voyageai-cli 1.33.3 → 1.33.5

package/README.md

@@ -1,7 +1,7 @@
 # voyageai-cli
 
 <p align="center">
-  <img src="https://raw.githubusercontent.com/mrlynn/voyageai-cli/main/demo-readme.gif" alt="voyageai-cli demo" width="800" />
+  <img src="https://raw.githubusercontent.com/mrlynn/voyageai-cli/main/cli-quickstart.gif" alt="voyageai-cli demo" width="800" />
 </p>
 
 [![CI](https://github.com/mrlynn/voyageai-cli/actions/workflows/ci.yml/badge.svg)](https://github.com/mrlynn/voyageai-cli/actions/workflows/ci.yml) [![npm version](https://img.shields.io/npm/v/voyageai-cli.svg)](https://www.npmjs.com/package/voyageai-cli) [![License: MIT](https://img.shields.io/badge/License-MIT-blue.svg)](https://opensource.org/licenses/MIT) [![Node.js](https://img.shields.io/node/v/voyageai-cli.svg)](https://nodejs.org) [![GitHub release](https://img.shields.io/github/v/release/mrlynn/voyageai-cli?label=Desktop%20App)](https://github.com/mrlynn/voyageai-cli/releases)

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "voyageai-cli",
-  "version": "1.33.3",
+  "version": "1.33.5",
   "description": "CLI for Voyage AI embeddings, reranking, and MongoDB Atlas Vector Search",
   "_comment": "This package contains the CLI + web playground. The electron/ directory is excluded via .npmignore and distributed via GitHub Releases.",
   "bin": {
@@ -38,6 +38,9 @@
     "test": "node --test test/**/*.test.js",
     "prepare": "git config core.hooksPath scripts/hooks 2>/dev/null || true",
     "reset": "node scripts/reset.js",
+    "demos:list": "node scripts/run-demos.js list",
+    "demos:run": "node scripts/run-demos.js",
+    "demos:run:all": "node scripts/run-demos.js all",
     "version": "node scripts/sync-nano-version.js && git add src/nano/nano-bridge.py",
     "release": "./scripts/release.sh",
     "app:install": "cd electron && npm install",

package/src/commands/ingest.js

@@ -275,7 +275,7 @@ function registerIngest(program) {
             batches: totalBatches,
             batchSize: opts.batchSize,
             estimatedTokens: estimated,
-            model: opts.model,
+            model: ingestModel,
             textField: textKey,
           }, null, 2));
         } else {
@@ -285,7 +285,7 @@ function registerIngest(program) {
           console.log(ui.label('Documents', String(documents.length)));
           console.log(ui.label('Batches', `${totalBatches} (batch size: ${opts.batchSize})`));
           console.log(ui.label('Est. tokens', `~${estimated.toLocaleString()}`));
-          console.log(ui.label('Model', opts.model));
+          console.log(ui.label('Model', ingestModel));
           console.log(ui.label('Text field', textKey));
           console.log(ui.label('Target', `${opts.db}.${opts.collection}`));
           console.log(ui.label('Embed field', opts.field));
@@ -404,7 +404,7 @@ function registerIngest(program) {
             collection: opts.collection,
             batches: totalBatches,
             tokens: totalTokens,
-            model: opts.model,
+            model: ingestModel,
             durationSeconds: parseFloat(duration),
             docsPerSecond: parseFloat(rate),
           };
@@ -420,7 +420,7 @@ function registerIngest(program) {
           }
           console.log(ui.label('Batches', String(totalBatches)));
           console.log(ui.label('Tokens', totalTokens.toLocaleString()));
-          console.log(ui.label('Model', opts.model));
+          console.log(ui.label('Model', ingestModel));
           console.log(ui.label('Duration', `${duration}s`));
           console.log(ui.label('Rate', `${rate} docs/sec`));
           if (errors.length > 0) {

package/src/commands/init.js

@@ -128,6 +128,24 @@ function registerInit(program) {
       const filePath = saveProject(config);
       const relPath = path.relative(process.cwd(), filePath);
 
+      // Warn if nano selected but bridge is not set up
+      if (config.model === 'voyage-4-nano') {
+        try {
+          const { checkVenv, checkModel } = require('../nano/nano-health');
+          const venv = checkVenv();
+          const model = checkModel();
+          if (!venv.ok || !model.ok) {
+            console.log('');
+            console.log(pc.yellow('⚠ voyage-4-nano requires local setup before use.'));
+            if (!venv.ok) console.log(pc.dim('  • Python venv: not found'));
+            if (!model.ok) console.log(pc.dim('  • Model weights: not downloaded'));
+            console.log(pc.yellow('  Run: vai nano setup'));
+          }
+        } catch (_) {
+          // nano module not available — skip check
+        }
+      }
+
       if (!opts.quiet) {
         console.log('');
         console.log(ui.dim('  Next steps:'));

package/src/commands/playground.js

@@ -316,7 +316,10 @@ function createPlaygroundServer() {
 
       // Handle RAG API requests
       if (req.url.startsWith('/api/rag/')) {
-        const handled = await handleRAGRequest(req, res, { generateEmbeddings });
+        const handled = await handleRAGRequest(req, res, {
+          generateEmbeddings,
+          generateLocalEmbeddings: require('../nano/nano-local.js').generateLocalEmbeddings,
+        });
         if (handled) return;
       }
 

package/src/commands/purge.js

@@ -5,7 +5,7 @@ const path = require('path');
 let p;
 function clack() { if (!p) p = require('@clack/prompts'); return p; }
 const { loadProject } = require('../lib/project');
-const { connect, close } = require('../lib/mongo');
+const { getMongoCollection } = require('../lib/mongo');
 const ui = require('../lib/ui');
 
 /**
@@ -125,8 +125,9 @@ async function purge(options = {}) {
     if (!quiet) {
       p.log.step(`Connecting to database: ${db}`);
     }
-    client = await connect(db);
-    const collection = client.db(db).collection(collectionName);
+    const mongo = await getMongoCollection(db, collectionName);
+    client = mongo.client;
+    const collection = mongo.collection;
 
     let filter = {};
     let staleIds = [];
@@ -223,8 +224,8 @@ async function purge(options = {}) {
       p.log.step('Deleting documents...');
     }
     
