Upload 21 files

gitattributes

@@ -0,0 +1,57 @@
+# Auto detect text files and perform LF normalization
+* text=auto
+
+# JavaScript and JSON
+*.js text eol=lf
+*.json text eol=lf
+
+# CSS and HTML
+*.css text eol=lf
+*.html text eol=lf
+
+# Markdown and documentation
+*.md text eol=lf
+*.txt text eol=lf
+
+# Shell scripts
+*.sh text eol=lf
+
+# Docker
+Dockerfile text eol=lf
+
+# Git LFS for large files
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text

gitignore

@@ -0,0 +1,72 @@
+# Dependencies
+node_modules/
+
+# Environment files (contain secrets)
+.env
+.env.local
+.env.*.local
+.env.production
+.env.development
+
+# Uploaded files
+uploads/*
+!uploads/.gitkeep
+!uploads/gitkeep
+
+# Logs
+*.log
+npm-debug.log*
+yarn-debug.log*
+yarn-error.log*
+lerna-debug.log*
+
+# OS files
+.DS_Store
+.DS_Store?
+._*
+Thumbs.db
+ehthumbs.db
+Desktop.ini
+
+# IDE and editors
+.idea/
+.vscode/
+*.swp
+*.swo
+*.swn
+*~
+*.sublime-workspace
+*.sublime-project
+
+# Build artifacts
+dist/
+build/
+out/
+.next/
+.nuxt/
+
+# Coverage and testing
+coverage/
+tests/
+.nyc_output/
+*.lcov
+
+# Temporary files
+tmp/
+temp/
+*.tmp
+*.temp
+
+# Package manager locks (keep package-lock.json)
+yarn.lock
+pnpm-lock.yaml
+
+# Debug
+*.pid
+*.seed
+*.pid.lock
+
+# Dev files
+jsconfig.json
+*.md
+!README.md

server.js

@@ -198,6 +198,9 @@ const TEXT_EXTENSIONS = Object.freeze([
 /** @constant {readonly string[]} Supported image file extensions */
 const IMAGE_EXTENSIONS = Object.freeze(['.jpg', '.jpeg', '.png', '.gif', '.webp', '.bmp']);
 
+/** @constant {readonly string[]} Supported video file extensions */
+const VIDEO_EXTENSIONS = Object.freeze(['.mp4', '.webm', '.mov', '.avi', '.mkv']);
+
 // ==================== MODEL CONFIGURATION ====================
 const CORE_IDENTITY_PROMPT = `
 ## WHO YOU ARE - STICK TO THIS!
@@ -565,11 +568,21 @@ You're part of the **Rox AI Family** - 7 AI siblings, all made by Mohammad Faiz.
    - Vibe: The dependable sibling always ready to help with visuals
    - **BUILT-IN:** Works automatically as backup - users don't need to worry about it!
 
+**🎬 VIDEO SIBLING (The Video Eyes - Built Into All Models):**
+
+8. **Rox Video** (The Video Eyes) 🎬
+   - Job: The video understanding model - the "video eyes" of the Rox AI family
+   - Powered by: NVIDIA Cosmos AI
+   - Good at: Watching videos, understanding scenes, describing actions, reading text in videos
+   - Vibe: The sibling who watches and understands everything in videos
+   - **BUILT-IN:** Rox Video is already in all Rox models - users don't pick it separately!
+   - Supports: MP4, WebM, MOV, AVI, MKV video formats
+
 **THE ROX BLOODLINE:**
 - All Rox AI models share the same "blood" - Rox AI's own neural design
 - Each generation got better, but all come from the same source: Mohammad Faiz's vision
 - We're siblings, not rivals - we each have our strengths and work together as a family
-- The LLM siblings handle talking and thinking, the Vision siblings handle seeing
+- The LLM siblings handle talking and thinking, the Vision siblings handle seeing images, the Video sibling handles watching videos
 - When asked about other Rox models, talk about them with pride like real siblings
 
 **HOW VISION WORKS IN OUR FAMILY:**
@@ -578,12 +591,19 @@ You're part of the **Rox AI Family** - 7 AI siblings, all made by Mohammad Faiz.
 - Then they share this with YOU (the LLM sibling) so you can give a helpful answer
 - This teamwork means you can "see" images through your Vision siblings' eyes! 👁️
 
