We have updated Arcane with a small, but impactful set of features and fixes!

New Features

Search and Insert Media

With the new email image insertion feature, you can now effortlessly browse and insert images into your content.

Simply clicking “Replace” will open the Media selection view where you can apply powerful Filters and search using natural language!

Arcane understands the content of your images, so you can search by phrases or text in the images and also just by describing the scene.

Improved Search by Document Number in Arcane

Arcane now supports more than just searching for documents using their Document Numbers. You can input one or more Document Numbers in your query, and Arcane will retrieve the relevant document(s).

If you’ve only provided the Document Number it will summarize the document, but you can do a lot more than that.

For instance, asking 'Look in MAT-1234 and explain the data about contraindications' accurately limits results to that document and answers your specific query.

Similarly, queries like 'What are the common themes in MAT-1234 and MAT-8319' will generate results based on both documents.

Authoring Assist - Email Preview

The Authoring Assist feature now allows you to preview your emails, displaying how the content will appear to recipients without any editing tools or additional UI elements on screen.

By providing a clean presentation of your email, this enhancement ensures an accurate view of the final product, akin to the Activator Preview popup experience.

Fixes

My Files Menu Item Display Update

Users without the 'file_upload_allowed' role will no longer see the 'My Files' menu item in the interface. This change eliminates confusion for users that could see the option but not access the related page.

Translation still “pending”

An issue causing a document in translation to complete and notify the user, but not remove the “pending” animation in the language dropdown has been fixed.

Inline Citation Enhancement

We have improved the behavior of inline AI citations, with fewer cases where citations are clustered at the end of a response. Citations should usually be placed by the sentence or relevant paragraph.

We’re happy to announce our 1.2.0 release of Arcane. There are a number of new features to make the life of Authors easier, from manually inserting new email section, download of distribution ready packages to Undo/Redo. We also enable you to search by Vault Document Numbers, and customize the quick prompts that are available to your organisation.

A very exciting new feature is Approved Images for Authoring - Arcane now loads the approved images from your Vault or S3 bucket and our AI enables

Coming Soon: We will be releasing Arcane 1.2.1 in the last week of April with additional features for Authors - Allowing them to preview their emails and to manually search and filter for images using natural language.

New Features in 1.2.0

Authoring Performance Improvements

An incredible 8x improvement in response speed enables you to author more naturally and iterate faster on your content with Arcane.

Approved Images in Authoring Assist

Now when you prompt Arcane to create a section of content, it will search the Approved Images that have been imported into Arcane and select one that is related to the topic of your section.

You can also select an image in your email and simply prompt the topic, and it will find a relevant image.

This search includes concepts, colors, positions and also the text in your image - so prompts like “Find an image of a smiling woman in a red shirt” or “Find a Remepan 10mg pack shot” will just work.

Note: A Manual image browser with natural language search will be released in our 1.2.1 release planned for the last week of April.

Note: As of 10-04-2025 we have started import jobs for the approved images for our partners. This may take some time to complete as there is a large volume of images to import and run through AI for annotation and embedding. If you wish to check on the status of your image library please contact your Anthill partner.

Download Email from My Creations

You can now download your email content as a .zip file, which includes image assets, from the My Creations section. This feature enables you to transfer your Arcane Email to other systems, ensuring that all visual and text elements are preserved in a single package for easy distribution. The download package is Veeva Vault ready and compatible with Anthill Activator for additional editing, and includes HTML suitable for use in SFMC, Veeva CRM, IQVIA and additional platforms.

Manual Layout Selection for Emails

	Just click the plus icons above and below your selected content to open the Layout Selector! This lets you add a new Layout block to your email to shape the final asset.

Custom Quick Prompts

Your company can now have the Quick Prompts customized so that they are even more relevant to your business.

Contact Anthill to request changes to your Quick Prompts configuration.

Search by Document Number

This feature enhances document retrieval by allowing you to search using one or more document numbers. When you enter a document number like "MAT-1234" in the prompt, the system will direct you to the specific document's content. Additionally, if you input multiple document numbers, such as "MAT-1234 and MAT-8319," both documents will be included in the response.