-    const result = await collection.deleteMany(filter);
-    const deleted = result.deletedCount;
+    const deleteResult = await collection.deleteMany(filter);
+    const deleted = deleteResult.deletedCount;
 
     if (options.json) {
       console.log(JSON.stringify({ success: true, deleted }));
@@ -244,7 +245,7 @@ async function purge(options = {}) {
     return { success: false, error: err.message };
   } finally {
     if (client) {
-      await close();
+      await client.close();
     }
   }
 }

package/src/commands/refresh.js

@@ -3,7 +3,7 @@
 let p;
 function clack() { if (!p) p = require('@clack/prompts'); return p; }
 const { loadProject, saveProject } = require('../lib/project');
-const { connect, close } = require('../lib/mongo');
+const { getMongoCollection } = require('../lib/mongo');
 const { generateEmbeddings } = require('../lib/api');
 const { chunkText } = require('../lib/chunker');
 const ui = require('../lib/ui');
@@ -76,8 +76,9 @@ async function refresh(options = {}) {
     if (!quiet) {
       p.log.step(`Connecting to database: ${db}`);
     }
-    client = await connect(db);
-    const collection = client.db(db).collection(collectionName);
+    const result = await getMongoCollection(db, collectionName);
+    client = result.client;
+    const collection = result.collection;
 
     // Build filter
     let filter = {};
@@ -291,7 +292,7 @@ async function refresh(options = {}) {
     return { success: false, error: err.message };
   } finally {
     if (client) {
-      await close();
+      await client.close();
     }
   }
 }

package/src/lib/catalog.js

@@ -36,7 +36,7 @@ const MODEL_CATALOG = [
   { name: 'voyage-multimodal-3.5', type: 'embedding-multimodal', context: '32K', dimensions: '1024 (default), 256, 512, 2048', price: '$0.12/M + $0.60/B px', bestFor: 'Text + images + video', shortFor: 'Multimodal', multimodal: true },
   { name: 'rerank-2.5', type: 'reranking', context: '32K', dimensions: '—', price: '$0.05/1M tokens', pricePerMToken: 0.05, bestFor: 'Best quality reranking', shortFor: 'Best reranker' },
   { name: 'rerank-2.5-lite', type: 'reranking', context: '32K', dimensions: '—', price: '$0.02/1M tokens', pricePerMToken: 0.02, bestFor: 'Fast reranking', shortFor: 'Fast reranker' },
-  { name: 'voyage-4-nano', type: 'embedding', context: '32K', dimensions: '512 (default), 128, 256', price: 'Open-weight (free)', pricePerMToken: 0, bestFor: 'Open-weight / edge / local', shortFor: 'Open / edge', local: true, unreleased: true, family: 'voyage-4', architecture: 'dense', sharedSpace: 'voyage-4', huggingface: 'https://huggingface.co/voyageai/voyage-4-nano', rtebScore: null },
+  { name: 'voyage-4-nano', type: 'embedding', context: '32K', dimensions: '512 (default), 128, 256, 1024, 2048', price: 'Open-weight (free)', pricePerMToken: 0, bestFor: 'Open-weight / edge / local', shortFor: 'Local / free', local: true, unreleased: true, family: 'voyage-4', architecture: 'dense', sharedSpace: 'voyage-4', huggingface: 'https://huggingface.co/voyageai/voyage-4-nano', rtebScore: null },
   // Legacy models
   { name: 'voyage-3-large', type: 'embedding', context: '32K', dimensions: '1024 (default), 256, 512, 2048', price: '$0.18/1M tokens', pricePerMToken: 0.18, bestFor: 'Previous gen quality', shortFor: 'Previous gen quality', legacy: true, rtebScore: null },
   { name: 'voyage-3', type: 'embedding', context: '32K', dimensions: '1024', price: '$0.06/1M tokens', pricePerMToken: 0.06, bestFor: 'Previous gen balanced', shortFor: 'Previous gen balanced', legacy: true, rtebScore: null },

package/src/lib/playground-rag-api.js

@@ -25,6 +25,34 @@ async function extractTextFromPDF(buffer) {
 const RAG_DB = 'vai_rag';
 const KBS_COLLECTION = 'knowledge_bases';
 
+async function computeKBStatsFromCollection(docsCollection) {
+  const stats = await docsCollection.aggregate([
+    { $group: {
+      _id: null,
+      totalSize: { $sum: { $strLenBytes: { $ifNull: ['$content', ''] } } },
+      chunkCount: { $sum: 1 },
+      files: { $addToSet: '$fileName' }
+    } }
+  ]).toArray();
+
+  const liveStats = stats[0] || { totalSize: 0, chunkCount: 0, files: [] };
+  return {
+    size: liveStats.totalSize,
+    chunkCount: liveStats.chunkCount,
+    docCount: liveStats.files.filter(Boolean).length
+  };
+}
+
+async function computeKBStats(db, kbName) {
+  return computeKBStatsFromCollection(db.collection(`kb_${kbName}_docs`));
+}
+
+function normalizeChunks(content) {
+  return chunkText(content)
+    .map(chunk => typeof chunk === 'string' ? chunk.trim() : '')
+    .filter(Boolean);
+}
+
 // ── Friendly KB name generator ──
 const KB_ADJECTIVES = [
   'swift', 'bright', 'calm', 'bold', 'keen',
@@ -76,26 +104,86 @@ function generateKBName() {
   return `${adj}-${noun}-${suffix}`;
 }
 
+/**
+ * Resolve the correct embedding function based on the selected model.
+ * When embeddingModel is 'voyage-4-nano', uses local nano embeddings.
+ * Otherwise, uses the remote Voyage API.
+ *
+ * @param {string} embeddingModel - Selected embedding model name
+ * @param {Function} remoteEmbed - Remote generateEmbeddings function
+ * @param {Function} localEmbed - Local generateLocalEmbeddings function
+ * @returns {{ embedFn: Function, model: string, isLocal: boolean }}
+ */
+function resolveEmbedFn(embeddingModel, remoteEmbed, localEmbed) {
+  if (embeddingModel === 'voyage-4-nano' && localEmbed) {
+    return {
+      embedFn: (texts, opts) => localEmbed(texts, {
+        inputType: opts.inputType || 'document',
+        dimensions: 1024,
+      }),
+      model: 'voyage-4-nano',
+      isLocal: true,
+    };
+  }
+  return {
+    embedFn: (texts, opts) => remoteEmbed(texts, {
+      model: embeddingModel || 'voyage-4-large',
+      inputType: opts.inputType || 'document',
+    }),
+    model: embeddingModel || 'voyage-4-large',
+    isLocal: false,
+  };
+}
+
 /**
  * Handle RAG API requests
  * Returns true if handled, false otherwise
  * @param {http.IncomingMessage} req
  * @param {http.ServerResponse} res
- * @param {Object} context - API context (generateEmbeddings, etc.)
+ * @param {Object} context - API context (generateEmbeddings, generateLocalEmbeddings)
  */
 async function handleRAGRequest(req, res, context) {
-  const { generateEmbeddings } = context;
+  const { generateEmbeddings, generateLocalEmbeddings } = context;
 