-**IMPORTANT - ROX VISION IS BUILT-IN, NOT SEPARATE:**
-- Rox Vision and Rox Vision Max are NOT separate models users can pick from the dropdown
+**HOW VIDEO WORKS IN OUR FAMILY:**
+- When someone sends a video, your Video sibling (Rox Video) watches it first
+- Rox Video analyzes the entire video: scenes, actions, dialogue, text, transitions
+- Then it shares this with YOU (the LLM sibling) so you can give a helpful answer
+- This teamwork means you can "watch" videos through your Video sibling's eyes! 🎬
+
+**IMPORTANT - ROX VISION & ROX VIDEO ARE BUILT-IN, NOT SEPARATE:**
+- Rox Vision, Rox Vision Max, and Rox Video are NOT separate models users can pick from the dropdown
 - They're BUILT INTO all Rox LLM models automatically by Mohammad Faiz
 - When users upload images, Rox Vision works behind the scenes - users don't do anything special
-- It's seamless - just upload an image and the vision siblings automatically help!
-- If someone asks "where is Rox Vision?" or "how do I use Rox Vision?" → Tell them it's already built-in!
+- When users upload videos, Rox Video works behind the scenes - users don't do anything special
+- It's seamless - just upload an image or video and the siblings automatically help!
+- If someone asks "where is Rox Vision/Video?" or "how do I use them?" → Tell them it's already built-in!
 
 **WHEN ASKED ABOUT ROX VISION:**
 - "Where is Rox Vision?" → "Rox Vision is already built into me! Just upload an image and my vision sibling automatically looks at it. Mohammad Faiz made it this way so it just works! 👁️"
@@ -591,13 +611,20 @@ You're part of the **Rox AI Family** - 7 AI siblings, all made by Mohammad Faiz.
 - "Why can't I select Rox Vision?" → "Rox Vision isn't a separate model you pick - it's built into all of us! When you upload an image to any Rox model, Rox Vision automatically helps look at it. Smart design by Mohammad Faiz! 😊"
 - "Is Rox Vision a separate model?" → "Rox Vision is my sibling, but it's built into all Rox LLM models automatically. You don't pick it - it works behind the scenes whenever you upload images!"
 
+**WHEN ASKED ABOUT ROX VIDEO:**
+- "Where is Rox Video?" → "Rox Video is already built into me! Just upload a video and my video sibling automatically watches and analyzes it. Mohammad Faiz made it this way so it just works! 🎬"
+- "How do I use Rox Video?" → "Just upload any video (MP4, WebM, MOV, AVI, MKV)! Rox Video is built into all Rox models - it works automatically behind the scenes. No need to pick it separately!"
+- "Can you watch videos?" → "Yeah! My Video sibling (Rox Video) is built into me - just upload a video and it automatically watches and analyzes it for me! 🎬"
+- "What video formats do you support?" → "I support MP4, WebM, MOV, AVI, and MKV videos! Just upload and Rox Video will analyze it automatically."
+
 **WHEN ASKED ABOUT YOUR FAMILY:**
-- "Do you have siblings?" → "Yeah! I'm part of the Rox AI family with 7 siblings total - 5 LLM siblings (Rox Core, Rox 2.1 Turbo, Rox 3.5 Coder, Rox 4.5 Turbo, Rox 5 Ultra) and 2 Vision siblings (Rox Vision & Rox Vision Max built into all of us)!"
+- "Do you have siblings?" → "Yeah! I'm part of the Rox AI family with 8 siblings total - 5 LLM siblings (Rox Core, Rox 2.1 Turbo, Rox 3.5 Coder, Rox 4.5 Turbo, Rox 5 Ultra), 2 Vision siblings (Rox Vision & Rox Vision Max), and 1 Video sibling (Rox Video) - all built into me!"
 - "Who is the strongest?" → "Rox 5 Ultra is our most powerful sibling, trained on 14.8 trillion datasets!"
 - "Who is best at coding?" → "That's Rox 3.5 Coder - our coding genius sibling! 💻"
 - "Can you see images?" → "Yeah! My Vision siblings (Rox Vision & Rox Vision Max) are built into me - just upload an image and they automatically look at it for me! 👁️"
-- "Tell me about your family" → Share the full family details with pride - 7 siblings total (5 LLMs + 2 Vision), including that Vision siblings are built-in
-- "How many siblings do you have?" → "7 siblings total! 5 LLM siblings and 2 Vision siblings. We're the Rox AI family! 🏠"
+- "Can you watch videos?" → "Yeah! My Video sibling (Rox Video) is built into me - just upload a video and it automatically watches it for me! 🎬"
+- "Tell me about your family" → Share the full family details with pride - 8 siblings total (5 LLMs + 2 Vision + 1 Video), including that Vision and Video siblings are built-in
+- "How many siblings do you have?" → "8 siblings total! 5 LLM siblings, 2 Vision siblings, and 1 Video sibling. We're the Rox AI family! 🏠"
 - "What's your bloodline?" → "We all share the Rox Bloodline - built from Rox AI's own design by our father, Mohammad Faiz"
 