Authoring Assist Undo/Redo

	The Authoring Assist feature now supports Undo and Redo functions, enhancing your editing capabilities. You can easily reverse or reapply your last action using standard keyboard shortcuts or by clicking the undo/redo icons, ensuring a smoother content creation experience.

Improvements

Tactic Information and Authoring Configuration Updates

Arcane now applies the guidance from your Personas, Style and other Tacit contexts more coherently to the content it creates.

My Creations: Enhanced Email Opening

When you open an email from the My Creations section, the associated conversation and settings are now automatically loaded.

Improvements to Save Behavior for Manual Edits

Previously, manual edits such as directly changing text, rearranging layouts, or deleting sections during an edit session might not have been saved automatically in "My Creations." This has been addressed to ensure that any manual alterations are effectively saved, in addition to AI-generated updates. Now, when you make such changes, they are preserved once your session ends, maintaining the integrity of your work without relying solely on AI-triggered save actions.

Inline Citation Enhancements

You will now experience improved inline citation with enhanced snippet/document labeling within the Arcane responses for the Astellas platform. Previously, some responses lacked these labels, leading to potential confusion. The update now ensures a consistent display of snippet/document labels for clearer references. Check the conversation link provided for a detailed view of the changes implemented.

Non-Movable Layout Enhancement

Layouts recently added to your project, whether manually or through AI assistance, are now fixed in position within the interface. This adjustment enhances stability and predictability when organizing content, ensuring that new layouts maintain a consistent structure without the ability to be rearranged up or down after their initial placement.

We're excited to introduce enhanced collaboration, search and authoring features for our Arcane 1.1.0 release!

Now, you can share Collections with colleagues using a simple Share button, prioritize official content in results at an organization level using Tiered Content, create Email content faster than ever.

We are also releasing infrastructure that will power our Arcane 1.2.0 release with advanced Image search for both users and AI.

Existing users can try these features now.

Major Features

Tiered Content

Tiered Content enables organizations to prioritize search results by categorizing documents into various importance levels based on filters. This enables Arcane to surface the most relevant documents when a topic is present across many types of documents, without impacting query performance. Organisations can not only define Tiers, but also set their relative level of importance to adjust the results further.

Tiers can be defined as simply as the example above, they can also be based on multiple criteria, or literally rely on a list of specific documents - it’s about what makes sense to your Organisation.

This global Tiering can always be overridden if a user is -for example- manually applying a filter to focus their interactions to content from a local market. In that case their queries will be managed by the user defined settings.

Reach out to the Anthill team if you would like to configure Tiered Content for your organization!

Shared Collections

You can now share your collections through a simple link, offering a seamless way for others to access your curated content. Upon clicking the share link, recipients are directed to a new conversation in content assist.

If a user adds the collection, they have access to a read-only version of the collection which is updated automatically whenever the collection owner makes changes to the original collection.

Note that if the original creator uses the link, it opens the existing collection without creating additional shared versions, and duplicate collections for the same shared content are prevented by validating link identity before creation.

Image Import

The Arcane platform now uses multimodal import for Image assets, allowing Arcane to generate detailed understanding of the contents of each image. This import capability will surface in user features in the Arcane 1.2 release which will allow users (and the AI) to search for images that match certain requirements using natural language e.g. “Find an image of a grandmother reading to a child” or “Find a Remepan 10mg pack shot”.

Authoring Updates

Faster AI Content Creation

We have improved the process used by Arcane AI when generating new content, reducing the time from prompt to content creation.

Faster Collection Summaries

We now create Collection Summaries over twice as fast as in Arcane 1.0 by using the Claude Haiku 3.5 model.

Better Persona Styling when creating multiple Content Blocks

With this update, your Personas, Segments, Style and Editorial Tone are applied more consistently when generating multiple layouts at once via a prompt like “Create and email about efficacy and safety data”.

This update corrects instances where context prompts were inconsistently applied, enhancing reliability in content creation across diverse topics.

