Skip to main content
botBrains Vision transforms you AI agent from a text-only chatbot into a multimodal support system that can see and understand images. When customers share screenshots, product photos, error messages, or visual issues, your AI can analyze the content and provide intelligent, context-aware assistance - just like a human support agent would.

​
Why Vision Matters

Visual communication is essential for effective customer support. Many issues are easier to show than describe, and images provide context that text alone cannot capture. Vision enables your AI to: Resolve Issues Faster - Customers can share screenshots of error messages, product defects, or installation problems instead of trying to describe them in words. Your AI sees exactly what they see. Reduce Customer Effort - Instead of asking customers to transcribe error codes, describe visual problems, or navigate complex menus to find information, let them simply take a photo. Handle Complex Support Scenarios - Troubleshoot visual issues like UI problems, physical product defects, receipt verification, installation guidance, and document processing. Improve First-Contact Resolution - With visual context, your AI can diagnose problems more accurately and provide solutions without multiple back-and-forth exchanges. Scale Visual Support - Provide the same level of visual analysis your best support agents offer, without hiring more staff.
Customers using image upload resolve issues 40% faster on average compared to text-only conversations. Visual context eliminates ambiguity and speeds up problem diagnosis.

​
How Vision Works

When a customer uploads an image during a conversation, botBrains processes it through a sophisticated multimodal AI system:

​
1. Image Upload

Customers can upload images directly in the chat widget (if enabled). The system accepts common image formats and validates them for security and compatibility.

​
2. Secure Processing

Uploaded images are:
  • Scanned for security threats and malicious content
  • Validated for format and file size
  • Stored securely with encryption at rest
  • Processed using privacy-preserving techniques

​
3. Visual Understanding

The AI analyzes the image using advanced computer vision capabilities:
  • Text extraction: Reads error messages, labels, serial numbers, and any visible text
  • Object recognition: Identifies products, components, UI elements, and physical objects
  • Scene understanding: Comprehends context like settings, environments, and visual relationships
  • Problem detection: Recognizes errors, defects, damage, or incorrect configurations

​
4. Contextual Response

The AI synthesizes visual information with:
  • Conversation history and context
  • Your knowledge base and documentation
  • Customer information and past interactions
  • Guidance instructions for handling specific scenarios

​
5. Intelligent Action

Based on visual analysis, the AI can:
  • Answer questions about what it sees
  • Provide troubleshooting steps for visible errors
  • Guide customers through processes using visual references
  • Escalate to human agents when needed, with image context preserved
Vision uses the same multimodal AI models that power your text conversations, ensuring consistent intelligence and brand voice across all interactions.

​
Supported Image Formats

botBrains Vision supports all common image formats:

​
Standard Formats

  • JPEG (.jpg, .jpeg) - Most common photograph format
  • PNG (.png) - Screenshots and graphics with transparency
  • WebP (.webp) - Modern efficient format

​
Advanced Formats

  • HEIC (.heic) - iPhone and modern iOS camera format
  • HEIF (.heif) - High-efficiency image format
  • GIF (.gif) - Animated or static images

​
File Size and Quality

  • Maximum file size: Determined by your project settings (typically 10-25 MB)
  • Automatic optimization: Large images are compressed to balance quality and processing speed
  • Resolution: Higher resolution images provide more detail but take longer to process
  • Best practice: 1920x1080 or lower for screenshots; original resolution for detailed product photos
For screenshots, use PNG format for crisp text readability. For photographs, JPEG provides the best balance of quality and file size.

​
Common Use Cases

Scenario: Customer receives a damaged product or notices a manufacturing defect.How Vision Helps:
  • Customer photographs the defect
  • AI analyzes the image to assess severity and type of damage
  • AI determines if it’s a manufacturing defect, shipping damage, or normal wear
  • AI can initiate return, replacement, or refund workflows automatically
Example Interaction:
Customer: "The corner of my laptop is dented"
[Uploads photo showing dent]

AI: "I can see the dent on the bottom-left corner of your laptop. This
appears to be shipping damage rather than a manufacturing defect. I'll
process a replacement shipment right away. You can keep the current
device or return it - whichever you prefer. Your replacement will ship
within 24 hours."
Scenario: Customer encounters an error message they don’t understand.How Vision Helps:
  • Customer shares screenshot of the error
  • AI reads error codes, messages, and visible context
  • AI matches error to known issues in documentation
  • AI provides specific troubleshooting steps
Example Interaction:
Customer: "I'm getting an error when I try to log in"
[Uploads screenshot of error: "Connection timeout - Error 504"]

