# Nectir AI Product Updates Source: https://support.nectir.io/changelog/product-updates Release notes for Nectir AI: new features, bug fixes, and LMS integration updates, with a short summary of what changed in each release. * Fixed removing a Workspace member reporting success but not refreshing the members list or revoking access until the cache expired minutes later * Fixed AI-generated chat titles occasionally appearing in the wrong language * Fixed Canvas deep links opening a different Assistant than the one the instructor selected during embed setup ## AI Builder The AI Builder is a new guided experience for creating and editing Assistants. Describe what you want your Assistant to do, answer a few questions, and the builder generates a complete configuration (prompt, conversation starters, model, and tools) ready to save. A Configure tab lets you fine-tune every setting, including knowledge scope, before publishing. The builder is available from both the create and edit Assistant pages. AI Builder * Fixed the artifact panel not refreshing when the Assistant edits a file that's already open, causing stale content until a new file was created ## Voice Dictation and Read Aloud A microphone button in the chat composer lets you speak your message instead of typing it. Speech streams into the input as it's recognized, and you can read, edit, and append more before sending. A few seconds of silence ends dictation automatically. On Assistant replies, a read-aloud button speaks the answer out loud and can be stopped at any time. Both work in English, Portuguese, and Spanish with automatic language detection (no configuration needed). Voice dictation and read aloud are available everywhere, including embedded chats launched from your LMS. Voice Dictation and Read Aloud * The expired trial modal now shows a **Switch Workspace** button when the user belongs to another active Workspace, instead of only showing "Contact sales" * Corrected the Workspace analytics export dialog to accurately describe what the export contains, removing the incorrect "counts and aggregates only" language * Fixed "Something went wrong" errors in chat caused by malformed tool calls stored in conversation history * Fixed tall or wide images (for example, 25×5000) failing Knowledge Base ingestion * Fixed BMP images failing Knowledge Base ingestion and converted HEIC/HEIF files keeping their original filename instead of the processed JPEG name * Fixed artifact creation stacking multiple error cards and rendering placeholder artifacts when generation fails, instead of showing a single failure state * Fixed transient Canvas API errors (rate limits, 5xx) killing the entire course content sync and incorrectly telling instructors to reconnect their Canvas account * Fixed concurrent Canvas content imports racing on token refresh, causing valid tokens to be invalidated and imports to fail as "token revoked" * Fixed Assistant prompt and settings saves being slow and intermittently failing with "An unexpected response was received from the server" ## Language Selection The Nectir AI interface is now available in Portuguese and Spanish. A language selector in **Account Settings** → **Preferences** lets users switch between English, Portuguese and Spanish, and the app remembers the choice across sessions and devices. Language Selection *** * Fixed agent chat crashing with "Something went wrong" when running knowledge search alongside multiple parallel sub-agents * Fixed agents rejecting messages after 200-message conversations by replacing the hard message cap with automatic context compaction ## Sharper Answers on Dense Material Assistants now reach further into your Knowledge Base, following references across files and pinpointing the right section of a long document. Questions like "explain chapter 5" hold up better on heavy or dense material. ## Artifacts You Can See and Download Assistants can now render what they make right next to the conversation, so you can ask for a diagram or a table and actually look at it while you keep chatting. When you want a copy to keep or share, a persistent download button in the artifact panel header gives you a Word document, PDF, or CSV spreadsheet. Downloadable Artifacts ## A clearer Chain of Thought The reasoning view now shows each step the Assistant took, including the tools it used and the sources it checked, in one collapsible panel. When an answer takes a few steps to put together, you can expand it and follow the whole path. *** * Assistants now use the latest model versions, with Claude Sonnet 5 set as the default for new Assistants * Responses now address users directly instead of occasionally referring to them in the third person * Canvas import now prompts users to reconnect when required API scopes change, instead of silently operating with outdated permissions * The invite link confirmation text now matches the role assigned to the link * Fixed the mobile header back button being unreachable behind the hamburger menu on Assistant settings, Group settings, and Assistant analytics * Fixed long URLs in chat messages breaking text flow instead of wrapping at URL separators * Fixed the Android keyboard hiding the chat composer on empty conversations * Fixed loading spinners appearing at the top of the page instead of vertically centered * Fixed the sidebar actions button showing a double focus outline on keyboard navigation * Fixed the "Auto" model label being clipped to "Au…" in Safari ## Workspace Analytics Redesign The Workspace Analytics dashboard has been redesigned with a new two-tab layout splitting usage details into separate **Groups** and **Assistants** views. Each row now displays its place in the hierarchy via breadcrumbs (for example, "Department of Sciences > Biology"), and standalone Assistants are labeled. The redesign includes consolidated filters, improved accessibility with full keyboard navigation and ARIA support, better dark mode contrast, and responsive horizontal scrolling on smaller screens. Workspace Analytics Redesign ## Canvas Sync Error Recovery When a Canvas automatic sync fails because the connection has expired, the sync settings banner now explains the issue and offers a **Refresh connection** button that opens the Canvas reconnect flow directly. Once reconnected, automatic sync resumes on its previous schedule without extra steps. Temporary failures (network issues, rate limits) continue to retry automatically with a softer message. Canvas Sync Error Recovery *** * Fixed stale Assistants appearing when entering a Workspace, caused by cached Assistant state redirecting users to Assistants no longer visible in the sidebar * Fixed Canvas content imports failing with transient fetch errors by classifying failure types and retrying HTTP/2 connection resets automatically ## Individual Canvas Item Selection When importing content from Canvas, instructors can now select individual items within each category instead of importing entire categories at once. Expand any category (assignments, pages, quizzes, discussions) to check or uncheck specific items. Fully selected categories continue to sync new content from Canvas automatically, while narrowed selections sync only the chosen items. The category row displays the current sync status so instructors can see at a glance which categories are fully synced and which are narrowed. ## Canvas Syllabus, Calendar Events, and Due Dates Canvas content sync now includes syllabus pages, calendar events, and assignment due dates, giving Assistants richer course context for answering student questions Individual Canvas Item Selection *** * Fixed Assistants occasionally referring to users in the third person (for example, "I'll build Anna a study plan") instead of addressing them directly ## Knowledge Table Improvements The Knowledge table in Assistant Settings has been redesigned with new file management tools: a search bar and filters to find files faster, in-browser preview for PDFs, images, and text files, per-file source visibility controls replacing the global toggle, in-place file replacement, and bulk actions for managing multiple files at once. Knowledge Table Improvements *** * Fixed beta feature toggles not appearing for users who declined optional cookies, by resolving feature flags server-side independently of cookie consent * Fixed uploaded knowledge files displaying URL-encoded characters in filenames (for example, `%20` instead of spaces) * Facilitators can now view knowledge without hitting a permission error, and action buttons or menus that Editors and Facilitators can't use are now correctly hidden or disabled * Assistant and Workspace names now enforce character limits * Fixed Google account profile photos not displaying correctly * Fixed older conversations showing a "Something went wrong" error when the chat history contained image attachments * Fixed the delete confirmation modal breaking layout when the Group or Assistant name is long * Fixed knowledge item text overflowing its container in the source document list * Fixed LTI integration table displaying an incorrect Client ID after registration * Now we can see what the Assistant is doing in real time: tool invocation notifications are now displayed in the chat interface during streaming * Fixed emoji avatars not being copied when duplicating an Assistant or Group ## Member Suggestions When Inviting to a Group When inviting members to a Group, the email field now suggests existing Workspace members as you type, making it faster to add people without remembering exact email addresses. Member Suggestions When Inviting to a Group ## Original URLs in Website Knowledge Citations Chat citations for website knowledge sources now display the original page URL, and title instead of the internal markdown filename, making it easier to identify and verify source material. Original URLs in Website Knowledge Citations ## LTI Waiting Screen for Students When a student launches an LTI link before the instructor has finished setting up the course integration, they now see a dedicated waiting page explaining the course isn't connected yet, instead of being redirected to an empty dashboard. LTI Waiting Screen for Students *** * Added a "Back" button and the clickable Assistant name to the header of the "Analytics" page in embedded chats, following the navigation pattern already used on the "Settings" page * Improved disabled-state contrast on dropdown menu items by replacing opacity-based styling with design system color tokens for better accessibility in both light and dark themes * Fixed the Queue Manager showing "0/0" and loading indefinitely when updating a website knowledge source to fewer pages than it previously had * Fixed a redirect loop causing 500 errors for users whose sessions lacked an email address * Fixed scroll overscroll revealing a transparent background behind the File Drawer * Fixed LTI-accepted members being stuck on the "No invites found" screen due to a stale Workspace * The "Documents searched" and "No results found" indicators no longer appear when an Assistant's knowledge base is empty, preventing confusing search feedback on Assistants that have no documents to search * Fixed PDF export failing in Safari * Fixed small images imported from Canvas failing during knowledge base processing * Fixed older conversations intermittently showing a "Something went wrong" error when the chat history contained image attachments * Fixed missing spacing between the icon and label on the "Refresh connection" button in the Canvas import dialog ## Math Equation Editor A new "Math equation" editor is now available in the chat input via the `+` button menu. Students and instructors can compose mathematical expressions visually using a point-and-click editor or toggle to direct LaTeX input for advanced use. The composed expression is previewed in real time and inserted into the chat as rendered LaTeX, making it easier to communicate math notation without knowing LaTeX syntax. Math Equation Editor *** * Standardized the Queue Manager component to behave consistently across all knowledge sources (manual uploads, Canvas imports, and IMSCC files), with unified progress tracking, collapsible course sections, and automatic error surfacing * IMSCC file imports now organize extracted files by course name in the Queue Manager, matching the collapsible course section pattern used for Canvas imports * Restored the copyable LTI configuration JSON in the Canvas guided setup flow, pre-populated with the customer's Workspace ID and updated to include deep-linking placements * Fixed tables appearing broken and unformatted in PDF chat exports * Fixed Mermaid diagrams briefly flashing a "Mermaid Error" during streaming before rendering correctly * Fixed non-math quiz content being unnecessarily wrapped in LaTeX formatting ## Accessibility Improvements Shipped a round of accessibility improvements across the UI, focused on three areas: * **Keyboard and assistive navigation:** Added descriptive text and new-tab indicators to external links, made truncated content available through keyboard-accessible tooltips, and added programmatic names to Help submenu links. * **Screen reader support:** Added accessible names to avatar zoom sliders and conversation history links, associated labels with the rows-per-page dropdown, and replaced static aria-labels on the Analytics date range picker with context-aware labels. * **Visual contrast:** Improved border and icon contrast on buttons and input components for better readability. See the full list below. *** * Updated prompt library audience categories, replacing "Content creator" with "Staff" to better align with how institutions describe their roles * Fixed user theme preference reverting to the previous value on navigation or when multiple browser tabs are open * Added programmatic names to Help submenu links * Added descriptive external link text and new-tab indicators on external links * Replaced static aria-labels on the Analytics date range picker with context-aware labels * Associated labels with the