getSafeLogger();
   const checks3 = [];
@@ -25281,9 +25333,10 @@ Stage and commit these files before running review.`
       const semanticCfg = config2.semantic ?? {
         modelTier: "balanced",
         rules: [],
-        timeoutMs: 600000
+        timeoutMs: 600000,
+        excludePatterns: [":!test/", ":!tests/", ":!*_test.go", ":!*.test.ts", ":!*.spec.ts", ":!**/__tests__/"]
       };
-      const result2 = await _reviewSemanticDeps.runSemanticReview(workdir, storyGitRef, semanticStory, semanticCfg, modelResolver ?? (() => null));
+      const result2 = await _reviewSemanticDeps.runSemanticReview(workdir, storyGitRef, semanticStory, semanticCfg, modelResolver ?? (() => null), naxConfig);
       checks3.push(result2);
       if (!result2.success && !firstFailure) {
         firstFailure = `${checkName} failed`;
@@ -25365,9 +25418,9 @@ async function getChangedFiles(workdir, baseRef) {
 }
 
 class ReviewOrchestrator {
-  async review(reviewConfig, workdir, executionConfig, plugins, storyGitRef, scopePrefix, qualityCommands, storyId, story, modelResolver) {
+  async review(reviewConfig, workdir, executionConfig, plugins, storyGitRef, scopePrefix, qualityCommands, storyId, story, modelResolver, naxConfig) {
     const logger = getSafeLogger();
-    const builtIn = await runReview(reviewConfig, workdir, executionConfig, qualityCommands, storyId, storyGitRef, story, modelResolver);
+    const builtIn = await runReview(reviewConfig, workdir, executionConfig, qualityCommands, storyId, storyGitRef, story, modelResolver, naxConfig);
     if (!builtIn.success) {
       return { builtIn, success: false, failureReason: builtIn.failureReason, pluginFailed: false };
     }
@@ -25466,7 +25519,7 @@ var init_review = __esm(() => {
         title: ctx.story.title,
         description: ctx.story.description,
         acceptanceCriteria: ctx.story.acceptanceCriteria
-      }, modelResolver);
+      }, modelResolver, ctx.config);
       ctx.reviewResult = result.builtIn;
       if (!result.success) {
         const pluginFindings = result.builtIn.pluginReviewers?.flatMap((pr) => pr.findings ?? []) ?? [];

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@nathapp/nax",
-  "version": "0.54.6",
+  "version": "0.54.8",
   "description": "AI Coding Agent Orchestrator — loops until done",
   "type": "module",
   "bin": {