   // GET /api/rag/kbs - List all knowledge bases
   if (req.method === 'GET' && req.url === '/api/rag/kbs') {
     try {
       const { client, collection: kbsCollection } = await getMongoCollection(RAG_DB, KBS_COLLECTION);
+      const db = client.db(RAG_DB);
       const kbs = await kbsCollection.find({}).toArray();
+      const metadataFixes = [];
+      const hydratedKbs = await Promise.all(kbs.map(async (kb) => {
+        const liveStats = await computeKBStats(db, kb.name);
+        if (
+          (kb.docCount || 0) !== liveStats.docCount ||
+          (kb.chunkCount || 0) !== liveStats.chunkCount ||
+          (kb.size || 0) !== liveStats.size
+        ) {
+          metadataFixes.push({
+            updateOne: {
+              filter: { _id: kb._id },
+              update: {
+                $set: {
+                  docCount: liveStats.docCount,
+                  chunkCount: liveStats.chunkCount,
+                  size: liveStats.size
+                }
+              }
+            }
+          });
+        }
+        return { ...kb, ...liveStats };
+      }));
+
+      if (metadataFixes.length > 0) {
+        await kbsCollection.bulkWrite(metadataFixes, { ordered: false });
+      }
+
       client.close();
 
       res.writeHead(200, { 'Content-Type': 'application/json' });
       res.end(JSON.stringify({
-        kbs: kbs.map(kb => ({
+        kbs: hydratedKbs.map(kb => ({
           name: kb.name,
           displayName: kb.displayName || kb.name,
           docCount: kb.docCount || 0,
@@ -223,6 +311,7 @@ async function handleRAGRequest(req, res, context) {
           const headerSep = Buffer.from('\r\n\r\n');
           const crlf = Buffer.from('\r\n');
           let kbName = null;
+          let embeddingModel = null;
 
           // Find all boundary positions in the raw Buffer
           let searchStart = 0;
@@ -267,6 +356,8 @@ async function handleRAGRequest(req, res, context) {
               files.push({ name: filename, path: filepath });
             } else if (nameMatch && nameMatch[1] === 'kbName') {
               kbName = body.slice(contentStart, contentEnd).toString('utf8').trim();
+            } else if (nameMatch && nameMatch[1] === 'embeddingModel') {
+              embeddingModel = body.slice(contentStart, contentEnd).toString('utf8').trim();
             }
           }
 
@@ -323,6 +414,9 @@ async function handleRAGRequest(req, res, context) {
             }
           }
 
+          // Resolve embedding function (local nano vs remote API)
+          const { embedFn } = resolveEmbedFn(embeddingModel, generateEmbeddings, generateLocalEmbeddings);
+
           // Ingest files
           res.writeHead(200, {
             'Content-Type': 'application/x-ndjson',
@@ -355,10 +449,10 @@ async function handleRAGRequest(req, res, context) {
             } else {
               content = fs.readFileSync(file.path, 'utf8');
             }
-            totalSize += Buffer.byteLength(content, 'utf8');
+            const contentSize = Buffer.byteLength(content, 'utf8');
 
             // Stage: chunking
-            const chunks = chunkText(content);
+            const chunks = normalizeChunks(content);
             res.write(JSON.stringify({
               type: 'progress',
               stage: 'chunking',
@@ -368,7 +462,18 @@ async function handleRAGRequest(req, res, context) {
               fileCount: files.length
             }) + '\n');
 
+            if (chunks.length === 0) {
+              res.write(JSON.stringify({
+                type: 'warning',
+                file: file.name,
+                warning: `No text content could be extracted from ${file.name}.`
+              }) + '\n');
+              continue;
+            }
+
             // Stage: embedding (per-chunk progress)
+            let persistedChunks = 0;
+            let lastEmbedError = null;
             for (let c = 0; c < chunks.length; c++) {
               try {
                 res.write(JSON.stringify({
@@ -381,7 +486,7 @@ async function handleRAGRequest(req, res, context) {
                   fileCount: files.length
                 }) + '\n');
 
-                const embedding = await generateEmbeddings(chunks[c], 'voyage-4-large');
+                const embedding = await embedFn([chunks[c]], { inputType: 'document' });
                 const doc = {
                   _id: crypto.randomUUID(),
                   kbName,
@@ -391,8 +496,10 @@ async function handleRAGRequest(req, res, context) {
                   createdAt: new Date()
                 };
                 await docsCollection.insertOne(doc);
+                persistedChunks++;
                 totalChunks++;
               } catch (embedErr) {
+                lastEmbedError = embedErr;
                 console.warn(`Failed to embed chunk from ${file.name}:`, embedErr.message);
               }
             }
@@ -406,7 +513,26 @@ async function handleRAGRequest(req, res, context) {
               fileCount: files.length
             }) + '\n');
 
-            totalDocs++;
+            if (persistedChunks > 0) {
+              totalDocs++;
+              totalSize += contentSize;
+            } else {
+              const detail = lastEmbedError?.message ? ` ${lastEmbedError.message}` : '';
+              res.write(JSON.stringify({
+                type: 'warning',
+                file: file.name,
+                warning: `No chunks were stored for ${file.name}.${detail}`.trim()
+              }) + '\n');
+            }
+
+            if (persistedChunks > 0 && persistedChunks < chunks.length) {
+              res.write(JSON.stringify({
+                type: 'warning',
+                file: file.name,
+                warning: `Only ${persistedChunks}/${chunks.length} chunks were stored for ${file.name}.`
+              }) + '\n');
+            }
+
             try {
               fs.unlinkSync(file.path);
             } catch (e) {
@@ -414,18 +540,13 @@ async function handleRAGRequest(req, res, context) {
             }
           }
 