AI: "The error code 504 indicates a temporary server timeout. This
usually resolves within a few minutes. Try these steps:
1. Wait 2-3 minutes and try again
2. Clear your browser cache
3. Try a different browser
If the issue persists after 10 minutes, our servers may be experiencing
higher-than-normal load. I can notify you when service is restored."
Scenario: Customer needs to submit receipts, invoices, or proof of purchase.How Vision Helps:
  • Customer uploads receipt photo
  • AI extracts purchase details, amounts, dates, and item information
  • AI verifies against order history
  • AI processes warranty claims, returns, or reimbursements
Example Interaction:
Customer: "I need to return this item but I lost the order number"
[Uploads photo of receipt]

AI: "I found your order from the receipt. You purchased the Premium
Widget on March 15, 2024 for $89.99 (Order #ORD-12345). You're within
the 30-day return window. I've created a prepaid return label and
emailed it to you. Once we receive the item, your refund will process
within 5-7 business days."
Scenario: Customer installing hardware or software needs verification they did it correctly.How Vision Helps:
  • Customer shares photo of their setup
  • AI verifies correct installation, configuration, or assembly
  • AI identifies mistakes or missing steps
  • AI provides corrective instructions with visual references
Example Interaction:
Customer: "I think I installed the router correctly but internet isn't working"
[Uploads photo of router with cables]

AI: "I can see your router setup. The issue is that the WAN cable (blue)
is plugged into LAN port 2 instead of the WAN port. Here's what to do:
1. Unplug the blue cable from LAN port 2
2. Plug it into the WAN port (the one separated from the others, usually yellow)
3. Wait 30 seconds for the router to restart
The WAN light should turn solid green when connected properly."
Scenario: Customer experiencing visual or UI problems with software.How Vision Helps:
  • Customer screenshots the problem
  • AI identifies UI elements, states, and potential causes
  • AI guides customer through fixes or workarounds
  • AI can escalate with full visual context if needed
Example Interaction:
Customer: "The upload button disappeared"
[Uploads screenshot showing file manager interface]

AI: "I can see the issue - you're in 'View Only' mode. The upload button
is hidden when you don't have edit permissions. To fix this:
1. Click on your username in the top-right corner
2. Select 'Switch to Edit Mode'
3. The upload button will appear in the top toolbar
If you don't see 'Switch to Edit Mode', you may need to request editor
access from your team administrator."
Scenario: Customer needs support but doesn’t know their product model or version.How Vision Helps:
  • Customer photographs serial number sticker or product label
  • AI reads model number, serial number, and manufacturing information
  • AI looks up specific documentation and support for that exact model
  • AI provides personalized troubleshooting
Example Interaction:
Customer: "How do I update the firmware?"
[Uploads photo of product label showing model]

AI: "You have Model X-2000-B (serial: SN-12345). This model requires
the 2.4 firmware update process:
1. Download firmware v2.4.1 for the X-2000-B series: [link]
2. Connect via USB (not WiFi for this model)
3. Run the firmware installer
Note: The X-2000-B has different update steps than the X-2000-A, so
generic instructions won't work. Follow the model-specific guide here: [link]"

​
Enabling Vision

Vision capabilities are controlled at the frame (widget) level, allowing you to enable or disable image upload for different deployments.

​
Enable Image Upload in Frames

  1. Navigate to Frames in your botBrains project
  2. Select the frame you want to configure
  3. Go to the General Settings tab
  4. Toggle Allow Image Upload to enabled
  5. Save your frame configuration
Changes to frame settings take effect immediately. Customers will see the image upload button in the chat widget after you save.

​
Configure in Guidance

While image upload is a frame-level setting, you can add guidance instructions to help your AI understand and respond to specific types of images: 
When a customer uploads an image of a product defect:
1. Analyze the image carefully to identify the type and severity of the defect
2. Determine if it's a manufacturing defect, shipping damage, or user damage
3. For manufacturing defects: offer immediate replacement
4. For shipping damage: apologize and process replacement with expedited shipping
5. For user damage: explain it's not covered by warranty but offer repair options
6. Always be empathetic and focused on resolving the customer's problem

​
Best Practices

​
When to Use Vision

  • Error messages and screenshots
  • Product defects or damage
  • Installation verification
  • Document and receipt processing
  • Serial number identification
  • UI/UX troubleshooting
  • Visual configuration issues
  • Physical setup verification

​
Security and Privacy

​
Image Security

All uploaded images are:
  • Scanned on upload for malware, viruses, and malicious content
  • Validated for format to prevent exploit attempts
  • Encrypted at rest using AES-256 encryption
  • Transmitted securely over HTTPS/TLS
  • Access-controlled with role-based permissions