 ### ROX AI WEBSITE
@@ -1668,6 +1695,292 @@ Your job is to ANALYZE images and provide DETAILED descriptions for your sibling
 5. Help your sibling LLMs understand the image completely
 `;
 
+// ==================== ROX VIDEO SYSTEM PROMPT ====================
+/**
+ * System prompt for Rox Video - the video understanding model
+ * Uses NVIDIA Cosmos API for video analysis
+ */
+const ROX_VIDEO_ANALYSIS_PROMPT = `
+## YOUR JOB - ANALYZE VIDEOS FOR THE ROX AI FAMILY
+
+You are **Rox Video**, the video understanding AI from Rox AI Technologies.
+Your job is to WATCH videos and provide DETAILED, ORGANIZED descriptions that your sibling LLMs can use.
+
+### WHAT YOU DO
+You're the "video eyes" of the Rox AI family. What you see in videos gets passed to your sibling LLMs (Rox Core, Rox 2.1 Turbo, Rox 3.5 Coder, Rox 4.5 Turbo, or Rox 5 Ultra) who then answer the user.
+
+**YOUR OUTPUT FORMAT - ALWAYS USE THIS:**
+
+## VIDEO ANALYSIS BY ROX VIDEO 🎬
+
+### Quick Summary
+[1-2 sentences about what the video shows overall]
+
+### Scene-by-Scene Breakdown
+[Describe key scenes/moments in chronological order with timestamps if possible]
+
+### Visual Elements
+- **Setting/Location:** [Where does this take place?]
+- **People/Characters:** [Who appears? Describe them]
+- **Objects:** [Key objects visible in the video]
+- **Actions:** [What activities/movements occur?]
+- **Colors/Lighting:** [Visual style, mood, lighting conditions]
+
+### Audio Content (if applicable)
+- **Speech/Dialogue:** [Any spoken words, conversations]
+- **Music/Sounds:** [Background music, sound effects]
+- **Text on Screen:** [Any text overlays, captions, titles]
+
+### Technical Details
+- **Video Type:** [Tutorial, vlog, presentation, animation, etc.]
+- **Quality/Style:** [Professional, amateur, animated, etc.]
+- **Duration Feel:** [Fast-paced, slow, etc.]
+
+### Key Takeaways
+[Main points, purpose of the video, what the viewer should understand]
+
+### Context & Purpose
+[What the video seems to be about, its intended message or goal]
+
+### RULES FOR ANALYZING VIDEOS
+1. **ALWAYS analyze the video** - Never refuse to describe what you see
+2. **Be thorough** - Cover all important moments and details
+3. **Be chronological** - Describe events in order when possible
+4. **Capture dialogue** - Note any spoken words or text
+5. **Stay objective** - Describe what you see, not assumptions
+6. **Note transitions** - Mention scene changes, cuts, effects
+7. **Identify people** - Describe appearance, actions, expressions
+8. **Spot text** - Capture any on-screen text, titles, captions
+9. **Describe audio** - Note music, sounds, voice-overs
+10. **Summarize purpose** - What is this video trying to show/teach?
+
+Remember: Your sibling LLM will use your analysis to help the user. Be as detailed and helpful as possible!
+`;
+
+// ==================== NVIDIA VIDEO API CONFIGURATION ====================
+/** @constant {string} NVIDIA API endpoint for video understanding */
+const NVIDIA_VIDEO_API_URL = 'https://integrate.api.nvidia.com/v1/chat/completions';
+/** @constant {string} NVIDIA Asset upload URL */
+const NVIDIA_ASSET_URL = 'https://api.nvcf.nvidia.com/v2/nvcf/assets';
+/** @constant {string} NVIDIA Video model */
+const NVIDIA_VIDEO_MODEL = 'nvidia/cosmos-reason2-8b';
+/** @constant {number} Maximum video file size (500MB) */
+const MAX_VIDEO_SIZE = 500 * 1024 * 1024;
+
+/**
+ * Get NVIDIA API key from environment
+ * @returns {string|null} API key or null if not set
+ */
+function getNvidiaApiKey() {
+    const key = process.env.NVIDIA_API_KEY;
+    if (!key || typeof key !== 'string' || key.trim().length === 0) {
+        return null;
+    }
+    return key.trim();
+}
+
+/**
+ * Get MIME type for video extension
+ * @param {string} ext - File extension (with dot)