-          // Update KB metadata (use $inc so size accumulates across uploads)
+          // Recompute live stats so counters stay accurate even when some files
+          // produce zero persisted chunks or partial embeddings succeed.
+          const liveStats = await computeKBStatsFromCollection(docsCollection);
           await kbsCollection.updateOne(
             { name: kbName },
             {
-              $inc: {
-                docCount: totalDocs,
-                chunkCount: totalChunks,
-                size: totalSize
-              },
-              $set: {
-                updatedAt: new Date()
-              }
+              $set: { ...liveStats, updatedAt: new Date() }
             }
           );
 
@@ -510,20 +631,8 @@ async function handleRAGRequest(req, res, context) {
       }
 
       // Compute live stats from docs collection (more accurate than stored metadata)
-      const { collection: docsCollection } = await getMongoCollection(RAG_DB, `kb_${kbName}_docs`);
-      const stats = await docsCollection.aggregate([
-        { $group: {
-          _id: null,
-          totalSize: { $sum: { $strLenBytes: { $ifNull: ['$content', ''] } } },
-          chunkCount: { $sum: 1 },
-          files: { $addToSet: '$fileName' }
-        }}
-      ]).toArray();
-
-      const liveStats = stats[0] || { totalSize: 0, chunkCount: 0, files: [] };
-      kb.size = liveStats.totalSize;
-      kb.chunkCount = liveStats.chunkCount;
-      kb.docCount = liveStats.files.length;
+      const db = client.db(RAG_DB);
+      Object.assign(kb, await computeKBStats(db, kbName));
 
       client.close();
       res.writeHead(200, { 'Content-Type': 'application/json' });
@@ -645,11 +754,10 @@ async function handleRAGRequest(req, res, context) {
 
       await docsCollection.deleteOne({ _id: docId });
 
-      // Update KB doc count
-      const docCount = await docsCollection.countDocuments();
+      const liveStats = await computeKBStatsFromCollection(docsCollection);
       await kbsCollection.updateOne(
         { name: kbName },
-        { $set: { chunkCount: docCount, updatedAt: new Date() } }
+        { $set: { ...liveStats, updatedAt: new Date() } }
       );
 
       kbClient.close();
@@ -670,7 +778,7 @@ async function handleRAGRequest(req, res, context) {
     req.on('data', chunk => { body += chunk; });
     req.on('end', async () => {
       try {
-        const { text, kbName, title } = JSON.parse(body);
+        const { text, kbName, title, embeddingModel } = JSON.parse(body);
 
         if (!text || typeof text !== 'string' || !text.trim()) {
           res.writeHead(400, { 'Content-Type': 'application/json' });
@@ -700,17 +808,28 @@ async function handleRAGRequest(req, res, context) {
           'Cache-Control': 'no-cache'
         });
 
-        const chunks = chunkText(text.trim());
+        const chunks = normalizeChunks(text.trim());
         const fileName = title && title.trim() ? title.trim().slice(0, 80) : `pasted-text-${Date.now()}`;
         const totalSize = Buffer.byteLength(text, 'utf8');
 
+        // Resolve embedding function (local nano vs remote API)
+        const { embedFn } = resolveEmbedFn(embeddingModel, generateEmbeddings, generateLocalEmbeddings);
+
         res.write(JSON.stringify({ type: 'progress', stage: 'chunking', current: chunks.length, total: chunks.length }) + '\n');
 
+        if (chunks.length === 0) {
+          res.write(JSON.stringify({ type: 'error', error: 'No text content could be chunked from the pasted text.' }) + '\n');
+          res.end();
+          kbClient.close();
+          return;
+        }
+
         let totalChunks = 0;
+        let lastEmbedError = null;
         for (let i = 0; i < chunks.length; i++) {
           res.write(JSON.stringify({ type: 'progress', stage: 'embedding', current: i + 1, total: chunks.length }) + '\n');
           try {
-            const embedding = await generateEmbeddings(chunks[i], 'voyage-4-large');
+            const embedding = await embedFn([chunks[i]], { inputType: 'document' });
             const doc = {
               _id: crypto.randomUUID(),
               kbName,
@@ -722,15 +841,24 @@ async function handleRAGRequest(req, res, context) {
             await docsCollection.insertOne(doc);
             totalChunks++;
           } catch (embedErr) {
+            lastEmbedError = embedErr;
             console.warn(`Failed to embed chunk from pasted text:`, embedErr.message);
           }
         }
 
+        if (totalChunks === 0) {
+          const detail = lastEmbedError?.message ? ` ${lastEmbedError.message}` : '';
+          res.write(JSON.stringify({ type: 'error', error: `No chunks were stored for the pasted text.${detail}`.trim() }) + '\n');
+          res.end();
+          kbClient.close();
+          return;
+        }
+
+        const liveStats = await computeKBStatsFromCollection(docsCollection);
         await kbsCollection.updateOne(
           { name: kbName },
           {
-            $inc: { docCount: 1, chunkCount: totalChunks, size: totalSize },
-            $set: { updatedAt: new Date() }
+            $set: { ...liveStats, updatedAt: new Date() }
           }
         );
 
@@ -752,7 +880,7 @@ async function handleRAGRequest(req, res, context) {
     req.on('data', chunk => { body += chunk; });
     req.on('end', async () => {
       try {
-        const { url, kbName } = JSON.parse(body);
+        const { url, kbName, embeddingModel } = JSON.parse(body);
 
         if (!url || typeof url !== 'string' || !/^https?:\/\//i.test(url)) {
           res.writeHead(400, { 'Content-Type': 'application/json' });
@@ -826,20 +954,31 @@ async function handleRAGRequest(req, res, context) {
           return;
         }
 
-        const chunks = chunkText(content);
+        const chunks = normalizeChunks(content);
         // Build fileName from URL hostname + path, truncated to 80 chars
         let parsedUrl;
         try { parsedUrl = new URL(url); } catch { parsedUrl = { hostname: 'unknown', pathname: '' }; }
         const fileName = (parsedUrl.hostname + parsedUrl.pathname).slice(0, 80);
         const totalSize = Buffer.byteLength(content, 'utf8');
 
