@zuvia-software-solutions/code-mapper 2.2.0 → 2.2.2

package/dist/core/embeddings/embedding-pipeline.js

@@ -115,18 +115,29 @@ export function fetchGraphContext(db, nodes) {
  * @returns Enriched text
  */
 export function enrichTextWithGraphContext(text, ctx) {
-    const parts = [];
-    if (ctx.module)
-        parts.push(`Module: ${ctx.module}`);
-    if (ctx.callers.length > 0)
-        parts.push(`Called by: ${ctx.callers.join(', ')}`);
-    if (ctx.callees.length > 0)
-        parts.push(`Calls: ${ctx.callees.join(', ')}`);
-    if (parts.length === 0)
+    if (!ctx.module && ctx.callers.length === 0 && ctx.callees.length === 0)
         return text;
     const lines = text.split('\n');
-    const insertIdx = lines.findIndex(l => l === '') || 2;
-    lines.splice(insertIdx, 0, ...parts);
+    // Append Module to the File: line (matches Python batch format)
+    if (ctx.module) {
+        const fileIdx = lines.findIndex(l => l.startsWith('File: '));
+        if (fileIdx >= 0) {
+            lines[fileIdx] += ` | Module: ${ctx.module}`;
+        }
+    }
+    // Insert callers/callees after the File: line
+    const insertParts = [];
+    if (ctx.callers.length > 0)
+        insertParts.push(`Called by: ${ctx.callers.join(', ')}`);
+    if (ctx.callees.length > 0)
+        insertParts.push(`Calls: ${ctx.callees.join(', ')}`);
+    if (insertParts.length > 0) {
+        const fileIdx = lines.findIndex(l => l.startsWith('File: '));
+        const insertAt = fileIdx >= 0 ? fileIdx + 1 : lines.findIndex(l => l === '') || 2;
+        for (let i = insertParts.length - 1; i >= 0; i--) {
+            lines.splice(insertAt, 0, insertParts[i] ?? '');
+        }
+    }
     return lines.join('\n');
 }
 /**

package/dist/core/embeddings/text-generator.js

@@ -10,6 +10,7 @@
  * The graph context enrichment (callers, callees, module) is applied
  * separately by the embedding pipeline — this module handles the per-node text.
  */
+import { expandIdentifier } from '../db/adapter.js';
 /** Extract filename from a file path */
 const getFileName = (filePath) => {
     const parts = filePath.split('/');
@@ -143,8 +144,12 @@ function extractSignature(content, label) {
 export const generateEmbeddingText = (node, _config = {}) => {
     const label = node.label;
     const parts = [];
-    // 1. Type + name
+    // 1. Type + name + expanded name for natural language bridge
     parts.push(`${label}: ${node.name}`);
+    const expanded = expandIdentifier(node.name);
+    if (expanded && expanded !== node.name.toLowerCase()) {
+        parts.push(expanded);
+    }
     // 2. First comment as natural language description
     const comment = extractFirstComment(node.content);
     if (comment)

package/dist/core/embeddings/types.d.ts

@@ -3,6 +3,8 @@ export declare const EMBEDDABLE_LABELS: readonly ["Function", "Class", "Method",
 export type EmbeddableLabel = typeof EMBEDDABLE_LABELS[number];
 /** Check if a label is embeddable */
 export declare const isEmbeddableLabel: (label: string) => label is EmbeddableLabel;
+/** Check if a file path is a test/fixture file (skip embedding, BM25 covers it) */
+export declare const isTestFile: (filePath: string) => boolean;
 /** Embedding pipeline lifecycle phases */
 export type EmbeddingPhase = 'idle' | 'loading-model' | 'embedding' | 'indexing' | 'ready' | 'error';
 /** Progress state emitted during embedding pipeline execution */

package/dist/core/embeddings/types.js

@@ -11,6 +11,10 @@ export const EMBEDDABLE_LABELS = [
 ];
 /** Check if a label is embeddable */
 export const isEmbeddableLabel = (label) => EMBEDDABLE_LABELS.includes(label);
+/** Test file patterns — these are searched via BM25, not semantic embeddings */
+const TEST_PATH_PATTERNS = ['/test/', '/tests/', '/spec/', '/fixtures/', '/__tests__/', '/__mocks__/', '.test.', '.spec.', '_test.', '_spec.'];
+/** Check if a file path is a test/fixture file (skip embedding, BM25 covers it) */
+export const isTestFile = (filePath) => TEST_PATH_PATTERNS.some(p => filePath.includes(p));
 // Jina Code 1.5B MLX — 1.54B params on Apple Silicon Metal
 // Matryoshka truncation to 256 dims (trained at this dim, <1% quality loss vs 1536)
 // Task-specific prefixes: nl2code queries, code passages

package/dist/core/incremental/refresh.js

@@ -428,9 +428,12 @@ export async function refreshEmbeddings(db, dirtyFiles, hasEmbeddings) {
             deleteEmbeddingsByFile(db, entry.relativePath);
         }
         // Step 2: Query new embeddable nodes for modified/created files
+        // Skip test/fixture files — BM25 handles them
+        const { isTestFile } = await import('../embeddings/types.js');
         const embeddableSet = new Set(EMBEDDABLE_LABELS);
         const modifiedPaths = dirtyFiles
             .filter(f => f.changeKind === 'modified' || f.changeKind === 'created')
+            .filter(f => !isTestFile(f.relativePath))
             .map(f => f.relativePath);
         if (modifiedPaths.length === 0)
             return;

package/dist/core/ingestion/workers/worker-pool.js

@@ -6,8 +6,8 @@ import fs from 'node:fs';
 import { fileURLToPath } from 'node:url';
 // Max files per postMessage to keep structured-clone memory bounded
 const SUB_BATCH_SIZE = 1500;
-// Per sub-batch timeout (likely a pathological file if exceeded)
-const SUB_BATCH_TIMEOUT_MS = 30_000;
+// Per sub-batch timeout — large codebases with big files need more time
+const SUB_BATCH_TIMEOUT_MS = 120_000;
 /** Create a pool of worker threads */
 export const createWorkerPool = (workerUrl, poolSize) => {
     // Validate worker script exists before spawning to prevent MODULE_NOT_FOUND crashes

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@zuvia-software-solutions/code-mapper",
-  "version": "2.2.0",
+  "version": "2.2.2",
   "description": "Graph-powered code intelligence for AI agents. Index any codebase, query via MCP or CLI.",
   "author": "Abhigyan Patwari",
   "license": "PolyForm-Noncommercial-1.0.0",