Everything you can customize in Taros
From widget design to AI behavior to multi-step forms – here's the full breakdown of what you can make your own.
One of the most common questions we get is "can I customize X?" The short answer is usually yes. Here's a full rundown of everything you can tweak, toggle, and personalize in Taros.
Widget appearance
The chat widget is the first thing your customers see, so it should look like it belongs on your site.
Colors and theming. Set a brand color, or use a gradient with separate start and end colors. Choose between light, dark, or auto theme mode. Pick from three visual styles: Classic (clean and familiar), Glass (frosted with blur effects), or Flat (minimal).
Button. Choose from 8 built-in icons or use your own custom image. Adjust size (48–80px), shape (circle, rounded, square), and color. Add an optional text label with configurable position and styling.
Position and layout. Place the widget in any corner. Fine-tune with pixel-level X and Y offsets. Set the z-index for stacking. Run it in bubble or inline mode.
Animations. Pick an idle animation (bounce, pulse, shake, rotate, and more) and an entrance animation (fade, slide, zoom, bounce, flip). Configure duration and delay for each. Add shadow effects with full control over color, blur, and offset.
Branding. Set your company name, choose an avatar (mascot image, custom image, or text initials), and write a welcome message.
AI model and behavior
You're not locked into one AI provider or personality.
Model selection. Choose between Mistral (all plans), OpenAI, or Anthropic (Pro and up). Each provider offers multiple model sizes. Set a fallback model in case the primary is unavailable. Adjust temperature for more creative or more precise responses.
Behavior templates. Start with a preset – Customer Support, Sales Assistant, FAQ Bot – or go fully custom. Each template generates a system prompt you can edit directly.
Conversation rules. Control whether the bot allows follow-up questions, how it cites sources, whether it can answer without knowledge base context, and how it identifies itself. Add custom instructions for anything else.
Fallback handling. Set what happens when the bot doesn't know the answer: custom fallback messages, escalation to a human with trigger keywords, or a specific no-knowledge message.
Image processing. Enable OCR so the bot can read text from images users upload.
Training and data sources
Multiple ways to give your assistant knowledge.
Website crawl. Point it at a URL and crawl a single page or your entire site. Re-crawl anytime to keep content fresh.
Document uploads. Upload PDFs, Word docs, and other files. Available on Basic and Pro.
Manual training. Add question-answer pairs directly. Great for FAQs and edge cases the crawler might miss.
Source management. Each data source is independent – create it once, assign it to multiple assistants. Set priority weights to control which sources the AI favors. Toggle sources active or inactive without deleting them.
Interaction flows and forms
Build multi-step forms that run inside the chat widget.
Field types. Text, email, number, phone, textarea, dropdown, radio buttons, checkboxes, date pickers, and hidden fields. Each with validation options like regex patterns, min/max length, and required/optional.
Conditional logic. Show or skip steps based on previous answers. Branch into different paths using conditions like equals, contains, greater than, and more.
Templates. Start from scratch or use a built-in template: Email Collection, Lead Qualification, Sales Assistant, or Tier-Based Routing.
Lead notifications. Send form submissions to an email address – immediately or as a daily digest. Pick which fields to include.
Webhooks. Fire webhooks when a flow completes, a step completes, or a specific branch is taken. Full audit log of every webhook call.
Multi-language. Translate any flow step into 19 languages with automatic fallback.
Deployment
Multiple ways to put your assistant in front of customers.
Website widget. One line of code. Works on any site.
Shopify. Native integration with automatic installation.
REST API. Full API access with per-bot API keys. Create, label, and revoke keys from the dashboard.
Domain restrictions. On paid plans, whitelist which domains can embed your widget. Supports wildcard patterns.
Analytics
Understand how your assistant is performing.
Session insights. Automatic extraction of categories, intents, topics, sentiment, resolution status, purchase signals, and unanswered questions from every conversation.
Daily aggregation. Total sessions, resolution rates, peak hours, top products mentioned, and intent breakdowns.
Chat logs. Full conversation history with search. Mark messages as reviewed. See which sources were used and how reliable the context was.
Message ratings. Users can upvote or downvote responses so you can spot quality issues.
Team and organization
Roles. Invite team members with role-based access.
Audit logs. Track every config change, bot enable/disable, member add/remove, and API key action. Includes actor, timestamp, and IP.
Session retention. Configure how long conversation data is stored (default: 90 days).
The point
You shouldn't have to compromise on how your AI assistant looks, sounds, or behaves. Whether you're a non-technical founder using the dashboard or a developer building through the API, everything in Taros is built to be yours to configure.
Not sure where to start? The defaults are sensible. Customize as much or as little as you want.