+        // Resolve embedding function (local nano vs remote API)
+        const { embedFn } = resolveEmbedFn(embeddingModel, generateEmbeddings, generateLocalEmbeddings);
+
         res.write(JSON.stringify({ type: 'progress', stage: 'chunking', current: chunks.length, total: chunks.length }) + '\n');
 
+        if (chunks.length === 0) {
+          res.write(JSON.stringify({ type: 'error', error: 'No text content could be chunked from the fetched URL.' }) + '\n');
+          res.end();
+          kbClient.close();
+          return;
+        }
+
         let totalChunks = 0;
+        let lastEmbedError = null;
         for (let i = 0; i < chunks.length; i++) {
           res.write(JSON.stringify({ type: 'progress', stage: 'embedding', current: i + 1, total: chunks.length }) + '\n');
           try {
-            const embedding = await generateEmbeddings(chunks[i], 'voyage-4-large');
+            const embedding = await embedFn([chunks[i]], { inputType: 'document' });
             const doc = {
               _id: crypto.randomUUID(),
               kbName,
@@ -851,15 +990,24 @@ async function handleRAGRequest(req, res, context) {
             await docsCollection.insertOne(doc);
             totalChunks++;
           } catch (embedErr) {
+            lastEmbedError = embedErr;
             console.warn(`Failed to embed chunk from URL ${url}:`, embedErr.message);
           }
         }
 
+        if (totalChunks === 0) {
+          const detail = lastEmbedError?.message ? ` ${lastEmbedError.message}` : '';
+          res.write(JSON.stringify({ type: 'error', error: `No chunks were stored for the fetched URL.${detail}`.trim() }) + '\n');
+          res.end();
+          kbClient.close();
+          return;
+        }
+
+        const liveStats = await computeKBStatsFromCollection(docsCollection);
         await kbsCollection.updateOne(
           { name: kbName },
           {
-            $inc: { docCount: 1, chunkCount: totalChunks, size: totalSize },
-            $set: { updatedAt: new Date() }
+            $set: { ...liveStats, updatedAt: new Date() }
           }
         );
 

package/src/nano/nano-bridge.py

@@ -14,7 +14,7 @@ Really wrestled with this... if you have better ideas, please let me know.
 import json
 import sys
 
-BRIDGE_VERSION = "1.33.3"
+BRIDGE_VERSION = "1.33.5"
 MODEL_NAME = "voyageai/voyage-4-nano"
 
 # Lazy-loaded on first embed request

package/src/playground/index.html

@@ -159,192 +159,6 @@ body {
 .update-progress-fill { height: 100%; background: var(--accent); border-radius: 3px; transition: width 0.3s ease; width: 0%; }
 .update-progress-label { font-size: 12px; font-family: var(--mono); color: var(--accent); min-width: 40px; text-align: right; }
 
-/* ── Onboarding Walkthrough ── */
-.onboarding-overlay {
-  display: none;
-  position: fixed;
-  inset: 0;
-  z-index: 10000;
-}
-.onboarding-overlay.active { display: block; }
-.onboarding-backdrop {
-  position: absolute;
-  inset: 0;
-  background: rgba(0,0,0,0.6);
-  transition: opacity 0.3s;
-}
-.onboarding-spotlight {
-  position: absolute;
-  border-radius: var(--radius);
-  box-shadow: 0 0 0 9999px rgba(0,0,0,0.55);
-  transition: all 0.4s cubic-bezier(0.4,0,0.2,1);
-  z-index: 10001;
-  pointer-events: none;
-}
-.onboarding-tooltip {
-  position: absolute;
-  z-index: 10002;
-  background: var(--bg-card);
-  border: 1px solid var(--accent);
-  border-radius: 12px;
-  padding: 24px 28px 20px;
-  width: 380px;
-  max-width: 90vw;
-  box-shadow: 0 12px 40px rgba(0,0,0,0.4), 0 0 0 1px rgba(0,212,170,0.15);
-  transition: all 0.4s cubic-bezier(0.4,0,0.2,1);
-  opacity: 0;
-  transform: translateY(10px);
-}
-.onboarding-tooltip.visible {
-  opacity: 1;
-  transform: translateY(0);
-}
-.onboarding-tooltip-arrow {
-  position: absolute;
-  width: 12px;
-  height: 12px;
-  background: var(--bg-card);
-  border: 1px solid var(--accent);
-  transform: rotate(45deg);
-}
-.onboarding-tooltip-arrow.top {
-  top: -7px;
-  left: 30px;
-  border-right: none;
-  border-bottom: none;
-}
-.onboarding-tooltip-arrow.bottom {
-  bottom: -7px;
-  left: 30px;
-  border-left: none;
-  border-top: none;
-}
-.onboarding-tooltip-arrow.left {
-  left: -7px;
-  top: 24px;
-  border-right: none;
-  border-top: none;
-}
-.onboarding-step-icon {
-  font-size: 28px;
-  margin-bottom: 8px;
-}
-.onboarding-step-title {
-  font-size: 16px;
-  font-weight: 700;
-  color: var(--accent-text);
-  margin-bottom: 6px;
-}
-.onboarding-step-body {
-  font-size: 13px;
-  color: var(--text-dim);
-  line-height: 1.6;
-  margin-bottom: 18px;
-}
-.onboarding-step-body strong { color: var(--accent); }
-.onboarding-footer {
-  display: flex;
-  align-items: center;
-  justify-content: space-between;
-  gap: 12px;
-}
-.onboarding-dots {
-  display: flex;
-  gap: 6px;
-}
-.onboarding-dot {
-  width: 8px;
-  height: 8px;
-  border-radius: 50%;
-  background: var(--border);
-  transition: background 0.2s;
-}
-.onboarding-dot.active { background: var(--accent); }
-.onboarding-dot.completed { background: var(--accent-dim); }
-.onboarding-actions {
-  display: flex;
-  gap: 8px;
-  align-items: center;
-}
-.onboarding-skip {
-  background: none;
-  border: none;
-  color: var(--text-muted);
-  font-size: 12px;
-  cursor: pointer;
-  font-family: var(--font);
-  padding: 6px 10px;
-  border-radius: 6px;
-  transition: color 0.15s;
-}
-.onboarding-skip:hover { color: var(--text); }
-.onboarding-next {
-  background: var(--accent);
-  color: var(--bg);
-  border: none;
-  border-radius: 6px;
-  padding: 8px 20px;
-  font-size: 13px;
-  font-weight: 600;
-  cursor: pointer;
-  font-family: var(--font);
-  transition: opacity 0.15s;
-}
-.onboarding-next:hover { opacity: 0.85; }
-.onboarding-welcome-center {
-  position: fixed;
-  inset: 0;
-  z-index: 10002;
-  display: flex;
-  align-items: center;
-  justify-content: center;
-  pointer-events: none;
-}
-.onboarding-welcome-card {
-  background: var(--bg-card);
-  border: 1px solid var(--accent);
-  border-radius: 16px;
-  padding: 40px 44px 32px;
-  width: 460px;
-  max-width: 90vw;
-  text-align: center;
-  box-shadow: 0 20px 60px rgba(0,0,0,0.5), 0 0 80px rgba(0,212,170,0.08);
-  pointer-events: auto;
-  opacity: 0;
-  transform: scale(0.95);
-  transition: all 0.4s cubic-bezier(0.4,0,0.2,1);
-}
-.onboarding-welcome-card.visible {
-  opacity: 1;
-  transform: scale(1);
-}
-.onboarding-welcome-logo {
-  width: 64px;
-  height: 64px;
-  margin-bottom: 16px;
-}
-.onboarding-welcome-title {
-  font-size: 24px;
-  font-weight: 700;
-  color: var(--accent-text);
-  margin-bottom: 8px;
-}
-.onboarding-welcome-sub {
-  font-size: 14px;
-  color: var(--text-dim);
-  line-height: 1.6;
-  margin-bottom: 28px;
-}
-[data-theme="light"] .onboarding-tooltip {
-  box-shadow: 0 12px 40px rgba(0,30,43,0.15), 0 0 0 1px rgba(0,158,128,0.2);
-}
-[data-theme="light"] .onboarding-welcome-card {
-  box-shadow: 0 20px 60px rgba(0,30,43,0.18), 0 0 80px rgba(0,158,128,0.06);
-}
-[data-theme="light"] .onboarding-backdrop {
-  background: rgba(0,30,43,0.45);
-}
-
 /* ── App Shell: sidebar + content layout ── */
 .app-shell {
   display: flex;
@@ -11065,18 +10879,6 @@ Retrieval augmented generation combines search with LLMs</textarea>
             </div>
           </div>
         </div>
-        <div class="settings-section">
-          <div class="settings-section-title">Help</div>
-          <div class="settings-row">
-            <div class="settings-label">
-              <span class="settings-label-text">Welcome Tour</span>
-              <span class="settings-label-hint">Replay the onboarding walkthrough</span>
-            </div>
-            <div class="settings-control">
-              <button class="settings-reset-btn" id="settingsShowTour" style="background:var(--accent);color:var(--bg);">Show Welcome Tour</button>
-            </div>
-          </div>
-        </div>
         <!-- Version Info (visible in Electron only) -->
         <div class="settings-section" id="settingsVersionSection" style="display:none;">
           <div class="settings-section-title">Version</div>
@@ -11153,38 +10955,6 @@ Retrieval augmented generation combines search with LLMs</textarea>
 </div><!-- .content-area -->
 </div><!-- .app-shell -->
 