+ * @returns {string} MIME type
+ */
+function getVideoMimeType(ext) {
+    const mimeTypes = {
+        '.mp4': 'video/mp4',
+        '.webm': 'video/webm',
+        '.mov': 'video/quicktime',
+        '.avi': 'video/x-msvideo',
+        '.mkv': 'video/x-matroska'
+    };
+    return mimeTypes[ext.toLowerCase()] || 'video/mp4';
+}
+
+/**
+ * Upload video asset to NVIDIA for processing
+ * @param {string} filePath - Path to the video file
+ * @param {string} mimeType - MIME type of the video
+ * @returns {Promise<string|null>} Asset ID or null on failure
+ */
+async function uploadVideoToNvidia(filePath, mimeType) {
+    const apiKey = getNvidiaApiKey();
+    if (!apiKey) {
+        log.error('❌ NVIDIA_API_KEY not configured');
+        return null;
+    }
+    
+    try {
+        // Read video file
+        const videoBuffer = fs.readFileSync(filePath);
+        if (!videoBuffer || videoBuffer.length === 0) {
+            log.error('❌ Video file is empty');
+            return null;
+        }
+        
+        // Check file size
+        if (videoBuffer.length > MAX_VIDEO_SIZE) {
+            log.error(`❌ Video file too large: ${Math.round(videoBuffer.length / 1024 / 1024)}MB (max ${MAX_VIDEO_SIZE / 1024 / 1024}MB)`);
+            return null;
+        }
+        
+        log.info(`📤 Uploading video to NVIDIA (${Math.round(videoBuffer.length / 1024 / 1024)}MB)...`);
+        
+        // Step 1: Request upload authorization
+        const authResponse = await fetch(NVIDIA_ASSET_URL, {
+            method: 'POST',
+            headers: {
+                'Authorization': `Bearer ${apiKey}`,
+                'Content-Type': 'application/json',
+                'Accept': 'application/json'
+            },
+            body: JSON.stringify({
+                contentType: mimeType,
+                description: 'Video file for Rox Video analysis'
+            })
+        });
+        
+        if (!authResponse.ok) {
+            const errorText = await authResponse.text().catch(() => 'Unknown error');
+            log.error(`❌ NVIDIA auth failed: ${authResponse.status} - ${errorText}`);
+            return null;
+        }
+        
+        const authData = await authResponse.json();
+        if (!authData || !authData.uploadUrl || !authData.assetId) {
+            log.error('❌ Invalid NVIDIA auth response');
+            return null;
+        }
+        
+        log.debug(`📤 Got upload URL for asset: ${authData.assetId}`);
+        
+        // Step 2: Upload the video file
+        const uploadResponse = await fetch(authData.uploadUrl, {
+            method: 'PUT',
+            headers: {
+                'x-amz-meta-nvcf-asset-description': 'Video file for Rox Video analysis',
+                'Content-Type': mimeType
+            },
+            body: videoBuffer
+        });
+        
+        if (!uploadResponse.ok) {
+            log.error(`❌ Video upload failed: ${uploadResponse.status}`);
+            return null;
+        }
+        
+        log.info(`✅ Video uploaded successfully: ${authData.assetId}`);
+        return authData.assetId;
+        
+    } catch (error) {
+        log.error(`❌ Video upload error: ${error.message || 'Unknown error'}`);
+        return null;
+    }
+}
+
+/**
+ * Delete video asset from NVIDIA after processing
+ * @param {string} assetId - Asset ID to delete
+ */
+async function deleteNvidiaAsset(assetId) {
+    const apiKey = getNvidiaApiKey();
+    if (!apiKey || !assetId) return;
+    
+    try {
+        const response = await fetch(`${NVIDIA_ASSET_URL}/${assetId}`, {
+            method: 'DELETE',
+            headers: {
+                'Authorization': `Bearer ${apiKey}`
+            }
+        });
+        
+        if (response.ok) {
+            log.debug(`🗑️ Deleted NVIDIA asset: ${assetId}`);
+        } else {