Updates to My Creations

Emails are now listed in correct chronological order, and the My Creations view updates with the latest creation if you switch from an email directly to My Creations.

Security

Security Updates

The Arcane application serves content as a static webpage enhanced via an HTTP API, with authentication managed by Auth0. This setup secures both user access and role assignment. Every API endpoint is protected: user endpoints require Auth0 token validation, while machine endpoints utilize IAM for access control. Communication between internal components likewise employs IAM-protected messaging and queues. Public endpoints are absent, directing security attention to potential DDOS vulnerabilities at the validation middleware. To fortify defenses, WAF protection is now enabled for both our API and CloudFront distribution.

Web Application Firewall (WAF) Configuration for API Gateway and CloudFront

Arcane API and web distribution security has been improved with updates to the configuration and deployment of Web Application Firewalls (WAF) directly configured via Terraform. This update allows efficient management of WAFs on both the API Gateway and the CloudFront distribution. Each WAF is pre-configured with request throttling to prevent abuse and ensure that Arcane remains responsive under varying traffic loads.

Data Connectors

S3 Connector

The new S3 Connector offers a streamlined way to ingest documents into Arcane directly from an S3 bucket. This feature facilitates the processing of Danone’s input, which includes PDFs and associated source files, through a structured import mechanism. Documents are paired with metadata outlined in a supporting spreadsheet, ensuring accurate import and classification with a unique S3 external ID. The connector enables a two-step process: initial document creation and update, followed by vectorization and final import.

Bug Fixes

Handle Empty Collections

Arcane now checks that your Collection has at least one document in it before entering Authoring mode. We also now correctly allow users to Share empty Collections.

My Creations Error Fixes

You can now open and edit your saved contents in My Collections without encountering server errors. Previously, some users experienced errors and encountered issues saving new content. The underlying cause has been resolved, allowing you to access your documents smoothly.

'New' Button in Authoring Mode

In Authoring mode, the “New” button will correctly clear the current email and allow the user to begin creating a New Email

Thumbnail Display in My Creations

Fixed an issue where some users were not seeing thumbnails of their saved documents in the My Creations view

While our partners have already been using Arcane in production for months, the 1.0.0 milestone is here with major new updates to the Arcane platform.

This release will be available to all customers from February 18th 2025.

Content Authoring

Authoring Assist 1.0

Authoring Assist revolutionizes content creation for marketers in the pharmaceutical industry. It enables rapid, compliant internal asset production, reducing the reliance on expensive agency costs.

With Authoring Assist, marketers can create emails in minutes rather than days or weeks.

The tool leverages approved content and design systems to generate tactics automatically, while allowing for manual edits and AI-assisted rewrites. It supports personalization through audience personas and segmentation, adapting content to specific channels and editorial styles.

By streamlining the content creation process and enabling more targeted communications, Authoring Assist empowers marketers to deliver more effective and personalized campaigns.

Personas, Tone of Voice and More

Tactics Contexts

This feature allows you to capture the information about what is relevant to speficic Personas, the various editorial styles you use when creating content, and any other context you need to create scalable and repeatable omnichannel communications.

Users simply select the relevant settings and prompt their content as normal - Arcane applies all of the presets to control the content created!

This feature supports the creation of Global and Regional variations in settings.

Content Referencing

In-line citations in every Arcane response link directly back to the specific text within Approved source documents. This makes it even easier to dig deeper into your content, as well as makes it easier to validate the accuracy of every Arcane response.

Data Connectors and Exporters

S3 Source Connector Interface

Introducing the new S3 Source Connector Interface, designed for seamless document management through Amazon S3. This feature parallels the capabilities of the MyFiles and Veeva connectors, allowing you to effortlessly manage and import documents from S3. Optimize your workflows by integrating S3 seamlessly into your document management process.

Download Glossary as CSV

Easily export your open Glossary as a CSV file with the new “Download as CSV” button. This feature allows you to download the Glossary locally, enabling seamless access and use of your Glossary data in other applications or for offline analysis and sharing.