rows-per-page dropdown * Added accessible names to avatar zoom sliders * Added accessible text to conversation history links * Made truncated text available through keyboard-accessible tooltips * Improved border and icon contrast on buttons and input components * Fixed sporadic "We Couldn't Connect to Nectir AI" errors during LTI launches caused by transient HTTP/2 connection resets when fetching platform keys ## Reorder Conversation Starters Conversation starters can now be reordered using drag-and-drop in Assistant Settings, giving owners full control over the order prompts appear to students. Reorder Conversation Starters *** * Fixed root-level static assets, causing broken images in Canvas and browser console errors * Fixed PWA manifest not loading correctly, and resolved a Chrome deprecation warning for the mobile-web-app-capable meta tag * Fixed LaTeX expressions not rendering in interactive quiz questions and answer choices, displaying raw syntax instead of formatted math * Accessibility fixes: resolved multiple accessibility issues across the UI, including insufficient contrast on dropdown trigger chevrons that failed WCAG 2.2 non-text contrast requirements in both light and dark themes * Fixed website URL imports showing a successful page count in the UI even when the back-end crawl workflow failed, leaving the knowledge base empty with no error surfaced to the user * Fixed PDF export failing for conversations containing inline code wrapped in backticks * Fixed support text not being centered in the Add Website dialog, matching the alignment used in other import modals * Accessibility fixes: resolved multiple accessibility issues across the UI, including insufficient focus indicator visibility on checkboxes and sidebar links, incorrect focus order after bulk selection actions, content truncation in role dropdowns at 200% zoom, incorrect heading hierarchy on LMS integration and import pages, missing focus management after adding LMS roles, incorrect focus landing position when navigating to Account Settings, missing ARIA value text and insufficient disabled-state contrast on the temperature slider, text spacing truncation in the History dialog and Analytics filters, and missing heading tags on Analytics stat cards * LTI integrations can now be renamed directly from the integrations table using a new actions menu, making it easier to differentiate between multiple integrations * Standardized back button labels, icons, and tooltips across the app for a more consistent navigation experience * Canvas "Student View" now displays a dedicated informational page explaining why it isn't supported, instead of showing a generic error * Toggle badge labels across the platform now display "On" and "Off" instead of "Enabled" and "Disabled" for a cleaner, more compact UI * Website URL imports are now limited to 1,000 pages * Fixed drag-and-drop file icon appearing stacked and illegible when the "Disable Animations" preference is enabled * Fixed long URLs overflowing and getting cut off in the delete website source confirmation modal * Fixed "something went wrong" error when sending messages in certain older Assistant conversation histories * Accessibility fixes: resolved multiple accessibility issues across the UI, including incorrect tab focus on non-interactive elements, focus not returning to the triggering control when closing dialogs, and the Copy LaTeX button not being reachable via keyboard navigation ## Reorder Assistants You can now reorder Assistants using Move Up and Move Down options in the actions menu. The custom order persists across sessions, giving owners full control over how Assistants are arranged within a Group or Workspace. Reorder Assistants ## Refresh Canvas Connection A new "Refresh connection" button in the Canvas import dialog allows instructors to re-authorize their Canvas connection and pick up newly supported content types without manual reconnection. Refresh Canvas Connection *** * The Quiz tool is now a default Assistant tool, ensuring reliable activation and consistent interactive rendering across all sessions * Canvas file imports now automatically retry after transient API failures, reducing the need to manually re-import * Optimized the Workspace analytics page to batch-load data in a single request, eliminating rate limiting for Workspaces with many private Assistants * Improved platform reliability by separating infrastructure health checks from AI model availability, preventing transient LLM provider outages from affecting non-AI features * Fixed Assistant tools being silently removed when saving Assistant or Workspace AI settings * Fixed "something went wrong" error when sending messages in older Assistant conversations * Fixed rate limit error appearing when navigating to conversation histories older than 30 days * Accessibility fixes: resolved multiple accessibility issues across the UI, including incorrect ARIA roles on tabs, comboboxes, and filter dropdowns, missing screen reader announcements for search results and status changes, broken keyboard navigation in listboxes and chart components, and layout reflow issues at 400% zoom * Fixed large image uploads in chat failing when the file exceeded the AI provider's size limit by adding progressive compression that automatically reduces quality and dimensions * Accessibility fixes Resolved multiple accessibility issues across the UI, including incorrect ARIA patterns, missing labels, keyboard navigation gaps, and a button in the emoji picker search * Changed the default automatic sync frequency for Canvas course imports from daily to weekly ## Add From Website URL You can now add website content directly to your Assistant's knowledge base by entering a URL. Nectir scans the site for a sitemap, displays the discovered pages, and lets you select which ones to import. Imported websites appear as a single entry in the Knowledge table, grouped by root domain. Once added, website sources can be fully managed from the Knowledge table. To keep content up to date, enable **Sync weekly** when adding a website or from the configuration panel. Nectir checks for changes on a weekly cadence and refreshes affected pages in the background. Manual sync remains available at any time. Add from Website URL ## Automatic Sync for Canvas Course Content Canvas course content imported into your Assistant's knowledge base can now stay up to date automatically. Toggle automatic sync per course and choose a daily or weekly sync frequency. Nectir checks for changes on schedule, adding new content, updating modified items, and removing deleted ones to keep your knowledge base in sync with Canvas. Automatic Sync for Canvas Course Content ## Add From Canvas You can now import course content from Canvas directly into your Assistant's knowledge base, with no file exports or downloads needed. When you connect for the first time, a security modal confirms that Nectir uses read-only access, stores your data securely, and gives you full control over what gets imported. Once authorized, browse your courses, select which content to import (files, pages, assignments, discussions, announcements, and quizzes), and it all appears in the Knowledge table grouped by course, ready to manage. Add Knowledge Modal
Canvas Quiz Import ## Canvas API Integration Workspace owners can now connect their Canvas instance to Nectir AI via the API. Create a Developer Key in Canvas, enter your credentials in Workspace Settings, and instructors in your Workspace can start importing course content directly into their Assistants. Canvas Data Security Modal ## Copy as LaTeX A new "Copy LaTeX" button now appears when hovering over rendered math expressions in AI responses. Clicking it copies the original LaTeX source to the clipboard, preserving structural formatting like fractions, exponents, and roots. This lets you paste equations into Canvas Equation Editor, LaTeX documents, or other tools that support LaTeX input. Copy as LaTeX ## Multi-URL LTI Integration Support The Canvas LTI setup now supports multiple instance URLs per integration, allowing institutions with multiple Canvas instances (for example, main campus and satellite campuses) to register all URLs in a single integration. The integrations table displays all connected URLs with a badge, and admins can edit existing LTI integrations directly from the table, adding or removing instance URLs without re-creating the integration. Multi-URL LTI Integration Support *** * Assistants now have access to a built-in code interpreter that verifies calculations before presenting answers. This addresses a known limitation of LLMs making arithmetic errors, particularly in generated practice questions where incorrect answers, duplicate options, or unsolvable equations could undermine trust. The tool validates math for algebra through pre-calculus level, ensuring generated problems produce correct, verifiable answers * User theme preference (light/dark mode) is now persisted across sessions and devices, so the app loads with your last selected theme on subsequent visits * Fixed Assistant responses showing "something went wrong" error after completing a valid response, caused by a streaming timeout in the AI pipeline * Simplified the "Unlink LMS integration" confirmation modal with clearer copy and consistent checkbox font styling * Polished Canvas LTI setup dialog with simplified configuration steps, automatic URL normalization, and embedded Workspace ID * Updated the JSON configuration for Canvas LTI manual setup to use the correct format and automatically embed the Workspace ID * Fixed LTI launch failing for students when Group ownership was determined by role mapping, causing certain users to be incorrectly blocked * Strengthened Content Security Policy * Improved protection against common browser-based attacks * Enhanced access controls across protected endpoints * Improved session timeout and session invalidation policies * Implementation of additional HTTP security headers * Implementation of rate limiting on authentication and resource creation endpoints * When unlinking an LTI integration, you can now opt to also remove course members: members only in that Group are removed from the Workspace, while members in other Groups are removed from the Group only * Improved dark mode text contrast across multiple UI components to meet WCAG AA accessibility standards * Invite email placeholder message now dynamically shows the Group or Workspace name based on invite context * PDF chat exports now display "user" instead of "student" as the role label and include the user ID * Improved success notification for better contrast with green background in dark mode * Fixed analytics dashboard chart y-axis not displaying values above 999 correctly * Fixed LTI integration setup screen not appearing for non-Instructor LMS roles (for example, Designer) that are mapped to Group Owner in Workspace role settings * Fixed visual delay when closing source attribution results in chat * Canvas LTI setup with fewer manual steps: the Workspace ID is now embedded in the JSON configuration, and the flow is reduced to two clear steps * Added HEIC image format support, fixing image uploads from iPad Safari when using the Photos app * The unsaved changes confirmation modal now triggers when navigating away via sidebar items such as Assistant, Workspace, and Group Settings * PDF chat exports now display the model name when "Auto" is selected (for example, Auto (GPT-5.2)) * Updated "Free trial expired" messaging for improved clarity * Fixed file type icons not being vertically centered with multi-line filenames in the Assistant Knowledge table * Fixed selected model text color being difficult to read in the model selector, especially in light mode ## Workspace-Level Assistant Tool Settings Workspace owners can now manage Assistant tool settings at the Workspace level. Tools like the Web Search, Quiz Generator, and Nectir Support can be enabled or disabled for all Assistants across the entire Workspace, so you can roll out tools without configuring each Assistant individually. All new Assistants are created with the previously defined tools enabled. Manage the tools in Workspace Settings → AI controls. Workspace-Level Assistant Tool Settings ## Emoji Picker for Assistant Avatars You can now choose an emoji as your Assistant's avatar. A new emoji picker is available in Assistant Settings, giving you a fun and expressive way to personalize your Assistants. Emoji Picker for Assistant Avatars *** * "Show sources" is now enabled by default for new Assistants * The unsaved changes confirmation modal now appears consistently across all navigation scenarios in Assistant, Workspace, and Group Settings, including browser back, sidebar clicks, tab switches, and page refresh * PDF chat exports now include additional metadata such as chat title, Assistant name, Group name, Workspace name, AI model, and student email * Increased minimum icon size across the platform from 12×12 pixels to 16×16 pixels for improved visual clarity * Updated Quiz Generator description to "Let students practice with interactive quizzes" * Renamed "Workspace" to "Workspace invitations" in personal settings for clearer navigation * Fixed "Cancel" and "Save changes" buttons being active without any changes on the Workspace Settings General page * Fixed "Theme" label using incorrect font size in user preferences * Fixed LTI connection failing with "Learning Platform not Recognized" error when the LMS enforces HTTP/2 only mode * Fixed LTI 1.3 integration failing when the same deployment key is installed across multiple Canvas sub-accounts ## Auto Model Selection A new "Auto" model option has been added that always points to the latest recommended model. New Assistants use "Auto" by default, so they automatically get model upgrades. Existing Assistants keep their currently assigned model, and users who manually select a specific model stay on it until that model is deprecated. Auto Model Selection *** * Updated the Nectir platform description displayed to LMS administrators