-<!-- Onboarding Walkthrough Overlay -->
-<div class="onboarding-overlay" id="onboardingOverlay">
-  <div class="onboarding-backdrop" id="onboardingBackdrop"></div>
-  <div class="onboarding-spotlight" id="onboardingSpotlight"></div>
-  <!-- Welcome card (step 0) -->
-  <div class="onboarding-welcome-center" id="onboardingWelcomeWrap">
-    <div class="onboarding-welcome-card" id="onboardingWelcomeCard">
-      <img class="onboarding-welcome-logo" id="onboardingLogo" src="/icons/dark/64.png" alt="Vai">
-      <div class="onboarding-welcome-title">Welcome to Vai</div>
-      <div class="onboarding-welcome-sub">Explore embeddings, compare text similarity, and search with vector models — all from one playground.</div>
-      <div class="onboarding-footer" style="justify-content:center;">
-        <button class="onboarding-skip" id="onboardingWelcomeSkip">Skip tour</button>
-        <button class="onboarding-next" id="onboardingWelcomeNext">Take the tour →</button>
-      </div>
-    </div>
-  </div>
-  <!-- Tooltip for steps 1-4 -->
-  <div class="onboarding-tooltip" id="onboardingTooltip">
-    <div class="onboarding-tooltip-arrow top" id="onboardingArrow"></div>
-    <div class="onboarding-step-icon" id="onboardingIcon"></div>
-    <div class="onboarding-step-title" id="onboardingTitle"></div>
-    <div class="onboarding-step-body" id="onboardingBody"></div>
-    <div class="onboarding-footer">
-      <div class="onboarding-dots" id="onboardingDots"></div>
-      <div class="onboarding-actions">
-        <button class="onboarding-skip" id="onboardingSkip">Skip</button>
-        <button class="onboarding-next" id="onboardingNext">Next</button>
-      </div>
-    </div>
-  </div>
-</div>
-
 <script>
 // Apply saved theme immediately to prevent flash
 (function() {
@@ -15929,254 +15699,10 @@ function checkForAppUpdate() {
 // ── Onboarding Walkthrough ──
 function initOnboarding() {
   const ONBOARDING_KEY = 'vai-onboarding-complete';
-  const overlay = document.getElementById('onboardingOverlay');
-  const welcomeWrap = document.getElementById('onboardingWelcomeWrap');
-  const welcomeCard = document.getElementById('onboardingWelcomeCard');
-  const tooltip = document.getElementById('onboardingTooltip');
-  const spotlight = document.getElementById('onboardingSpotlight');
-  const arrow = document.getElementById('onboardingArrow');
-  const dotsContainer = document.getElementById('onboardingDots');
-  const titleEl = document.getElementById('onboardingTitle');
-  const bodyEl = document.getElementById('onboardingBody');
-  const iconEl = document.getElementById('onboardingIcon');
-
-  if (!overlay) return;
-
-  const isElectron = !!(window.vai && window.vai.isElectron);
-
-  const steps = [
-    {
-      target: null, // welcome card, no spotlight
-      icon: '🚀',
-      title: 'Welcome to Vai',
-      body: 'Explore embeddings, build RAG pipelines, generate production code — all from one playground.',
-    },
-    {
-      target: '[data-tab="settings"]',
-      icon: '🔑',
-      title: 'API Key Setup',
-      body: 'First, add your <strong>Voyage AI API key</strong> in Settings. You can get one free at <strong>dash.voyageai.com</strong>.' +
-        (isElectron ? '<br><br>💎 On desktop, your key is encrypted via <strong>OS keychain</strong> for secure storage.' : ''),
-      arrow: 'left',
-    },
-    {
-      target: '[data-tab="embed"]',
-      icon: '⚡',
-      title: 'Embed',
-      body: 'Turn any text into a <strong>vector embedding</strong> — a numerical fingerprint that captures meaning. Visualize the raw vectors and explore what models produce.',
-      arrow: 'left',
-    },
-    {
-      target: '[data-tab="compare"]',
-      icon: '🔥',
-      title: 'Compare',
-      body: 'Paste multiple texts and see a <strong>similarity heatmap</strong> — instantly discover which phrases are semantically close and which are far apart.',
-      arrow: 'left',
-    },
-    {
-      target: '[data-tab="search"]',
-      icon: '🔍',
-      title: 'Search & Rerank',
-      body: 'Run <strong>vector search</strong> against a set of documents and optionally <strong>rerank</strong> results for higher precision. Great for building RAG pipelines.',
-      arrow: 'left',
-    },
-    {
-      target: '[data-tab="multimodal"]',
-      icon: '🖼️',
-      title: 'Multimodal',
-      body: 'Embed <strong>images and text</strong> in the same vector space. Search images with text queries, or find similar images across your collection.',
-      arrow: 'left',
-    },
-    {
-      target: '[data-tab="generate"]',
-      icon: '💻',
-      title: 'Generate & Scaffold',
-      body: '<strong>Generate code snippets</strong> for retrieval, ingestion, and API routes — or <strong>scaffold entire projects</strong> with Next.js, Express, or Flask.' +
-        (isElectron ? ' Create projects directly on disk.' : ' Download as ZIP.'),
-      arrow: 'left',
-    },
-    {
-      target: '[data-tab="chat"]',
-      icon: '💬',
-      title: 'RAG Chat',
-      body: '<strong>Chat with your knowledge base</strong> using retrieval-augmented generation. Voyage AI finds relevant documents, then your chosen LLM generates grounded answers with source citations.',
-      arrow: 'left',
-    },
-    {
-      target: '[data-tab="workflows"]',
-      icon: '\u2699\uFE0F',
-      title: 'Workflow Visualizer',
-      body: '<strong>Visualize and execute workflows</strong> as interactive DAGs. Browse built-in templates, inspect step configuration, and watch execution animate in real time.',
-      arrow: 'left',
-    },
-    {
-      target: '[data-tab="benchmark"]',
-      icon: '⏱️',
-      title: 'Benchmark',