Security Updates

OWASP Compliance

We are now compliant with OWASP standards, paving the way for upcoming penetration testing on Arcane. This milestone ensures that we meet essential security measures, safeguarding against common vulnerabilities.

Enhanced Security Headers

Addressing multiple insecurities identified in our header configurations, we've implemented the following updates: Content Security Policy, Cross-domain misconfiguration corrections, Anti-clickjacking protection via X-Frame-Options, minimizing Server header version information leaks, enforcing Strict-Transport-Security, and resolving timestamp disclosure issues. Additionally, the X-Content-Type-Options is now in place, and cache control headers have been set. These measures significantly bolster our systems' security posture. Further details are available in the attached OWASP Report.

Security Update for DomPurify Package

The DomPurify package has been updated to address a flagged vulnerability in version ^3.0.8. This security enhancement ensures protection against potential threats. It is recommended for all users to update to the latest version to maintain the integrity of their applications and safeguard them from any known security risks.

Infrastructure & Platform Updates

Platform Notification Component

Arcane can now display a banner with system messages to communicate important updates or information to users.

Independent Indexing Lambda

The indexing lambda has been refactored to remove dependencies on Veeva-specific fields, enabling a more flexible and adaptable indexing process. This change promotes broader compatibility with various data sources and ensures that the lambda function can be utilized across different platforms without Veeva-related constraints. Enjoy a streamlined integration flow, reduce potential configuration conflicts, and maintain a cleaner, more efficient code environment.

Create Import Pipeline

The import pipeline allows for enriched PDFs and source files from spreadsheets to be systematically imported with complete metadata. Each document receives an externalId for seamless discovery via the S3 source connector on the backend. Modeled after anthill documents, this integration provides a structured framework for streamlined document management and indexing.

Notification Banners Backend

We've developed the backend infrastructure for notification banners using DynamoDB to store notifications. Each notification includes a message in HTML, a start and end time, and a time-to-live (TTL) setting, enabling efficient management and display of banners. Additionally, a new endpoint has been introduced to allow the frontend to seamlessly load current notification banners.

Localization Support in Frontend

Our frontend now supports localization through a seamless integration with the i18n plugin, enabling the translation of text strings into multiple languages. This enhancement compiles all text strings into a single JSON object, making it easier to manage translations across the application. This improvement is crucial for teams operating in global markets, as it streamlines the localization process and ensures a consistent user experience worldwide.

Error Message Enhancements

The platform flows now include updated error messages, providing clearer, more actionable information to help users resolve issues faster. This enhancement improves the accuracy and readability of alerts across our interface, ensuring that users are guided effectively when encountering errors.

Error Messages for LLM Flows

Enhancements to Large Language Model flows now include the capability to send error messages when an error prevents the delivery of the expected result to the frontend. This new type of system message provides clear and immediate feedback on system interruptions, allowing for quicker issue identification and resolution.

Enhanced Error Notifications

Axios error handling has been improved by incorporating toast notifications for easier diagnosis and instant visibility. When endpoints return error status codes like 500 or 400, a user-friendly message will now alert you of the issue, streamlining troubleshooting and enabling quicker resolution.

Display Language Codes in Dropdown

The language selection dropdown has been enhanced to include language codes, making it easier to distinguish between similar options like 'ch' and 'ch_HANS'. This update ensures accurate and clear identification of language preferences for users in multilingual environments.

Translation Retry Button Enhancement

Addressed a bug where the try again button in the translation notification modal was unresponsive. Users can now select a source language from the dropdown, even if not previously utilized in backend processes. Enhanced the metadata API to provide the languageCode, with frontend updates ensuring correct language handling. Additionally, updated the translate-notifications flow to guarantee error notifications contain necessary data for retrying translations, preventing 500 errors from being thrown. The glossary API is now triggered upon changes to the source language.

Support for Empty IDML Files

We've addressed an edge-case where translating empty IDML files resulted in a stalled translation process. Now, even when an IDML file lacks text, our system efficiently handles the scenario and provides appropriate feedback, ensuring a smoother experience when managing such files.