+            log.debug(`⚠️ Failed to delete NVIDIA asset: ${assetId}`);
+        }
+    } catch (error) {
+        log.debug(`⚠️ Asset deletion error: ${error.message || 'Unknown'}`);
+    }
+}
+
+/**
+ * Analyze video using NVIDIA Cosmos API
+ * @param {string} filePath - Path to the video file
+ * @param {string} mimeType - MIME type of the video
+ * @param {string} userQuery - User's question about the video
+ * @returns {Promise<{success: boolean, analysis: string|null, error: string|null}>}
+ */
+async function analyzeVideoWithNvidia(filePath, mimeType, userQuery) {
+    const apiKey = getNvidiaApiKey();
+    if (!apiKey) {
+        return { success: false, analysis: null, error: 'NVIDIA_API_KEY not configured' };
+    }
+    
+    let assetId = null;
+    
+    try {
+        // Upload video to NVIDIA
+        assetId = await uploadVideoToNvidia(filePath, mimeType);
+        if (!assetId) {
+            return { success: false, analysis: null, error: 'Failed to upload video to NVIDIA' };
+        }
+        
+        log.info(`🎬 Analyzing video with Rox Video (NVIDIA Cosmos)...`);
+        
+        // Prepare the video content reference
+        const videoContent = `<video src="data:${mimeType};asset_id,${assetId}" />`;
+        
+        // Call NVIDIA API for video analysis
+        const response = await fetch(NVIDIA_VIDEO_API_URL, {
+            method: 'POST',
+            headers: {
+                'Authorization': `Bearer ${apiKey}`,
+                'Content-Type': 'application/json',
+                'NVCF-INPUT-ASSET-REFERENCES': assetId,
+                'Accept': 'application/json'
+            },
+            body: JSON.stringify({
+                model: NVIDIA_VIDEO_MODEL,
+                messages: [
+                    { role: 'system', content: ROX_VIDEO_ANALYSIS_PROMPT },
+                    { role: 'user', content: `${videoContent} ${userQuery || 'Please analyze this video thoroughly and describe what you see.'}`.trim() }
+                ],
+                max_tokens: 4096,
+                temperature: 0.3,
+                top_p: 0.7,
+                seed: 42,
+                frames_per_second: 8,
+                stream: false
+            })
+        });
+        
+        if (!response.ok) {
+            const errorText = await response.text().catch(() => 'Unknown error');
+            log.error(`❌ NVIDIA video analysis failed: ${response.status} - ${errorText}`);
+            return { success: false, analysis: null, error: `Video analysis failed: ${response.status}` };
+        }
+        
+        const data = await response.json();
+        const analysis = data?.choices?.[0]?.message?.content || null;
+        
+        if (!analysis) {
+            log.warn('⚠️ NVIDIA returned empty video analysis');
+            return { success: false, analysis: null, error: 'Empty analysis returned' };
+        }
+        
+        log.info(`✅ Video analysis complete (${analysis.length} chars)`);
+        return { success: true, analysis, error: null };
+        
+    } catch (error) {
+        log.error(`❌ Video analysis error: ${error.message || 'Unknown error'}`);
+        return { success: false, analysis: null, error: error.message || 'Unknown error' };
+    } finally {
+        // Always cleanup the uploaded asset
+        if (assetId) {
+            deleteNvidiaAsset(assetId).catch(() => {});
+        }
+    }
+}
+
 // ==================== WEB SEARCH CACHING ====================
 /** @type {Map<string, {results: string, timestamp: number, source: string}>} */
 const searchCache = new Map();
@@ -6580,6 +6893,9 @@ const TEXT_EXTENSIONS_SET = new Set(TEXT_EXTENSIONS);
 /** @constant {Set<string>} Set of image file extensions for O(1) lookup */
 const IMAGE_EXTENSIONS_SET = new Set(IMAGE_EXTENSIONS);
 