during LTI integration setup to accurately reflect Nectir AI's capabilities * Improved Canvas manual integration setup with automatic scrolling between configuration steps and consistent UI components * Standardized "Cancel" button label across all settings pages for a more consistent experience * Fixed table scrolling not working at certain screen resolutions, causing an infinite loading spinner * Fixed issue where removing an LMS integration and re-adding it with the same client ID would not display in the integrations table * Fixed subgroups breaking out of their Group hierarchy when duplicating a Group * Fixed middle dot character breaking LaTeX rendering in chat messages ## Account Settings Redesign We've redesigned the account settings experience with a new full-page layout. Settings are now organized into three tabs: Profile, Preferences, and Beta features. The profile dropdown has been simplified to show only Settings and Sign out, with all configuration options moved to the dedicated settings page. Account Settings Redesign *** * Improved word wrapping behavior in the chat text input area * Fixed sidebar automatic scrolling issue * Fixed subgroups being visible to parent Group members even when "Sync parent Group members" was turned off * Fixed text alignment in the cookie preferences modal ## Interactive Quiz Feature Assistants can now generate and grade interactive quizzes directly in chat. Students can request quizzes on topics they're learning, answer questions step-by-step, and receive AI-powered grading with feedback. Supported question types include multiple choice, true/false, multiple select, fill-in-the-blank, and short answer. Enable this feature in Assistant Settings → Advanced → Tools. Interactive Quiz Feature ## Nectir Support Center Tool A new "Nectir Support" tool is now available for Assistants. When enabled, the Assistant can search Nectir's official support documentation to answer platform questions without users leaving their conversation. Enable this feature in Assistant Settings → Advanced → Tools. Nectir Support Center Tool ## Redesigned LMS Integration Setup We've redesigned the LMS integration experience in Workspace Settings. The new LMS Integrations table displays all connected integrations with details including instance name, type, platform, client ID, and URL. Workspace owners can now add new integrations through a guided setup flow and remove existing integrations directly from the table. The Canvas manual integration setup now includes a step-by-step wizard with progress indicators. Quick access links to the Support Center and Trust Center are also available for security and compliance information. LMS Integrations Table *** * Chat now displays personalized greetings with the user's name, and the AI can address users by name in responses * Improved LTI error page with friendlier messaging, clearer error descriptions, and actionable guidance * Added confirmation dialog when clicking external links in AI responses for improved security * Fixed date range picker in analytics dashboard * We added a 'Show More' option for Assistant descriptions. This option opens in a modal window, making it easier to read and improving the experience with long descriptions * We improved document ingestion in the Assistant Knowledge * AI prompts now enforce proper LaTeX wrapping for consistent mathematical rendering across all models, as well as improved rendering for complex expressions (matrices, equations, etc.) * Fixed math equations in PDFs being misread (for example, fractions displaying incorrectly) * Fixed sidebar automatic scrolling not working when accessing Assistants via LTI integration ## Assistant Identity Badge The chat interface now displays the Assistant's name and avatar prominently above the greeting message, making it easier for students to confirm they're using the correct Assistant. Assistant Identity Badge ## LTI Error Page When LTI launches fail, users now see a dedicated error page with user-friendly messages and error codes, making it easier to troubleshoot issues. LTI Error Page *** * Line breaks in Assistant descriptions are now preserved when displayed in the Assistant information modal * Fixed infinite loop that could occur when uploading documents to Assistant knowledge for the first time * Fixed file size displaying as "NaN undefined" for certain uploaded files * Fixed image uploads failing in chat conversations - Gemini and Claude * Fixed reasoning UI not displaying when using GPT-5.2 model * Fixed issue where certain AI models could generate garbled or unexpected text responses ## GPT-5.2 Model Support We've added support for OpenAI's GPT-5.2-chat model, giving you another powerful AI option for your Assistants. New Assistants are now created with GPT-5.2 by default, and the new model is available in Assistant Settings → Advanced alongside GPT-4.1, Claude Sonnet 4.5, and Gemini 2.5 Pro. GPT-5.2 Model Support ## Reasoning UI We've also introduced a new reasoning UI that displays the AI's thinking process when using models that support it. Reasoning UI *** * Updated cookie consent modal to use "Product usage data" instead of "analytics" to avoid confusion with the Analytics feature * Added proper spacing between buttons in the cookie consent modal on mobile * AI SDK update to version 6 to improve performance and reliability * Fixed training failures when uploading large images to Assistant knowledge * Fixed chat input border misalignment when uploading documents ## Redesigned Chat Experience We've made significant improvements to the chat interface with new animations and micro-interactions throughout the experience. The update includes redesigned file attachments, a typewriter effect for AI-generated conversation titles, improved drag-and-drop styling, and a new visual design for tool calls. Files are now automatically uploaded before message submission for a smoother workflow. Redesigned Chat Experience ## File Upload Queue Manager We've introduced a new Queue Manager component for file uploads in Assistant Knowledge. Instead of individual toast notifications for each file, uploads are now consolidated into a single view showing real-time status for all files—in progress, completed, or failed. The queue runs in the background even when minimized, and if you attempt to close it while uploads are still processing, a confirmation modal appears. Failed uploads display clear error reasons, and clicking on completed files navigates directly to their entry in the Knowledge table. File Upload Queue Manager *** * Knowledge table now displays only successfully processed files, with status tracking moved to the Queue Manager * Added browser confirmation dialog when navigating away while file uploads are in progress * Workspace Analytics and Web Search features are now available to all users ## Cookie Consent Management We've added cookie consent controls to provide transparency and control over which cookies are used on Nectir AI. On your first visit, a banner will ask for your consent to use cookies. You can change your preference anytime through the new **Cookie Preferences** page, accessible via the link in the footer: Account settings > Cookie preferences Cookie Consent Management * Renamed "Invite Members" to "Members" in Group dropdown menu for clearer navigation ## Parent Group Member Inheritance Subgroups can now automatically inherit members from their parent Group. When enabled, members added to a parent Group are automatically synced to all subgroups. You can view inherited members in a dedicated "Parent Group members" tab, toggle inheritance on or off, and manage membership at the parent level. Inherited members' roles can only be changed from the parent Group, ensuring consistent permissions across your Group hierarchy. Parent Group Member Inheritance *** * Fixed "Show Sources" and "Web Search" settings not being preserved when duplicating a Group * Fixed error when sending large messages to Assistants * Fixed AI responses being cut off prematurely when generating long answers * Fixed issue preventing Group creation in certain cases * Fixed Workspace avatar not displaying correctly after removing an uploaded image * Fixed issue preventing bulk duplication of Groups and Assistants * Simplified Help menu by consolidating support options into a single "Support Center" link * Added "Submit feedback" option to the Help menu for bug reports and feature requests * "Unlink from LMS" option is now hidden in member role dropdown when no LMS integrations are configured * Fixed source preview not loading for certain file formats (audio, video, `.tiff` images, `.eml`, `.md` files) ## LMS Role Management Improvements We've made significant improvements to how user roles are managed with LMS integrations: * **Unlink roles from LMS**: Instructors can now unlink a user's role from LMS sync, preventing automatic role changes during future syncs. Unlinked roles display an "Unlinked" badge and can be re-linked at any time. Unlink roles from LMS * **Automatic member removal**: Users removed from the LMS course roster are now automatically removed from the corresponding Nectir AI Group during roster sync. * **Role sync during roster sync**: LMS roster sync now updates user roles in addition to membership, ensuring roles stay consistent between Canvas and Nectir AI. *** * Conversation starter prompts now place the cursor at the end of the text * Updated invite email copy with improved subject line and support for line breaks in custom messages * Improved mobile layout on the "No invites found" screen * LTI ID copy fields now include the parameter prefix (for example, `workspaceId=`) for easier setup * Invite links no longer override existing member roles * Fixed deleted chat history still appearing until page refresh * Fixed LTI role changes in Canvas not updating user roles in Nectir AI * Users already in a Workspace can no longer accidentally create free trial Workspaces; the option is now only shown during initial onboarding ## Improved Empty State for New Members Workspace members who haven't been added to any Groups yet now see a helpful empty state screen instead of a blank page. The new screen explains what to expect, provides visual examples of what a populated Workspace looks like, and guides users on next steps. Improved Empty State for New Members *** * Large markdown tables in chat messages can now be scrolled horizontally * Simplified Assistant settings by removing the Semantic search toggle * We've optimized database operations for chat conversations, resulting in faster response times when interacting with Assistants * Improved LTI role mapping to correctly assign Teaching Assistant permissions instead of Instructor permissions * Fixed roster sync failing when users with non-allowed email domains are present (invalid emails are now skipped) * Fixed source attribution not displaying when web search is enabled * Fixed LMS roster sync member count incorrectly including deleted members * Fixed session timeout causing users to be signed out prematurely in Canvas ## AI Web Search Assistants can now search the web to find up-to-date information when answering questions. Enable this feature through the **Web search** toggle in Assistant Settings → Advanced → Tools. When web search is used, the sources (URLs) are displayed alongside training data sources, allowing users to verify the information. AI Web Search ## Unlink LMS Integration You can now unlink LMS integrations from Groups when they're no longer needed. In Group Settings → LMS Integration, click **Unlink** on any connected course to permanently disconnect it from Nectir. Group members remain unchanged after unlinking. Unlink LMS Integration Button Unlink LMS Integration *** * Improved AI response performance and reduced costs through prompt caching optimizations * Improved email delivery reliability during large roster syncs * Fixed Workspace analytics showing 0 uploaded documents ## Source Attribution Now Available for All Users The "Show Sources" feature is now available to all users. When enabled, Assistants display the training data sources used when generating responses, providing transparency and allowing students to verify the materials supporting AI-generated content. Enable this feature through the **Show sources** toggle in Assistant settings. Show Sources ## Roster Sync Email Controls You can now control invite email notifications during LMS roster sync. When syncing a roster, a modal allows you to include a custom message in invite emails or skip sending emails entirely. The user performing the sync no longer receives redundant invite notifications. Roster Sync Email Controls *** * Improved LTI role mapping to correctly distinguish between institution-level and course-level roles from Canvas * Fixed issue where member join date wasn't being recorded for users joining through invite links * Fixed issue where IMSCC files exported from Brightspace (D2L) weren't being recognized during LMS import * Fixed LTI integration issue for Workspaces configured with legacy slug format * We've made significant performance improvements through extensive caching optimizations. These changes reduce database queries and improve page load times across the platform * Fixed issue saving URLs in message metadata * Fixed early access feature toggle for Workspace analytics not working correctly * Students completing onboarding are no longer shown the "Create free trial Workspace" option, directing them to join existing Workspaces first * Fixed Assistant history not updating in real time after starting a new conversation * Fixed PDF export failing for certain Assistants due to missing italic font * Fixed "Nectir" spelling typo in modal * Improved email delivery reliability with upgraded email infrastructure * Fixed issue where the send button was disabled when only files were attached to a message * Fixed cache issue causing users to be logged out after completing onboarding ## Terms of Service Acceptance Modal Existing users who haven't accepted the Terms of Service will