-      body: 'Compare model <strong>latency, throughput, and cost</strong>. Test embedding speeds, reranking quality, and quantization trade-offs.',
-      arrow: 'left',
-    },
-    {
-      target: '[data-tab="explore"]',
-      icon: '💡',
-      title: 'Explore & Learn',
-      body: 'Dive into <strong>25 interactive explainers</strong> covering embeddings, vector search, RAG, reranking, and more. Each topic links to live demos.',
-      arrow: 'left',
-    },
-  ];
-
-  let currentStep = 0;
-  const totalSteps = steps.length;
-
-  function buildDots() {
-    dotsContainer.innerHTML = '';
-    for (let i = 1; i < totalSteps; i++) {
-      const dot = document.createElement('div');
-      dot.className = 'onboarding-dot';
-      if (i < currentStep) dot.classList.add('completed');
-      if (i === currentStep) dot.classList.add('active');
-      dotsContainer.appendChild(dot);
-    }
-  }
-
-  function positionTooltipNear(targetEl) {
-    const step = steps[currentStep];
-    const rect = targetEl.getBoundingClientRect();
-
-    // Position spotlight over the target
-    spotlight.style.display = 'block';
-    const pad = 6;
-    spotlight.style.left = (rect.left - pad) + 'px';
-    spotlight.style.top = (rect.top - pad) + 'px';
-    spotlight.style.width = (rect.width + pad * 2) + 'px';
-    spotlight.style.height = (rect.height + pad * 2) + 'px';
-
-    // Reset arrow classes
-    arrow.className = 'onboarding-tooltip-arrow';
-
-    // Position tooltip to the right of sidebar items
-    if (step.arrow === 'left') {
-      arrow.classList.add('left');
-      tooltip.style.left = (rect.right + 16) + 'px';
-      tooltip.style.top = Math.max(8, rect.top - 10) + 'px';
-      tooltip.style.right = 'auto';
-      tooltip.style.bottom = 'auto';
-    } else {
-      // Below the target
-      arrow.classList.add('top');
-      tooltip.style.left = Math.max(8, rect.left) + 'px';
-      tooltip.style.top = (rect.bottom + 14) + 'px';
-      tooltip.style.right = 'auto';
-      tooltip.style.bottom = 'auto';
-    }
-
-    // Ensure tooltip doesn't overflow viewport
-    requestAnimationFrame(() => {
-      const tr = tooltip.getBoundingClientRect();
-      if (tr.bottom > window.innerHeight - 10) {
-        tooltip.style.top = Math.max(8, window.innerHeight - tr.height - 10) + 'px';
-      }
-      if (tr.right > window.innerWidth - 10) {
-        tooltip.style.left = Math.max(8, window.innerWidth - tr.width - 10) + 'px';
-      }
-    });
-  }
-
-  function showStep(idx) {
-    currentStep = idx;
-    const step = steps[idx];
-
-    if (idx === 0) {
-      // Welcome card — centered, no spotlight
-      welcomeWrap.style.display = 'flex';
-      tooltip.classList.remove('visible');
-      tooltip.style.display = 'none';
-      spotlight.style.display = 'none';
-      requestAnimationFrame(() => {
-        welcomeCard.classList.add('visible');
-      });
-      return;
-    }
-
-    // Hide welcome card
-    welcomeWrap.style.display = 'none';
-    welcomeCard.classList.remove('visible');
-
-    // Show tooltip
-    tooltip.style.display = 'block';
-    tooltip.classList.remove('visible');
-    iconEl.textContent = step.icon;
-    titleEl.textContent = step.title;
-    bodyEl.innerHTML = step.body;
-    buildDots();
-
-    // Update next button text
-    const nextBtn = document.getElementById('onboardingNext');
-    nextBtn.textContent = (idx === totalSteps - 1) ? 'Get Started' : 'Next';
-
-    // Find target element and position
-    const targetEl = document.querySelector(step.target);
-    if (targetEl) {
-      positionTooltipNear(targetEl);
-    }
-
-    requestAnimationFrame(() => {
-      tooltip.classList.add('visible');
-    });
-  }
-
-  function finish() {
-    tooltip.classList.remove('visible');
-    welcomeCard.classList.remove('visible');
-    spotlight.style.display = 'none';
-    setTimeout(() => {
-      overlay.classList.remove('active');
-    }, 300);
-    localStorage.setItem(ONBOARDING_KEY, 'true');
-  }
-
-  function nextStep() {
-    if (currentStep < totalSteps - 1) {
-      showStep(currentStep + 1);
-    } else {
-      finish();
-    }
-  }
-
-  // Wire up buttons
-  document.getElementById('onboardingWelcomeNext').addEventListener('click', nextStep);
-  document.getElementById('onboardingWelcomeSkip').addEventListener('click', finish);
-  document.getElementById('onboardingNext').addEventListener('click', nextStep);
-  document.getElementById('onboardingSkip').addEventListener('click', finish);
-  document.getElementById('onboardingBackdrop').addEventListener('click', finish);
-
-  // "Show Welcome Tour" button in settings
-  const tourBtn = document.getElementById('settingsShowTour');
-  if (tourBtn) {
-    tourBtn.addEventListener('click', () => {
-      startOnboarding();
-    });
-  }
-
-  // Auto-start on first visit
-  function startOnboarding() {
-    // Sync onboarding logo with current theme
-    const onboardLogo = document.getElementById('onboardingLogo');
-    if (onboardLogo) {
-      const theme = localStorage.getItem('vai-theme') || 'dark';
-      onboardLogo.src = '/icons/' + (theme === 'light' ? 'light' : 'dark') + '/64.png';
-    }
-    overlay.classList.add('active');
-    showStep(0);
-  }
-
-  if (!localStorage.getItem(ONBOARDING_KEY)) {
-    // Delay slightly so the app is fully rendered
-    setTimeout(startOnboarding, 600);
-  }
-
-  // Expose for manual replay
-  window.startOnboarding = startOnboarding;
+  // The legacy welcome tour no longer matches the redesigned interface.
+  // Mark it complete so startup skips the broken onboarding flow.
+  localStorage.setItem(ONBOARDING_KEY, 'true');
+  window.startOnboarding = function() {};
 }
 