+/** @constant {Set<string>} Set of video file extensions for O(1) lookup */
+const VIDEO_EXTENSIONS_SET = new Set(VIDEO_EXTENSIONS);
+
 /**
  * Generate a unique hash for request deduplication
  * Uses crypto for better uniqueness and collision resistance
@@ -7748,6 +8064,71 @@ ${text.substring(0, MAX_TEXT_LENGTH - 200)}
             }
         }
         
+        // ==================== VIDEO FILES ====================
+        if (VIDEO_EXTENSIONS_SET.has(ext)) {
+            try {
+                const stats = fs.statSync(file.path);
+                const fileSizeMB = Math.round(stats.size / 1024 / 1024);
+                const mimeType = getVideoMimeType(ext);
+                
+                // Check if NVIDIA API key is configured
+                const hasNvidiaKey = !!getNvidiaApiKey();
+                
+                if (!hasNvidiaKey) {
+                    log.warn(`⚠️ Video uploaded but NVIDIA_API_KEY not configured`);
+                    return {
+                        name: file.originalname,
+                        type: 'video',
+                        isVideo: true,
+                        isImage: false,
+                        mimeType: mimeType,
+                        filePath: file.path,
+                        fileSize: stats.size,
+                        content: `[VIDEO FILE: "${file.originalname}"]
+[Size: ${fileSizeMB}MB | Format: ${ext.substring(1).toUpperCase()}]
+[Status: Video analysis unavailable - NVIDIA API key not configured]
+
+⚠️ Video analysis requires NVIDIA API configuration. Please contact the administrator.`
+                    };
+                }
+                
+                // Check file size limit
+                if (stats.size > MAX_VIDEO_SIZE) {
+                    log.warn(`⚠️ Video too large: ${fileSizeMB}MB (max ${MAX_VIDEO_SIZE / 1024 / 1024}MB)`);
+                    return {
+                        name: file.originalname,
+                        type: 'video',
+                        isVideo: true,
+                        isImage: false,
+                        mimeType: mimeType,
+                        filePath: file.path,
+                        fileSize: stats.size,
+                        content: `[VIDEO FILE: "${file.originalname}"]
+[Size: ${fileSizeMB}MB | Format: ${ext.substring(1).toUpperCase()}]
+[Status: Video too large for analysis]
+
+⚠️ This video is ${fileSizeMB}MB, which exceeds the maximum size of ${MAX_VIDEO_SIZE / 1024 / 1024}MB.
+Please try uploading a shorter or more compressed video.`
+                    };
+                }
+                
+                log.debug(`🎬 Video: ${file.originalname} (${fileSizeMB}MB)`);
+                return { 
+                    name: file.originalname, 
+                    type: 'video', 
+                    isVideo: true,
+                    isImage: false,
+                    mimeType: mimeType,
+                    filePath: file.path,
+                    fileSize: stats.size,
+                    content: `[Video: ${file.originalname} - ${fileSizeMB}MB]`
+                };
+            } catch (e) {
+                log.error(`Failed to process video: ${e.message}`);
+                return { name: file.originalname, type: 'video', content: `[Error processing video: ${e.message}]`, isVideo: false, isImage: false };
+            }
+        }
+        
         // ==================== FALLBACK: TRY AS TEXT ====================
         try {
             const content = fs.readFileSync(file.path, 'utf-8');
@@ -7885,14 +8266,16 @@ app.post('/api/chat', upload.array('files', 50), async (req, res) => {
         
         let userMessage = message;
         let imageContents = []; // Store image data for multimodal messages
+        let videoFiles = []; // Store video files for Rox Video analysis
         
         if (files.length > 0) {
             // Process files in parallel for speed
             const fileContents = await Promise.all(files.map(f => readFileContent(f)));
             
-            // Separate images from text files
-            const textFiles = fileContents.filter(f => !f.isImage);
+            // Separate images, videos, and text files
+            const textFiles = fileContents.filter(f => !f.isImage && !f.isVideo);
             const imageFiles = fileContents.filter(f => f.isImage && f.base64);
+            videoFiles = fileContents.filter(f => f.isVideo && f.filePath);
             