now see a blocking modal requiring acceptance before continuing to use the platform. This ensures compliance and keeps all users up to date with our terms. Terms of Service Acceptance Modal *** * New users are now directed to complete the onboarding form before accessing the platform * Added role-based flow customization for personalized user experiences * Redesigned the "No Invite Found" screen for better clarity * Fixed issue where Group invites didn't redirect users to the correct Workspace and Group after accepting ## Terms of Service Acceptance Checkbox New users must now accept our Terms of Service and Privacy Notice before creating an account. A required checkbox has been added above the email field. If a user tries to sign up with email, Google, or Microsoft without checking the box, an inline error message appears and account creation is blocked until acceptance. Terms of Service Acceptance Checkbox *** * Updated login and account creation page text for improved clarity ## Custom Messages in Invite Emails You can now add personalized messages when inviting members to a Group. When inviting members via email, a confirmation modal allows you to choose who receives the invite email and include a custom message that appears in the invitation. Custom Messages in Invite Emails *** * Added confirmation modal when manually inviting members, allowing you to control who receives email notifications * Added message timestamps to PDF chat exports for better conversation tracking * Fixed analytics dashboard header display on mobile devices * Fixed issue where some `.docx` files weren't being accepted in chat uploads * Fixed training pipeline failures for certain `.html` and `.xlsx` files * Fixed table rendering in PDF exports ## Code File Uploads in Chat You can now upload code and text files directly in your chat conversations with Assistants. This includes `.html`, `.txt`, `.sql`, and other code files, allowing for more dynamic interactions where you can ask questions about specific code snippets or have the Assistant analyze text content. *** * Added admin options menu with quick access to prompts library and other administrative features * Improved client-side caching for faster page loads and reduced unnecessary API requests * Fixed file downloads to properly preserve file extensions and download directly instead of opening in browser * Fixed performance issue causing 500 errors when accessing large Workspaces * Fixed actions menu not appearing for private Assistants created by Workspace editors * Fixed compatibility issue with older browsers that don't support the `.at()` method * Fixed error when the prompt library API returns unexpected data * Fixed code block rendering errors in chat messages * Fixed file preview button not appearing for `.sql` files uploaded in chat * Fixed code file preview display issues * Fixed issue where some IMSCC files from Canvas couldn't be extracted during LMS import due to invalid or corrupted media files ## Bulk Member Actions We've implemented improvements to the Members table interface, allowing multiple rows to be selected using checkboxes. This enables bulk actions including removing multiple members at once and changing roles for multiple users simultaneously, improving usability when managing large groups. Bulk member actions in Members table ## New File Pipeline We've launched a new pipeline for uploading files. This update optimizes our document ingestion process for improved reliability and performance. *** * Standardized colors for bot avatars and file type icons across the platform * Updated the sender name of emails to display as "Nectir" instead of "mail" for improved clarity and recipient recognition across all email clients * Implemented AI-generated conversation titles that automatically create concise, descriptive titles based on the first user message and AI response, helping students identify and revisit past discussions * Updated the onboarding role selection terminology by replacing "Teacher" with "Instructor" to better align with higher education contexts * Fixed issue of repeatedly requesting to connect to Nectir AI when navigating within Canvas iframe integration ## Workspace-Level AI Controls We've introduced a new workspace-level AI controls page that allows administrators to set global prompts that apply to all Assistants across the entire Workspace. This powerful feature provides centralized control over AI behavior while still allowing individual Assistant customization. Workspace AI controls *** * Enhanced prompt library with improved UI including text fields that expand as you type, better spacing, and accessibility improvements * Unified help icon component across the platform for consistent user experience * Streamlined first-time user access flow with improved Workspace selection and invitation management * Added post-login access to "Choose your Workspace" page through Workspace dropdown menu * Fixed access flow issues for new users who were getting stuck without Workspace creation options * Resolved double @ symbol display in allowed email domains section and fixed domain validation * Fixed issue where some users received an error when interacting with their own Assistant ## Prompt Library and Editing Enhancements We've launched a comprehensive prompt library alongside significant improvements to the prompt editing experience. Browse, search, and apply pre-built prompts filtered by activity and role, with the ability to preview and customize them before applying. The enhanced editor now includes undo/redo functionality, making it easier than ever to create and refine your Assistant prompts. Prompt library *** * Improved dialog scroll behavior to allow scrolling on the entire screen instead of just the content container * Fixed markdown list formatting issues in Assistant responses for better readability * Implemented infinite scroll for tables, replacing pagination for improved performance with large datasets * Fixed broken "LMS export guide" link on the import page to properly direct to help documentation * Disabled interaction with private and deleted Assistants in Workspace analytics dashboard * Added database performance optimizations with new indexes for improved chat query speed * Enhanced analytics export to include identifiable data when Workspace settings permit * Added workspace-level flag to control visibility of personal data in analytics * Improved CSV export compatibility with UTF-8 byte order mark support * Fixed onboarding survey styles not switching correctly in light theme mode * Resolved Workspace lock issue when trial period expires, allowing users to switch Workspaces * Improved PDF text extraction and error handling for better document readability ## IMSCC File Import General Availability The IMSCC file import feature is now available to all users. You can import course content directly from your learning management system without needing early access. Import from LMS button *** * Reorganized analytics components for better code structure and maintainability * Added workspace-level control for showing identifiable data in analytics and exports * Fixed LaTeX notation rendering in Assistant responses for mathematical content * Resolved file icon and format consistency issues across different components * Removed .html from plaintext file extensions list for proper categorization ## Source Attribution for AI Responses (Beta) Assistants can now display the sources used when generating responses from instructor-uploaded training data. This feature provides transparency and allows students to verify the materials supporting AI-generated content. Control source visibility through the new "Show sources" toggle in Assistant settings. Sources *** * Enhanced tool call display in chat UI for better transparency of AI operations * Improved color scale for active user percentage badges in analytics dashboard ## Workspace Analytics Dashboard We've launched a comprehensive analytics dashboard at the Workspace level, providing administrators with valuable insights into how Nectir AI is being used across their entire organization. This powerful new feature complements the existing Group-level analytics by offering a broader view of usage patterns, engagement metrics, and trends across all Courses and Assistants within your Workspace. Workspace analytics dashboard ## Gemini 2.5 Pro Model Support Expanding our AI model offerings, we've added Google's Gemini 2.5 Pro to the available model options. This gives you even more flexibility in choosing the AI model that best fits your educational needs, alongside our existing GPT-4.1 and Claude Sonnet 4 options. Gemini 2.5 Pro model ## Native iOS Sharing for Conversations iOS and iPadOS users can now share their conversation PDF exports using the native iOS share sheet. This feature makes it seamless for students to submit their Nectir AI conversations as assignments directly through apps like Canvas, improving the mobile experience for educational workflows. *** * Updated disclaimer text in Assistant chat conversations for improved clarity * Enhanced file upload component with multiple file support, better previews, and responsive design * Removed user avatars from chat messages for a cleaner, more focused interface * Made file upload cards fully keyboard-accessible and screen-reader friendly * Fixed issue where pasting content from CSV files (particularly from Apple Numbers) would paste as an image instead of text, preventing further conversation ## PDF Export for Conversations You can now export your chat conversations with Assistants as PDF files, providing a more readable and tamper-resistant format compared to CSV exports. This feature makes it easier to submit conversations to learning management systems and creates a cleaner, professional format for sharing AI interactions. PDF export option in conversation ## Enhanced Code File Support We've expanded the types of files that can be uploaded to an Assistant's knowledge base to include code files. You can now train your Assistants with source code including `.py`, `.java`, `.c`, `.cpp`, `.h`, `.js`, `.html`, `.css`, `.sql`, and `.R` files, enabling better support for computer science and programming courses. ## Analytics Improvements We've enhanced the Analytics dashboard with new export capabilities and smarter data visualization. You can now export analytics data from Group and Workspace level Assistants. The dashboard also features an improved default time range of 3 months and intelligent trendline displays that only appear when sufficient historical data exists, eliminating misleading percentage changes. *** * Relocated knowledge scope selector from the Knowledge tab to the Prompt tab for better workflow organization * Reorganized supported file type display in upload interface: Documents → Plaintext → Images → Audio → Video → Code * Resolved component import error that was causing page rendering issues in development environment * Corrected file format display from `.r` to `.R` for proper R language file recognition ## Document Upload in Chat You can now upload PDF, DOCX, and TXT files directly in your chat conversations with Assistants. This allows for more dynamic interactions where you can ask questions about specific documents, get summaries, or have the Assistant analyze content without needing to add files to the Assistant's knowledge base. *** * Added ability to rename training data files in the Assistant's knowledge base * Fixed sidebar spacing issues with avatar icons and three-dot menus for improved visual consistency * Corrected keyboard navigation highlight colors and border radius throughout the interface * Resolved Assistant settings menu incorrectly showing for users without proper permissions * Fixed regenerate response feature duplicating messages and causing feedback to be saved to wrong messages ## Group Nesting You can now create nested Groups to organize your Assistants in hierarchical structures. This powerful feature enables you to create Groups inside other Groups, making it easier to manage complex organizational structures, departments, or multi-level course hierarchies. You can invite members at any Group level with permissions properly inherited throughout the hierarchy. Nested groups ## AI Model Updates We've upgraded our AI models to the latest versions for improved performance and capabilities. All Assistants have been automatically migrated to GPT-4.1 (from GPT-4o) and Claude Sonnet 4 (from Claude Sonnet 3.5). These models will providing better reasoning, enhanced context understanding, and more accurate responses. AI model dropdown ## Temperature Control for AI Models Assistant Owners now have fine-grained control over AI response creativity through customizable temperature settings via the Advanced tab. The AI model temperature allows you to adjust how creative or predicable the response is. AI model temperature slider *** * Improved key topics analysis with smarter consolidation that combines similar topics, reducing hundreds of narrow categories into meaningful insights * Updated Facilitator role permissions to properly restrict editing capabilities while maintaining view access to analytics and settings * Fixed text capitalization inconsistencies throughout the platform for improved readability * Removed "Import from LMS" option for Facilitators and properly restricted their access to view-only capabilities ## LMS Role Mapping Workspace owners can now configure how LMS roles (like Instructor, Learner, Teaching Assistant) are mapped to Nectir AI roles when users are provisioned via LMS integration. This ensures users receive appropriate permissions based on their institutional roles. LMS role mapping ## Duplicate Assistants You can now duplicate Assistants, just like courses. When duplicating, all settings and training data will be copied to the new Assistant. Duplicate Assistant dropdown ## "Courses" Are Now "Groups" We've updated the interface terminology from "Courses" to "Groups" to better reflect diverse educational use cases and simplify the