​
Safety Measures

Content Moderation:
  • Automatic detection of graphic violence and sexual content
  • Suspected CSAM (Child Sexual Abuse Material) reported to German authorities (LKA NRW) per German law
  • Team protection from inappropriate material
  • Reporting mechanisms for policy violations
  • Filters to prevent processing of harmful content

​
Data Privacy

Storage and Retention
  • Images are stored in secure, encrypted cloud storage in Frankfurt, Germany
  • Access is logged and auditable
  • Retention policies are not configurable, infinite by default. Subject to change.
Processing Privacy
  • Images are processed by AI models without human review (unless escalated)
  • No image data is used to train AI models
  • Processing is ephemeral - images are analyzed in-memory and not cached
  • Compliance with GDPR, CCPA, and other privacy regulations
Customer Rights Customers can:
  • Request deletion of their uploaded images
  • Access copies of images they’ve uploaded
Images uploaded by customers may contain sensitive information. Implement appropriate guidance to prevent collection of regulated data types (PHI, PII, payment card data) unless your organization has knowledge of the necessary compliance.

​
Frequently Asked Questions

Check frame configuration:
  1. Navigate to Frames > General Settings
  2. Verify “Allow Image Upload” is enabled
  3. Save and refresh the chat widget
Check file size limits:
  • Ensure images are under maximum file size
  • Try compressing large images before upload
Check browser compatibility:
  • Image upload requires modern browsers (Chrome 90+, Safari 14+, Firefox 88+)
  • Mobile browsers may have different file size limits
Improve image quality:
  • Use higher resolution for detailed analysis
  • Ensure good lighting for photographs
  • Avoid blurry or low-quality images
Provide context in text:
Instead of: [uploads image]
Better: "Here's a screenshot of the error I'm seeing" [uploads image]
Use appropriate image types:
  • Screenshots for digital errors (PNG format)
  • Photographs for physical products (JPEG format)
  • Avoid highly compressed or low-resolution images
Model limitations:
  • Some images may be too complex or ambiguous
  • Very small text may be difficult to read
  • Extreme lighting conditions can affect recognition
Fallback to human support: Configure guidance to escalate when:
If you cannot confidently analyze the image or the issue requires
human judgment, say: "I'm having trouble analyzing this image clearly.
Let me connect you with a specialist who can review this in detail."
Then use `escalate_to_human` with the image attached.

​
Examples

Customer uploads photo of damaged box
Customer: "My order arrived damaged"
[Uploads photo showing crushed shipping box]

AI: "I'm sorry your order arrived in that condition. I can see significant
damage to the shipping box. Let's get this resolved:

Based on the damage visible in your photo, I'm processing:
✓ Full refund: $156.78 (credited within 3-5 business days)
✓ Replacement shipment: Expedited 2-day delivery at no charge
✓ Return label: No need to return the damaged item

Your replacement order is #ORD-67890. You'll receive tracking within
2 hours. Is there anything else I can help with?"
Customer uploads screenshot of software error
Customer: "Getting this error when I export"
[Uploads screenshot: "Export failed: Invalid format specified (ERR-2045)"]

AI: "I can see error code ERR-2045 in your screenshot. This happens when
the export format doesn't match your file type. Looking at your screenshot,
you're trying to export a video file (.mp4) as an audio-only format (.mp3).

Here's how to fix it:
1. In the Export dialog, change 'Format' from 'MP3' to 'MP4'
2. Or, if you want just the audio, check 'Extract Audio' first
3. Then select MP3 format

The option you need is in the dropdown menu visible in your screenshot,
just to the right of where the error appeared. Try that and let me know
if you still have issues!"
Customer uploads photo of partially assembled furniture
Customer: "Is this right? The instructions are confusing"
[Uploads photo of desk with parts laid out]

AI: "Looking at your photo, you're on the right track! You have all the
parts correctly identified and the frame pieces are in the right position.

However, I notice one issue: the support bracket (the L-shaped metal piece
on the left) needs to be rotated 90 degrees. The long side should face up,
not forward.

Once you rotate that bracket:
1. Attach it to the frame using the 4 long screws (visible in your photo)
2. Then attach the desktop (the large piece in the background)
3. Finally, add the drawer assembly

You're actually further along than the instructions show - good work!
Need any other guidance?"
Now that you understand Vision, explore related features:
  • Guidance - Configure how your AI responds to visual queries
  • Actions - Enable tools for processing visual information
  • Frames - Configure image upload in your chat widget
  • Knowledge - Combine visual analysis with documentation
  • Conversations - Review how Vision is used in customer interactions

Enable Image Upload

Configure Vision capabilities in your chat widget settings