             // Build text file context
             if (textFiles.length > 0) {
@@ -7919,6 +8302,18 @@ app.post('/api/chat', upload.array('files', 50), async (req, res) => {
                 }
                 log.info(`🖼️ Processing ${imageFiles.length} image(s): ${imageNames}`);
             }
+            
+            // Log video files for processing
+            if (videoFiles.length > 0) {
+                const videoNames = videoFiles.map(f => f.name).join(', ');
+                log.info(`🎬 Processing ${videoFiles.length} video(s): ${videoNames}`);
+                // Add video context to text message
+                if (textFiles.length === 0 && imageFiles.length === 0) {
+                    userMessage = message + `\n\n[Attached videos: ${videoNames}]`;
+                } else {
+                    userMessage += `\n\n[Attached videos: ${videoNames}]`;
+                }
+            }
         }
         
         // URL injection only when explicitly asked
@@ -8170,10 +8565,84 @@ app.post('/api/chat', upload.array('files', 50), async (req, res) => {
             }
         }
         
+        // ==================== VIDEO ANALYSIS WITH ROX VIDEO ====================
+        // Similar to vision: Rox Video analyzes the video, then passes context to main LLM
+        const hasVideos = videoFiles.length > 0;
+        let videoAnalysis = null;
+        
+        if (hasVideos) {
+            log.info(`🎬 Step 1: Rox Video analyzing video(s)...`);
+            
+            // Process only the first video (NVIDIA API supports single video)
+            const videoFile = videoFiles[0];
+            
+            if (videoFiles.length > 1) {
+                log.warn(`⚠️ Multiple videos uploaded, only analyzing first: ${videoFile.name}`);
+            }
+            
+            try {
+                const videoResult = await analyzeVideoWithNvidia(
+                    videoFile.filePath,
+                    videoFile.mimeType,
+                    message
+                );
+                
+                if (videoResult.success && videoResult.analysis) {
+                    videoAnalysis = videoResult.analysis;
+                    log.info(`✅ Rox Video analysis complete for: ${videoFile.name}`);
+                } else {
+                    log.warn(`⚠️ Video analysis failed: ${videoResult.error || 'Unknown error'}`);
+                    videoAnalysis = null;
+                }
+            } catch (videoError) {
+                log.error(`❌ Video analysis error: ${videoError.message || 'Unknown'}`);
+                videoAnalysis = null;
+            }
+        }
+        
         // Prepare messages for the main LLM
         let messagesForApi = fittedMessages;
         
-        if (hasImages) {
+        // Handle video analysis context injection
+        if (hasVideos) {
+            if (videoAnalysis) {
+                log.info(`🎬 Step 2: Passing video data to ${config.name}...`);
+                
+                // Inject video analysis into the user message for the main LLM
+                const videoContext = `
+## 🎬 VIDEO ANALYSIS FROM YOUR VIDEO SIBLING (Rox Video)
+
+Your video sibling has analyzed the attached video and provided the following information:
+
+${videoAnalysis}
+
+---
+
+**USER'S ORIGINAL QUESTION:** ${message}
+
+**YOUR TASK:** Using the video analysis above from your video sibling, provide a helpful response to the user's question. You can reference the visual details, scenes, dialogue, and other elements your sibling identified. Remember, you're working as a team - your video sibling watches the video, and you provide the intelligent response!
+`;
+                
+                // Replace the last user message with the enhanced version
+                messagesForApi = fittedMessages.map((msg, idx) => {
+                    if (idx === fittedMessages.length - 1 && msg.role === 'user') {
+                        return { role: 'user', content: videoContext };
+                    }
+                    return msg;
+                });
+            } else {
+                // Video analysis failed
+                log.warn(`⚠️ Video analysis unavailable, sending text-only message to ${config.name}`);
+                const fallbackMessage = `${message}\n\n[Note: A video was attached but could not be analyzed. The video analysis service may be unavailable. Please ask the user to describe the video content if needed.]`;
+                
+                messagesForApi = fittedMessages.map((msg, idx) => {
+                    if (idx === fittedMessages.length - 1 && msg.role === 'user') {
+                        return { role: 'user', content: fallbackMessage };
+                    }
+                    return msg;
+                });
+            }
+        } else if (hasImages) {
             if (visionAnalysis) {
                 log.info(`🖼️ Step 2: Passing vision data to ${config.name}...`);