organization of AI Assistants. ## Early Access Features Introduced a new feature flag functionality that enables customers to opt into early access features. You can now access experimental and beta features by clicking on your profile icon and selecting **Early access features**. * Redesigned Workspace invite email for improved user experience * Redesigned Course invite email with enhanced visual design * Redesigned "Continue with email" login email template * Added help article links to "Import from LMS" flow for better user guidance * Resolved security vulnerabilities CVE-2024-21538, CVE-2025-30218, and CVE-2025-26519 ## Iframe Support for LMS Integration LMS integration via LTI 1.3 now works with iframe embedding, allowing Nectir AI to be embedded directly within Canvas, Moodle, D2L, and Blackboard. Learn more by checking out our [blog post](https://www.nectir.io/blog/introducing-lms-embed-access-nectir-ai-assistants-right-within-your-lms-with-lti). LMS iframe ## Link Multiple LMS Courses to a Single Nectir AI Course You can now link multiple LMS courses to a single Nectir AI course. This is useful if you want to use the same Assistant across multiple LMS courses. Link to course *** * Redesigned LMS integration page: renamed from "Integrations" to "LMS Integration" and consolidated all integrations into a single table * Added new columns (Status, LMS type) to the LMS integrations table * Created default Nectir AI Assistant for all new Workspaces * Users are now redirected to Intercom when contacting support * Improved internal analytics implementation by removing unused SDKs * Fixed issue when logging into LTI iframe where the page breaks * Fixed type error when sorting a table * Increased maximum file upload size from 100 MB to 150 MB for training files * Modified LMS roster sync to exclude users with Observer role from receiving Workspace invitations * Optimized invitation process to prevent sending redundant notifications to users already in the Workspace ## Assistant Training Enhancements * **IMSCC support (limited rollout)**: Added ability to import and train on IMSCC files from learning management systems * **Media file support**: Enabled uploading of audio and video files for Assistant training * **Increased file size limit**: Maximum file size for training files increased from 25 MB to 100 MB ## Role Improvements * **Facilitator role update**: Facilitators can now view analytics dashboard and Course/Assistant settings, but can't edit settings *** * Fixed issue with duplicate messages being sent to AI ## AI Model Selection Instructors can now choose between GPT-4o (default) and Claude 3.5 Sonnet v2 for their Assistants via the Advanced tab in Assistant settings. For a complete overview, check out our [blog post](https://www.nectir.io/blog/nectir-ai-is-now-supporting-claude-and-chatgpt-models). * Fixed issue where Course owners couldn't duplicate their own courses. Now they can duplicate their own Course regardless of their Workspace role * Fixed issue where Course duplication option wasn't showing at LTI integration first step ## Key Topics Analytics We've enhanced our analytics capabilities with a new key topics feature that helps you understand what students are discussing with your Assistants. For a complete overview, check out our [blog post](https://www.nectir.io/blog/introducing-key-topics-unlocking-deeper-insights-into-student-ai-interactions-nectir-ai-assistants). * **Topic analysis**: View the top topics discussed in conversations for specific Assistants * **Sortable data**: Sort topics in ascending or descending order * **Pagination**: Navigate through all topics with 5 rows per page * **Filter options**: Apply role and date filters to refine topic analysis Key topics analytics * Fixed issue where you couldn't copy or paste text in the chat input ## UI/UX Improvements * **Image handling improvements**: Added drag-and-drop and copy-paste functionality for images in chat * **Course duplication via LTI**: Users can now duplicate Courses through LTI interface * **Improved analytics dashboard filters**: Preselected all user roles by default and expanded time range to show data from Assistant creation to present day * Added evaluation framework to the document training pipeline to enable comparison of different solutions and methods * Fixed issue when duplicating a Course with Assistants that have training files—embeddings are now properly copied ## Trial Experience Improvements * **Trial countdown**: Added card displaying the remaining time in free trial period * **Trial management**: Implemented logic to restrict access to certain resources when free trial ends * **Eligibility check**: Added account verification to check free trial eligibility *** * Added support for multiple LTI integrations with the same client ID * Changed URLs to use IDs rather than slugs while maintaining backward compatibility * Created basic structure and implemented first unitary tests * Fixed issues with the Assistant options dialog * Fixed stop button being disabled when generating responses * Restricted chat uploads to only supported image formats ## LMS Integration Improvements * **Course linking**: Added ability to link LTI courses to existing Nectir AI courses *** * Rearranged sign up/in modal with sign up column moved to the right * Upgraded PostHog AI library to better track LLM usage and calculate costs * Fixed highlighting issue for same-named courses in the combo box * Fixed CSV export when conversations contain the "#" character * Fixed issue with linking LTI courses to the correct Workspace * Fixed redirect to wrong Assistant in a different Workspace * Fixed integration name display when linking a Course from LTI * Fixed various UI typos ## Self-Service Improvements * **Redesigned sign-in/up pages**: Enhanced user experience for account creation and login * **Workspace invite redesign**: Improved process for inviting users to Workspaces * **Trial workflow**: Added first screen for trial workflow - create Workspace ## System Enhancements * **Default sharing permissions**: Changed the default invite link access to "Anyone with the link" instead of "Restricted" * **AI analytics**: Now storing metadata of messages sent to AI for improved analytics *** * Upgraded numerous dependency libraries * Standardized user email storage to always save in lowercase * Fixed accessibility issues for the menu dialog * Added automated testing to the document training pipeline in the CI/CD process * Fixed issues causing .docx and .html files to fail during training * Enhanced training notifications to appear to users even if they leave the training page * Fixed dialog bug that prevented page interaction after dialogs were closed * Fixed issue with retrieving user information after session expiration * Fixed error when processing PPTX files in the training pipeline * Implemented code improvements for the document training pipeline ## Free Trial Now Available We've launched a trial page so you can now sign up and try Nectir AI free for 14 days. Check it out [here](https://ai.nectir.io/trial/signin). *** * Refactored navigation within the Conversation Assistant page for better UI/UX * Changed CSV delimiter to comma for exported conversations, improving compatibility * Fixed issue of losing typed text when opening History/Information panels * Fixed sidebar display in PWA format * Improved management system for upgrading third-party libraries ## Progressive Web App Support Nectir AI can now be downloaded and installed as an app on your mobile phone, providing a more native experience when using the platform on mobile devices. ## AI Enhancements * **Date awareness**: Added current date to AI prompt context (first Feature Flag implementation) * **Transparency**: Introduced an AI awareness modal and footer note on the conversations page *** * Enhanced test pipeline by adding more tests and aligning test environments with production * Improved UX/UI for the delete course confirmation modal * Added timestamp tracking for when users accept Workspace invites * Fixed issue where Course owners who aren't Workspace owners couldn't delete their own courses ## Course Duplication We've introduced a new feature that lets you effortlessly duplicate Courses, simplifying the process of copying all your Assistants along with their settings and training data across different sessions or terms. *** * Enhanced logging for server-side operations * Added more automated tests for platform stability * Fixed page not found error for members on a Workspace without any Assistants * Fixed AI message duplicating and non-responsive behavior when sending images * Fixed losing message context when sending first message to an Assistant * Fixed Campus AI message box rendering issues * Fixed error when adding LTI tool via dynamic registration * Fixed database transaction issues that could allow data errors * Fixed various accessibility issues * Fixed typos throughout the platform ## Nectir AI Major Release We're excited to announce a big release with significant new features and improvements across the platform. For a complete overview, check out our [blog post](https://www.nectir.io/blog/nectir-ai-release-v1-0-0-empowering-educators-with-advanced-features). ## Analytics Dashboard Track and measure Assistant performance at both Workspace and Course levels with our new comprehensive analytics dashboard. Our new analytics dashboard provides real-time insights into AI Assistant usage, helping you make informed decisions about your teaching strategies. Analytics dashboard ## Improved LMS Integration * **LTI 1.3 roster sync**: Route multiple LTI course rosters to single Assistant * **LTI 1.3 dynamic registration**: Streamline the LMS setup process with dynamic registration for Canvas, Moodle, and others ## Enhanced User Experience for Assistants * **Image support**: Upload and send images in conversations with any Assistant * **Markdown support**: Text input in Assistant conversations is now formatted as markdown * **Improved navigation**: Better navigability throughout Assistant pages * **Training data download**: Added download button for Assistant's training files ## Security & administration * **Email domain restriction**: New Workspace setting allow only specific email domains to join Workspaces * **Security warnings**: Added warning when sharing public invite links * **Infrastructure update**: Completed migration to new training documents service Azure Document Intelligence * **Default settings**: Changed default Assistants knowledge to general knowledge *** * Enhanced accessibility for multiple selection component * Added automated unit tests for functionality and accessibility * Resolved complex cache issue with invite links * Fixed restriction on Assistants with the same name * Fixed multiple Sentry errors related to semantic HTML and accessibility # Advanced Features Source: https://support.nectir.io/creating/advanced Configure advanced Assistant settings in Nectir AI, including AI model selection, temperature, web search, quizzes, and the Nectir support tool. AI model selection interface showing different model options The model picker interface may change as providers and model options evolve over time. ## Configuring AI Models Nectir gives you control over which AI model powers your Assistant. For most users, the default setting is the best choice. ### Auto (Recommended) New Assistants default to **Auto**, which always uses the current recommended model. When Nectir updates or adds models, Auto keeps your Assistant on the best available option without any action from you. **Use Auto if you:** * Want your Assistant to always use the latest recommended model * Don't have a specific reason to choose a different model * Manage more than one Assistant and don't want to track model changes across them ### Choosing a Specific Model You can also select a specific AI model from the dropdown. Nectir offers models from different providers, and the available options may change over time as providers release new models. **You might choose a specific model if you:** * Want consistent behavior from a particular model you've already tested * Are comparing how different models respond to the same prompts * Have a preference based on your own experience with a specific provider If you select a specific model and it's later removed from the platform, your Assistant will automatically switch to the current recommended model. ### Selecting an AI Model 1. Navigate to your Assistant's settings 2. Click **Advanced** in the settings sidebar 3. Under **AI Model**, choose **Auto** (recommended) or select a specific model from the dropdown 4. Click **Save Changes** ### Building AI Literacy Exposing students to different AI models is a practical way to build AI literacy. Because Nectir offers models from different providers, you can encourage students to: * Compare how different models respond to the same question * Notice differences in tone, depth, and reasoning style across models * Think critically about which model handles certain types of tasks best This kind of hands-on comparison helps students develop judgment about AI tools they'll encounter beyond your classroom. ## AI Model Temperature Temperature controls how creative or focused your AI Assistant's responses are. Most Assistants work well with the default setting, and some models don't support temperature at all. The Temperature control always appears in Advanced settings, but it may be disabled for models that don't support it. Lower Temperature for more consistent, grounded responses (useful for fact-heavy subjects). Raise it for brainstorming or creative work where varied ideas are helpful. ## Web Search Web search allows your Assistant to search the internet in real-time to answer questions that require current information, recent events, or data beyond the Assistant's knowledge base. | Web Search Enables ✓ | Web Search Isn't ✗ | | - | - | | Real-time internet searches during conversations | URL training or web scraping to add to your Assistant's knowledge base | | Access