gitnexus 1.6.5-rc.8 → 1.6.5-rc.9

package/dist/core/embeddings/http-client.js

@@ -30,8 +30,11 @@ const readConfig = () => {
     const rawDims = process.env.GITNEXUS_EMBEDDING_DIMS;
     let dimensions;
     if (rawDims !== undefined) {
+        if (!/^\d+$/.test(rawDims)) {
+            throw new Error(`GITNEXUS_EMBEDDING_DIMS must be a positive integer, got "${rawDims}"`);
+        }
         const parsed = parseInt(rawDims, 10);
-        if (Number.isNaN(parsed) || parsed <= 0) {
+        if (parsed <= 0) {
             throw new Error(`GITNEXUS_EMBEDDING_DIMS must be a positive integer, got "${rawDims}"`);
         }
         dimensions = parsed;
@@ -73,9 +76,22 @@ const safeUrl = (url) => {
  * @param model - Model name for the request body
  * @param apiKey - Bearer token (only used in Authorization header)
  * @param batchIndex - Logical batch number (for error context)
- * @param attempt - Current retry attempt (internal)
+ * @param dimensions - Optional output-vector size. When provided, sent as
+ *   the `dimensions` field in the request body. Endpoints that implement
+ *   Matryoshka truncation (OpenAI text-embedding-3-*, Cohere embed-v3,
+ *   Voyage) return a truncated vector at that size; endpoints that do not
+ *   recognise the field may ignore it or return 400. Leave
+ *   `GITNEXUS_EMBEDDING_DIMS` unset for strict backends that reject
+ *   unknown fields.
  */
-const httpEmbedBatch = async (url, batch, model, apiKey, batchIndex = 0) => {
+const httpEmbedBatch = async (url, batch, model, apiKey, batchIndex = 0, dimensions) => {
+    const requestBody = {
+        input: batch,
+        model,
+    };
+    if (dimensions !== undefined) {
+        requestBody.dimensions = dimensions;
+    }
     let resp;
     try {
         resp = await resilientFetch(url, {
@@ -85,7 +101,7 @@ const httpEmbedBatch = async (url, batch, model, apiKey, batchIndex = 0) => {
                 'Content-Type': 'application/json',
                 Authorization: `Bearer ${apiKey}`,
             },
-            body: JSON.stringify({ input: batch, model }),
+            body: JSON.stringify(requestBody),
         }, {
             breakerKey: HTTP_BREAKER_KEY,
             retry: { maxAttempts: HTTP_MAX_RETRIES + 1, baseDelayMs: HTTP_RETRY_BACKOFF_MS },
@@ -130,7 +146,7 @@ export const httpEmbed = async (texts) => {
     for (let i = 0; i < texts.length; i += HTTP_BATCH_SIZE) {
         const batch = texts.slice(i, i + HTTP_BATCH_SIZE);
         const batchIndex = Math.floor(i / HTTP_BATCH_SIZE);
-        const items = await httpEmbedBatch(url, batch, config.model, config.apiKey, batchIndex);
+        const items = await httpEmbedBatch(url, batch, config.model, config.apiKey, batchIndex, config.dimensions);
         if (items.length !== batch.length) {
             throw new Error(`Embedding endpoint returned ${items.length} vectors for ${batch.length} texts ` +
                 `(${safeUrl(url)}, batch ${batchIndex})`);
@@ -164,7 +180,7 @@ export const httpEmbedQuery = async (text) => {
     if (!config)
         throw new Error('HTTP embedding not configured');
     const url = `${config.baseUrl}/embeddings`;
-    const items = await httpEmbedBatch(url, [text], config.model, config.apiKey);
+    const items = await httpEmbedBatch(url, [text], config.model, config.apiKey, 0, config.dimensions);
     if (!items.length) {
         throw new Error(`Embedding endpoint returned empty response (${safeUrl(url)})`);
     }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "gitnexus",
-  "version": "1.6.5-rc.8",
+  "version": "1.6.5-rc.9",
   "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",