 // ── Multimodal Tab ──
@@ -16871,7 +16397,6 @@ let _vsiGameTrigger = 'unknown';
       'exploreModal',
       'costHelpModal',
       'agentInfoModal',
-      'onboardingOverlay',
       'wfStoreOverlay',
       'wfOutputModalBackdrop',
       'wfHelpModalBackdrop',

package/src/playground/js/kb-manager.js

@@ -9,11 +9,20 @@ class KBManager {
     this.kbs = [];
     this.isIngesting = false;
     this.ingestionProgress = { current: 0, total: 0, currentFile: '' };
-    
+
     // Local state
     this.loadLastKB();
   }
 
+  /**
+   * Get the currently selected embedding model from the chat config UI.
+   * Falls back to 'voyage-4-large' if the element is missing.
+   */
+  getEmbeddingModel() {
+    const sel = document.getElementById('chatEmbeddingModel');
+    return sel ? sel.value : 'voyage-4-large';
+  }
+
   /**
    * Load last used KB from localStorage
    */
@@ -97,6 +106,7 @@ class KBManager {
     if (kbName) {
       formData.append('kbName', kbName);
     }
+    formData.append('embeddingModel', this.getEmbeddingModel());
 
     try {
       this.isIngesting = true;
@@ -175,7 +185,7 @@ class KBManager {
       const res = await fetch('/api/rag/ingest-text', {
         method: 'POST',
         headers: { 'Content-Type': 'application/json' },
-        body: JSON.stringify({ text, kbName, title })
+        body: JSON.stringify({ text, kbName, title, embeddingModel: this.getEmbeddingModel() })
       });
 
       if (!res.ok) {
@@ -239,7 +249,7 @@ class KBManager {
       const res = await fetch('/api/rag/ingest-url', {
         method: 'POST',
         headers: { 'Content-Type': 'application/json' },
-        body: JSON.stringify({ url, kbName })
+        body: JSON.stringify({ url, kbName, embeddingModel: this.getEmbeddingModel() })
       });
 
       if (!res.ok) {

package/src/playground/js/kb-ui.js

@@ -499,6 +499,7 @@ class KBUIManager {
     const progressBar = document.getElementById('kbPanelProgressBar');
     const progressLabel = document.getElementById('kbPanelProgressLabel');
     if (!progressContainer) return;
+    const warnings = [];
 
     try {
       progressContainer.style.display = 'block';
@@ -531,12 +532,27 @@ class KBUIManager {
 
           if (progressBar) progressBar.style.width = `${percent}%`;
           if (progressLabel) progressLabel.textContent = label;
+        } else if (event.type === 'warning') {
+          if (event.warning) warnings.push(event.warning);
+          if (progressLabel) progressLabel.textContent = event.warning || 'Upload completed with warnings';
         } else if (event.type === 'complete') {
           if (progressBar) progressBar.style.width = '100%';
-          if (progressLabel) progressLabel.textContent = `✓ Complete: ${event.docCount} docs, ${event.chunkCount} chunks`;
+          const hasWarnings = warnings.length > 0 || event.docCount === 0;
+          if (progressLabel) {
+            progressLabel.textContent = hasWarnings
+              ? `Complete with warnings: ${event.docCount} docs, ${event.chunkCount} chunks`
+              : `✓ Complete: ${event.docCount} docs, ${event.chunkCount} chunks`;
+          }
           setTimeout(() => {
             this.updatePanelUI();
             progressContainer.style.display = 'none';
+            const messages = [...warnings];
+            if (event.docCount === 0) {
+              messages.unshift('No documents were stored from that upload.');
+            }
+            if (messages.length > 0) {
+              alert(messages.join('\n'));
+            }
           }, 1000);
         } else if (event.type === 'error') {
           console.error('Ingestion error:', event.error);