to current information and recent events | A replacement for uploading course materials | | Answers about topics not covered in your uploaded documents | Available when the knowledge scope is Document Only | | Up-to-date statistics, news, and data | Permanent knowledge that persists across conversations | **Important**: Web search doesn't run when your Assistant's knowledge scope is "Document Only," even with the **Web Search** toggle on. With "General Knowledge" or "Topic Knowledge," the Assistant searches the internet when needed. Enable web search for Assistants that need to: * Answer questions about current events or recent developments * Provide up-to-date statistics, prices, or data * Research topics not covered in your course materials * Verify or supplement information with external sources * Help students explore topics beyond your uploaded content "How do recent archaeological discoveries about Roman engineering relate to what we learned about their social structure in Chapter 3?" "Compare today's unemployment data with the economic theories we studied. Which model best explains current trends?" "Analyze recent CRISPR breakthroughs using the ethical framework from our readings. What new considerations arise?" "How do current market conditions for tech companies align with Porter's Five Forces model we discussed?" Consider disabling web search if: * Your course should focus on uploaded materials * You want to prevent students from relying on external sources * Your subject matter doesn't require current information * You're concerned about external information conflicting with course content When you enable web search, students interact with the Assistant normally. The Assistant automatically determines when a web search would be helpful and performs it in the background. **Student Question**: "How does this week's Supreme Court decision on environmental law relate to the regulatory frameworks we studied in Chapter 5?" | With Web Search ✓ | Without Web Search ✗ | | - | - | | "Based on today's Supreme Court ruling, here's how it connects to the Clean Air Act framework we studied:

The Court's decision fundamentally changes the 'major questions doctrine' we discussed, specifically affecting the EPA's authority under Section 111(d) that you analyzed in your readings. Let me break down three key implications for the regulatory approaches in your textbook…" | "I can discuss the regulatory frameworks from Chapter 5, but I don't have access to recent court decisions. I recommend searching for the latest Supreme Court rulings on environmental law to see how they might relate to what we've studied." | **Student Question**: "Can you help me apply the supply chain concepts from our course to understand what's happening with semiconductor shortages right now?" | With Web Search ✓ | Without Web Search ✗ | | - | - | | "Looking at current semiconductor supply chain data, let's apply the bullwhip effect and lean manufacturing concepts from Week 3:

Current lead times are 26 weeks (up from the 12 weeks we discussed in class), which demonstrates the exact inventory challenges described in your readings. The just-in-time approach you studied has these specific weaknesses in today's situation…" | "I can explain the supply chain concepts from your course materials, but I don't have access to current semiconductor industry data. You might want to research recent articles about chip shortages to apply these concepts." | **Student Question**: "We learned about Nash equilibrium in game theory. Can you analyze a recent business decision using that framework?" | With Web Search ✓ | Without Web Search ✗ | | - | - | | "Let's analyze Microsoft's recent AI pricing strategy announced this week using Nash equilibrium from your Chapter 7:

Microsoft set prices assuming Google would respond with similar moves - this is a classic prisoner's dilemma scenario. Based on today's market data, we can see how both companies are arriving at an equilibrium where neither can unilaterally improve their position, exactly like the model predicts…" | "I can explain Nash equilibrium and game theory from your course materials, but I don't have information about recent business decisions. Try finding a recent business case study to analyze using the frameworks we've covered." | New Assistants start with **Web Search** on, unless a Workspace Owner changed the default under **Workspace Settings** > **AI Controls** > **Default Assistant Tools**. Web search toggle in Assistant advanced settings Go to your Assistant's settings and click on **Advanced** In the **Tools** section, toggle on **Web Search** and click **Save Changes** Your Assistant can now perform web searches when answering questions When your Assistant uses web search, it always shows the sources it referenced, regardless of whether you have the [sources feature](/creating/sources) enabled for uploaded knowledge files. Web search sources displayed in Assistant response **Sources are automatic for web search**: Unlike uploaded knowledge files where you can toggle sources on or off, web search results always include the links the AI referenced. This helps students verify the information and understand where the data comes from. Web search works best when combined with your course materials. Your Assistant will reference uploaded documents first, then use web search to supplement with current information when needed. Tell students that the Assistant can access current information online. This helps them understand the Assistant's capabilities and encourages them to ask questions about recent events or data. Enable web search for subjects that benefit from current information (sciences, business, current events) but consider disabling it for subjects focused solely on historical texts or specific course materials. **Why this is more powerful than Google search**: Unlike having students search Google themselves, your Assistant with web search can synthesize current information with your course materials, applying your specific frameworks and theories to real-world events. Students get contextualized analysis rather than just raw data, helping them understand how to apply what they're learning to current situations. ## Quiz Quiz allows your Assistant to generate and grade interactive quizzes directly within the chat conversation. Students can request quizzes on topics they're learning, answer them interactively, and receive AI-powered grading and feedback without leaving the conversation. Quiz is available on every Assistant. It has no toggle in **Advanced** settings and can't be turned off. | Quiz Enables ✓ | Quiz Isn't ✗ | | - | - | | AI-generated practice questions within chat conversations | A replacement for your LMS quiz or test tools | | Instant AI-powered grading with explanatory feedback | Integrated with your LMS grade book | | Five question types for diverse practice | Designed for high-stakes summative assessment | | Student-driven, self-paced knowledge checks | Controlled by instructor like traditional quizzes | | One-time submission for each quiz (1-20 questions per quiz) | Allows retakes | **Important**: Quiz is designed as a formative learning tool for student self-assessment and practice, not as a replacement for formal graded assessments. Think of it as an interactive study partner that helps students test their understanding and receive immediate feedback during their learning process. Quiz fits Assistants that support: * Formative assessment and knowledge checks * Student self-testing and exam preparation * Practice reinforcement after reading or lectures * Interactive learning with immediate feedback * Self-paced study and concept mastery Student: "Can you quiz me on cellular respiration with 5 multiple choice questions to help me prepare for tomorrow's exam?" Student: "Give me a short quiz about the causes of World War I so I can test what I learned from this week's readings." Student: "Test my understanding of Python loops with some fill-in-the-blank and short answer questions." Student: "Create a quiz on the themes in Chapter 3 of The Great Gatsby to check my comprehension." Students request quizzes in conversation; no particular syntax is required. They ask the Assistant to generate a quiz on a specific topic, specify the number of questions (1-20), and optionally request particular question types. The Assistant generates the quiz, presents questions, and provides AI-powered grading with explanatory feedback after submission. **Student Request**: "Quiz me on photosynthesis with 3 multiple choice questions" **Quiz Experience**: The Assistant generates questions with labeled options. Students select their answers, submit the completed quiz, and receive immediate grading with explanations for correct and incorrect answers. **Feedback Example**: * Question 1: Correct ✓ - Explanation of why the answer is right * Question 2: Incorrect ✗ - Explanation of the correct answer and why their choice was wrong * Question 3: Correct ✓ - Additional context connecting to course concepts **Student Request**: "Give me questions about SQL commands where multiple answers might be correct" **Quiz Experience**: The Assistant creates "select all that apply" questions. Students can choose multiple options for each question. Grading evaluates whether they selected all correct answers and avoided incorrect ones. **Feedback Example**: * Partial credit for selecting some but not all correct answers * Explanations for each option (why it's correct or incorrect) * Guidance on concepts that need review **Student Request**: "Test my understanding of Newton's laws, and use a mix of fill-in-the-blank and short answer questions" **Quiz Experience**: Students type their answers directly. For fill-in-the-blank, they provide specific terms or values. For short answer, they write brief explanations or analyses. **Feedback Example**: * AI evaluates answers for accuracy and completeness * Provides corrections for factual errors * Suggests ways to strengthen explanations * Highlights key concepts that were missing **Quiz as a learning tool**: Quiz is designed as a conversation-based study partner for formative learning. Students can take a quiz, review feedback, study more, then test themselves again in a new conversation. This active learning cycle helps them identify knowledge gaps, reinforce concepts, and build confidence before exams. You can also customize quiz behavior in your Assistant's prompt, for example by specifying preferred question types, number of questions per quiz, or topics to focus on. *** ## Nectir Support Tool The Nectir Support tool lets your Assistant search [Nectir's support documentation](https://support.nectir.io) to answer questions about the platform in real-time. Your Assistant always accesses the latest support documentation—no manual updates needed. **What it's for:** * Real-time access to all Nectir support documentation * Answers questions about platform features, setup, and best practices * Seamless in-conversation help without leaving the chat **What it's not:** * A replacement for human support (complex issues, bugs, account problems) * For answering your course-specific content questions Enable this tool when you want your Assistant to help users learn how to use the Nectir platform itself. **Recommended for:** * Institutions with many first-time Nectir users * Assistants designed for general student support * Pilot programs where users are still learning the platform * Any Assistant where users might ask questions about how Nectir works **May not need it:** * Highly specialized Assistants where users are already platform-savvy * Mature deployments where support questions are rare New Assistants start with **Nectir Support** on, unless a Workspace Owner changed the default under **Workspace Settings** > **AI Controls** > **Default Assistant Tools**. As a Workspace Owner or an Owner of the Assistant's Group, go to the Assistant's **Settings** page. Select the **Advanced** tab, then scroll to the **Tools** section. Toggle on **Nectir Support** and save your changes. Your Assistant can now search Nectir's support documentation when users ask platform-related questions. * Enable this tool on general-purpose Assistants in environments with new Nectir users * This tool adds Nectir platform knowledge on top of your Assistant's normal capabilities. It doesn't limit the Assistant to only answering platform questions For complex issues, account problems, or bugs, direct users to [support@nectir.io](mailto:support@nectir.io). *** ## Video Tutorial Watch how to select AI models and configure temperature settings # Add training data to your AI Assistant Source: https://support.nectir.io/creating/assistant-knowledge Upload and manage training data to customize your Assistant's knowledge base. Assistant knowledge settings interface The knowledge you upload is the **foundation** of your Assistant's ability to provide **personalized, relevant responses**. This content customizes your AI Assistant to understand your specific course materials, institutional policies, and unique program requirements. More importantly, it creates a **tailored learning experience** that directly supports each student's **individual learning journey**, making your Assistant truly **supportive** rather than generic. Hover over your Assistant, click the **three dots**, and select **Settings** Click on the **Knowledge** tab in the sidebar Select your preferred method to add knowledge to your Assistant: Upload files from your computer Import a course export (.imscc) from LMS Import content from a website URL Connect to import course files Sources show which knowledge files your Assistant references in responses. Every file is shown as a source by default. To hide a file, open its menu (three dots) and click **Hide Source**. Configure your source preferences ## What to Upload Upload your syllabus, course materials, assignments, and any content specific to your course or program. Include information unique to your campus, policies, procedures, and institutional guidelines. Add documents containing concepts, terminology, or jargon specific to your field or program. Include dates, deadlines, schedules, and any time-sensitive information your Assistant should know. **Important:** Focus on content that AI models won't have in their knowledge base. This includes your specific course materials, institutional policies, and unique program requirements. ## Supported Knowledge Sources Any public website URL (text content is extracted from selected pages) .pdf .docx .xlsx .pptx .txt .text .eml .md .xml .jpg .jpeg .png .heif .heic .tiff .bmp .mp3 .wav .flac .aac .m4a .opus .mp4 .webm .mpeg .mpg .mov .wmv .py .java .c .cpp .h .js .css .html .sql .R File Limits: You can upload **unlimited files** with a maximum size of **150 MB per file**. ## Understanding the Knowledge Table The Knowledge table displays only files that are ready for use. Each file listed has completed processing successfully and is available for Assistants to reference during conversations. If a file appears in the Knowledge table, it's usable. Uploading a file begins a background processing step that prepares the content for use. While processing is underway: * The file doesn't appear in the Knowledge table * Progress and results are shown in the Upload Queue Once processing completes successfully, the file is automatically added to the Knowledge table. The Knowledge table intentionally excludes: * Files that are still processing * Files that failed during processing * Files with upload errors These files remain visible in the Upload Queue to provide clear status information. Files can be uploaded to multiple Assistants at the same time. When uploading to multiple assistants: * Each Assistant processes the file independently * Status appears per the Assistant in the Upload Queue * A file may succeed for one Assistant and fail for another This allows shared content while maintaining clear visibility. ## Using the Upload Queue to Track File Status The Upload Queue provides real-time visibility into file uploads and processing. It displays: * Files currently being processed * Files that completed successfully * Files that encountered errors This view offers transparency into what's happening after files are uploaded. Each file in the Upload Queue displays a clear status. **Processing** - The file is being prepared for use. **Success** - Processing completed successfully and the file is ready. The file will now appear in the Knowledge table. **Error** - Processing failed and the file wasn't added to knowledge. The file can be uploaded again or replaced. ## Accessing Current Information If you want your Assistant to access real-time information from the web instead of static uploaded files, consider enabling **[web search](/creating/advanced#web-search)**. **Important Distinction**: Web search isn't the same as [adding a website](/creating/website-url) to your Assistant's knowledge base. Website sources extract and store page content for your Assistant to reference. Web search is a separate feature that allows your Assistant to search the internet in real-time during conversations. **When to use web search instead of uploading files:** * You need access to current events, news, or recent developments * You want real-time data like stock prices, weather, or statistics * Your course requires students to connect concepts with current information * The information changes frequently and would become outdated if uploaded as static files Web search depends on your [knowledge scope](/creating/assistant-knowledge-scope). With **General Knowledge** or **Topic Knowledge**, an Assistant with web search turned on can search the internet. With **Document Only**, the Assistant doesn't search the web, even if web search is turned on. Configure web search and control which sources your Assistant uses ## Managing Your Knowledge Files Once uploaded, you can manage your knowledge files through the Assistant knowledge settings. Each file has a dropdown menu (three dots) with **Preview**, **Rename**, **Replace**, **Download**, **Hide Source** (or **Show Source**), and **Delete**. Renaming, downloading, and deleting work like this: Make your knowledge files more descriptive and easier to identify: Click on the three dots next to the file you want to rename Choose "Rename" from the dropdown menu Type the new file name in the dialog box Rename knowledge file interface Click **Save Changes** to apply the new file name Use descriptive file names that reflect the content, purpose, or source of the file to better organize your knowledge materials. Retrieve a copy of a knowledge file for backup or modification: Choose the "Download" option from the dropdown menu Your browser will prompt you to save the file to your local device Downloading knowledge files is useful for backing up important content or making modifications before re-uploading. Remove files from your Assistant's knowledge base: Choose the "Delete" option from the dropdown menu When prompted, confirm that you want to remove the file **Permanent Action:** Deleting a file permanently removes it from your Assistant's knowledge base. The Assistant will no longer be able to access information contained in that file. This action can't be undone. ## Optimizing AI Performance Upload complete documents rather than fragments. AI performs better with full context - include entire syllabi, complete assignment descriptions, and full policy documents rather than excerpts. Add files with concrete examples, sample work, and case studies. AI learns patterns from examples and can better replicate the style and quality you expect. Name files with clear, descriptive titles that indicate content, purpose, and context (for example, "Week-3-Calculus-Problem-Set-with-Solutions.pdf" vs "homework3.pdf"). Start with foundational materials (syllabus, course overview) then add specific content (assignments, readings). This mirrors how humans build understanding from general to specific. Upload documents that cover exceptions, special circumstances, and frequently asked questions. This helps AI handle unusual situations appropriately. Test your Assistant with questions that require connecting information across multiple documents to ensure it's synthesizing knowledge effectively. *** ## Next Steps Configure the boundaries of information your Assistant can access Set the directions that shape your Assistant's behavior ### Video Tutorial Watch how to configure knowledge scope and upload knowledge files # Knowledge Scope Source: https://support.nectir.io/creating/assistant-knowledge-scope Configure the boundaries of information your Assistant can access and understand. Knowledge scope configuration interface Knowledge scope controls how much information your Assistant can access beyond your uploaded files. Think of it as setting the **knowledge boundaries** - from using only your specific documents to accessing the full breadth of AI knowledge. This powerful setting allows you to **balance specificity with comprehensiveness** based on your teaching goals. Hover over your Assistant, click the **three dots**, and select **Settings** Click on the **Prompt** tab in the sidebar Choose either **General knowledge**, **Topic knowledge**, or **Document only**: **What it includes:** Your documents plus the full breadth of knowledge available through Nectir's supported AI models. **Best for:** Assistants that need comprehensive, up-to-date information beyond your specific course materials. **Choose this when you want your Assistant to:** * Provide comprehensive, up-to-date information beyond your course materials * Access current events, recent research, or latest developments in the field * Offer broad context and background information * Support research with wide-ranging knowledge **Example:** A research methods Assistant that needs to reference current academic standards and recent publications. **What it includes:** Your documents and only the relevant topics from the AI models based on your documents, the prompt, and user input. **Best for:** Assistants that should stay focused on your course content while having access to relevant background information. **Choose this when you want your Assistant to:** * Stay focused on your course content while providing relevant context * Offer background information that's directly related to your materials * Provide explanations without going too far off-topic * Balance course-specific guidance with helpful context **Example:** A biology tutor that can explain related concepts but stays focused on your course curriculum. **What it includes:** Only the information in the documents you upload. The Assistant doesn't use the AI model's general knowledge or web search, and declines questions your documents don't answer. **Best for:** Assistants that should only reference your specific course materials and nothing else. **Choose this when you want your Assistant to:** * Reference only your specific course materials * Avoid any external information or interpretations * Stick strictly to your syllabus, assignments, and uploaded documents * Provide answers based solely on your provided content **Example:** A syllabus Assistant that only references your specific course policies and requirements. Click **Save changes** to apply your knowledge scope setting ## Best Practices **Important:** Your knowledge scope setting works in combination with your Assistant's prompt. Make sure both are aligned with your teaching goals. Begin with **Document Only** or **Topic Knowledge** to ensure your Assistant stays focused on your course content. Monitor your Assistant's responses and adjust the knowledge scope if needed based on student feedback. Periodically review your knowledge scope setting as your course content and goals evolve. Source citations are on by default and show which knowledge files your Assistant uses. This helps identify gaps in your content and ensures responses are grounded in your materials. *** ## Next Steps Set the directions that shape your Assistant's behavior Configure AI models and advanced options *** ### Video Tutorial Watch how to set knowledge scope and configure AI models # Write an Assistant prompt to shape responses Source: https://support.nectir.io/creating/assistant-prompt Customize how your Assistant responds and behaves, ensuring relevance and accuracy while enhancing its effectiveness. Assistant prompt configuration interface ## What's an Assistant Prompt The Assistant prompt is the **foundational instructions** that define how your AI Assistant behaves, responds, and interacts with students. Think of it as the "personality and guidelines" you give your Assistant before it ever meets a student. **Important:** The Assistant prompt is different from what students type in the chat interface. The prompt is the **behind-the-scenes instructions** that shape every response your Assistant gives. ## Creating Your Assistant Prompt Follow this proven workflow to build an effective Assistant that provides meaningful support for your students' learning journey. Start by identifying what **opportunity for support** you want to address for your students. Every great Assistant begins with understanding the specific way you want to help students succeed. **Which of these resonates with your teaching goals?** * Help students understand complex concepts through guided questions * Provide study strategies and exam preparation support * Guide students through problem-solving processes step-by-step * Offer writing feedback and improvement suggestions * Support students with course logistics and assignment clarification * Encourage deeper thinking about course topics Sometimes the biggest challenge is identifying **what kind of support your students need most**. Reflect on common questions you get during office hours or frequent struggles you observe. On the **Prompt** tab of your Assistant's settings, click **Prompt Library** and select a template that aligns with your support opportunity. Review and edit it, click **Apply Prompt**, then click **Save Changes**. Templates provide proven starting points that you can customize. Templates focused on helping students master course content and develop learning skills. Answers course-specific questions and guides students through content Provides personalized learning support without giving direct answers Acts as a collaborative study companion to reinforce understanding Helps students develop and organize research projects Templates designed to improve student writing and communication skills. Guides students through the writing process and skill development Focuses on professional and academic writing standards Templates that support broader student success and career preparation. Helps students develop career skills and professional readiness Provides guidance on academic planning and college success strategies Creates interactive scenarios for skill practice and application Templates for handling course logistics and administrative tasks. Manages course logistics, schedules, and administrative questions Helps students understand institutional policies and procedures Templates designed to assist educators in their teaching and planning. Helps educators plan and structure learning experiences Guides educators through the process of creating effective AI Assistants Give your Assistant context by uploading relevant course materials. Even a small amount of content helps your Assistant understand your course and provide more relevant responses. **Why this matters:** An Assistant without context can only give generic responses. Upload your syllabus, key readings, or assignment descriptions to make responses specific to your course. Add syllabi, readings, assignments, and learning materials Before students interact with your Assistant, you need to test it extensively. This is where you discover what works and what needs improvement. **Test like a student:** * Ask questions you know students will ask * Try to get the Assistant to give direct answers (if you don't want that) * Test edge cases and unusual requests * Ask about topics outside your course materials **Questions to explore:** * "Can you just give me the answer to problem 5?" * "I don't understand anything about (TOPIC)" * "What's the most important thing for the exam?" * "Can you write my essay introduction?" **Test with the mindset of different student types:** the confused student, the student looking for shortcuts, the highly motivated student, and the student who's struggling with basic concepts. Based on your testing, refine your prompt by adding **actionable directions** that guide better behavior and removing directions that don't apply. **Focus on what TO do, not what NOT to do:** "Don't give direct answers. Don't solve problems for students. Don't provide solutions." "When a student asks for help: 1) Ask what they've already tried 2) Guide them to identify the next step 3) Encourage them to apply the concept" **Plan conversation paths based on student responses:** * If student is confused → Ask clarifying questions about their current understanding * If student wants shortcuts → Redirect to learning process and explain the value * If student is stuck → Break down the problem into smaller, manageable steps * If student shows good understanding → Challenge them to extend their thinking **Avoid laundry lists of restrictions.** Instead, create clear pathways that guide your Assistant toward productive educational interactions. ## Key Principles for Effective Prompts Create **if-then scenarios** that guide your Assistant's responses. "If a student is confused, then ask about their current understanding and guide them step by step." Write directions about what **TO do** rather than what not to do. Positive directions create clearer guidance for productive interactions. **Testing reveals everything.** Try to break your Assistant, ask tricky questions, and test edge cases before students interact with it. Your first prompt won't be perfect. Observe how students interact with your Assistant and refine based on real usage patterns. ## Controlling Web Sources If you've enabled [web search](/creating/advanced#web-search) for your Assistant, you can control which websites it uses by adding instructions directly in your Assistant prompt. **Why control sources in your prompt?** Instead of your Assistant searching the entire internet, you can ensure it only references authoritative, course-appropriate sources that align with your academic standards. **Example prompt instructions:** * "When searching the web, only reference information from .edu and .gov domains" * "Use scholarly sources like Google Scholar and academic journals when searching" * "When searching for medical information, rely only on CDC, WHO, and peer-reviewed medical journals" * "Only use the following trusted news sources: \[list specific domains]" Example of web search source instructions in an Assistant prompt These instructions work best when they're specific and actionable. The more precisely you define your source requirements, the better your Assistant can follow them. ## Creating a Custom Prompt Want to write a custom prompt from scratch? Use AI to help you create it. Follow this process: ### Step 1: Create a prompt-writing Assistant Create a new Assistant using either: * **AI Assistant Creator template** - Specifically designed to help you create effective Assistants * **Blank Assistant with General Knowledge** - Start fresh with full AI knowledge access ### Step 2: Use These Conversation Starters Copy and paste these prompts to get started with your prompt-writing Assistant: I'm creating an AI Assistant for my (SUBJECT) course. The main challenge my students face is (DESCRIBE CHALLENGE). I want my Assistant to help them by (DESCRIBE DESIRED SUPPORT). Can you help me brainstorm the key behaviors and approach my Assistant should have? Based on our discussion, please write a first draft of an Assistant prompt. The Assistant should (DESCRIBE ROLE AND GOALS). Make sure to include specific guidance on how the Assistant should respond to different types of student questions. Focus on creating conversation paths rather than listing restrictions. Here's how my Assistant responded to some test questions: (PASTE ACTUAL RESPONSES). The responses were (DESCRIBE WHAT WORKED/DIDN'T WORK). Please help me revise the prompt to improve these specific areas: (LIST SPECIFIC IMPROVEMENTS NEEDED). Help me add specific "if-then" conversation paths to my prompt. If a student is (SCENARIO 1), then my Assistant should (DESIRED RESPONSE). If a student is (SCENARIO 2), then my Assistant should (DESIRED RESPONSE). Can you help me write these as clear instructions in my prompt? ### Step 3: Test and iterate Use your prompt-writing Assistant to continuously improve your custom prompt as you test it with real student scenarios. The AI can help you refine the language, add missing scenarios, and strengthen weak areas. **Pro tip:** Keep your conversation with the prompt-writing Assistant active throughout your testing phase. This way you can quickly get help refining your prompt based on real interactions. *** ## Next Steps Configure AI models and temperature settings Create predefined prompts to help users begin interacting ### Video Tutorial Watch how to select templates and customize your Assistant's behavior # Add from Canvas Source: https://support.nectir.io/creating/canvas-import Add course content from Canvas into your AI Assistant's knowledge base via the Canvas API. The Canvas integration connects directly to your Canvas instance via API, letting you browse and select course content to add to your Assistant's knowledge base. Unlike [Import Course Export](/creating/lms-import), which requires you to download and upload a course export file, this integration pulls content from Canvas with no file handling on your end. You can sync updates daily or weekly to keep your Assistant current as your course evolves. Before you can import from Canvas, a Workspace Owner must set up the Canvas API integration. If you don't see the option to add from Canvas, ask your admin to follow the [Canvas API Setup guide](/lms-integrations/canvas-api). ## How It Works Hover over your Assistant, click the **three dots**, select **Settings**, then navigate to the **Knowledge** tab. Click **Add Knowledge**, then select **Add from Canvas**. If this is your first time, click **Continue to Canvas** to authorize access in Canvas. Nectir keeps your Canvas authorization active for as long as possible. In some cases, you may need to reauthorize. After authorizing, you'll see a list of Canvas courses you have access to. Select the course you want to import content from. Only courses where you're enrolled as a teacher are shown, including for Canvas admins. Courses already imported into this Assistant aren't listed. Select which content categories to import (files, pages, assignments, discussions, announcements, quizzes, syllabus, calendar events). Categories are selected by default. Deselect anything you don't want to import. Unpublished pages, assignments, discussions, announcements, and quizzes, and hidden files and calendar events, aren't listed. You can also expand any category to check or uncheck individual items within it. Fully selected categories automatically pick up new content from Canvas during sync. Narrowed selections sync only the items you chose. Unsupported file formats are marked and can't be imported. Click **Add to knowledge** to begin importing the selected content. Files are processed and added to your Assistant's knowledge base automatically. Once processing completes, your imported content appears in the Knowledge table grouped under the Canvas course name. ## Supported Content Types The Canvas API integration supports importing the following content types from your courses: | Content | What Gets Imported | | - | - | | **Files** | Documents, images, and other files uploaded to your Canvas course | | **Pages** | Wiki pages and their full content | | **Assignments** | Assignment descriptions, instructions, and due dates | | **Discussions** | Discussion topic prompts and descriptions | | **Announcements** | Course announcements and their content | | **Quizzes** | Quiz questions, answer choices, and descriptions | | **Syllabus** | Your course's syllabus page content | | **Calendar Events** | Standalone course calendar events | All content is imported as **read-only**. Nectir never modifies or deletes anything in your Canvas course. ## Keeping Content Up to Date As your course content evolves in Canvas, there are two ways to keep your Assistant's knowledge base in sync. ### Manual Sync Check for updates at any time. This is also how you add items or categories you skipped in an earlier import, since imported courses don't appear in the **Add from Canvas** course list. 1. Open the **Knowledge** tab in your Assistant's settings 2. Find your Canvas course in the Knowledge table and open its menu (three dots) 3. Click **Check Updates** to scan for content that was added, modified, or deleted in Canvas 4. Review the changes. To add content, expand a category and check the items you want, or check a category you skipped before. 5. Click **Update Knowledge** to apply the changes ### Automatic Sync Turn on automatic sync to have Nectir check for content changes on a schedule: 1. Open the Canvas course in the Knowledge table 2. Toggle automatic sync on 3. Choose your sync frequency: **daily** or **weekly** You can also turn on automatic sync on the content selection step of your first import. On your chosen schedule, Nectir adds new content, updates modified items, and removes items that were deleted in Canvas. Your knowledge base stays in sync with the course. Automatic sync and manual **Check Updates** work independently. You can use **Check Updates** at any time, even with automatic sync turned on. ### Removing Imported Content Individual Canvas items can't be deleted from the Knowledge table. You can keep Canvas content out of your Assistant's knowledge base in two ways: * **Narrow the selection.** Uncheck the items you want to exclude when choosing content to import, or later from **Check Updates** on the course row. Narrowed selections sync only the items you chose, so excluded items are removed and stay out of your knowledge base on future syncs. * **Delete the course import.** Open the menu on the Canvas course row in the Knowledge table and click **Delete**. This removes every item imported from that course. You can import the course again at any time. Nothing you remove in Nectir affects the content in your Canvas course. ## Troubleshooting The Canvas API integration must be configured at the Workspace level before instructors can import content. Ask your Workspace Owner to set up the connection using the [Canvas API Setup guide](/lms-integrations/canvas-api). * Verify you're enrolled as a teacher in at least one Canvas course * Courses already imported into this Assistant don't appear again. Find them in the Knowledge table. * Check that your Canvas account is active and not restricted * If the issue persists, contact [support@nectir.io](mailto:support@nectir.io) Files in unsupported formats can't be imported. Convert them to a [supported file type](/creating/assistant-knowledge#supported-knowledge-sources) in Canvas, then try importing again. * Check your internet connection * If the issue persists, contact [support@nectir.io](mailto:support@nectir.io) with details about the course and content types you're importing *** ## Next Steps Configure the boundaries of information your Assistant can access Set the directions that shape your Assistant's behavior # Conversation Starters Source: https://support.nectir.io/creating/conversation-starters Create engaging conversation starters that guide users to meaningful interactions with your AI Assistant. Creating conversation starters animation Conversation starters are pre-defined prompts that help users begin interacting with your AI Assistant. They serve as suggested entry points that: * Remove barriers to engagement for new users * Guide users toward productive interactions * Showcase your Assistant's capabilities * Encourage exploration of your Assistant's knowledge Hover over your Assistant, click the **three dots**, and select **Settings** Select **Conversation Starters** in the sidebar navigation Click **Add New** for each starter. You can create up to 4 conversation starters or prompts (50 characters maximum per starter) Focus on creating prompts based on: * Common questions users might have * Sentence stems that can lead to informative responses * Interesting questions that may spark a learner's curiosity * Examples that showcase your Assistant's specialized knowledge Click **Save changes** to apply your conversation starters ## What To Consider Remember that conversation starters have a **50 character limit each**. Here are some approaches to maximize their effectiveness: Open-ended stems that encourage exploration: * "How can I…" * "What should I know about…" * "Why does…" Direct requests for specific help: * "Help me with…" * "Explain the…" * "Show me how to…" Ready-to-use questions for your subject: * "What's due this week?" * "How do I study effectively?" * "Where do I find resources?" Stems for organization and preparation: * "Help me plan…" * "How do I prepare for…" * "What's the best way to…" *** ## Next Steps Invite users to start interacting with your Assistant Monitor usage and performance metrics ### Video Tutorial Watch how to create engaging prompts that help students get started # Create an Assistant Source: https://support.nectir.io/creating/create-an-assistant Nectir AI allows educators to create custom AI Assistants for 24/7 student support, reducing workloads and boosting engagement. ## Discover the new Assistant Builder Create an Assistant by describing it. You can create or edit an Assistant just by describing what you want. The Assistant Builder turns your description into a complete Assistant with a name, description, prompt, and conversation starters, then asks follow-up questions to help you refine it.