Skip to main content
Skip table of contents

HALO

14-04-2025 | Audience Feature Changes

We have released a new update that removes "audience" as a stand-alone feature and incorporates it as a value set within the Context framework.

What does this mean for existing sources?

  • A new audience context variable has been added for sources previously categorized as "internal" or "external."

  • For sources labeled with the audience "all", no context variable has been set. These sources will continue to be retrieved as before, maintaining the same logic.

  • As a result, the retrieval process for existing sources remains unaffected by this change.

What does this mean for newly added sources?

  • When a new source is added to the knowledge base, it will not have an audience context variable assigned by default. This is equivalent to the "all" audience setting in the previous implementation.

  • To classify a new source as "internal" or "external", simply add the audience context variable to the respective source and select the appropriate category.

Note:

  • For newly created profiles, the audience context variable must be manually added to the Contexts table.

  • For existing profiles, the audience context variable has already been added to the Contexts table by default.

12-03-2025 | HALO

Bug fixes:

  • Fixed an issue with the agents/command palette when selecting an item

  • Fixed an AVA issue that prevented adding steps to a tool in certain situations

Improvements

  • Reorganized Profile Settings: We've removed the single Profile Settings page and split its content into two separate dialogs:

    • A Profile dialog that includes basic settings like name, API Key, and Product Token

    • A dedicated dialog on the Knowledge page for Scheduled Sync and Webcrawler Blacklist settings

  • Added a new Settings dialog under the Agents page to simplify management of root agents

05-03-2025 | HALO

New features:

  • System Context: Introduced the system context feature. Now you can use current_datetime variable and change its timezone as needed

  • Agents & Tools Management: Added new filters and sorting options to enhance the management of agents and tools.

  • Copy/Duplicate Agents & Tools: Enabled copying of tools and agents between profiles

Improvements

  • Context Secret Column: removed this column, the secret values have an eye button to toggle visibility

  • Bulk actions for tables: improved delete dialogs and added bulk actions for most tables

Bug fixes

  • HTTP Step saving bug: Fixed a bug with the HTTP step getting reset on save

  • Tool Analytics and Saving Enhancements: Fixed issues related to tool steps analytics graphs

25-02-2025 | HALO

Improvements:

  • A confirmation popup will now appear if you attempt to navigate away from the page while editing agents or tools. This helps prevent accidental navigation and loss of unsaved changes.

  • Improved code block highlighting in Test Center, Ava messages, and Conversation History, featuring a convenient "Copy" button.

  • Introduced a "Fallback Language" option on the Style page. This language will be used if the automated language detection fails. It is set to English by default

21-02-2025 | HALO

Bug Fixes:

  • Resolved an issue with user interaction node

  • Fixed an error where updating audience setting for file uploads was resulting in a 422 status code.

  • Addressed a problem where audience updates for file uploads failed due to a faulty dirty check logic.

  • Corrected an issue that caused the agent editor to refresh and lose progress after saving settings.

  • Fixed a bug in the JSON editor that ignored subsequent keyboard events after the use of the '/' command.

17-02-2025 | HALO

Bug fixes

  • Fixed issue with HTTP step URL arrows keys functionality

  • Resolved tool description editing problems

  • Fixed JSON mode button overlap issue

  • Corrected source references in test center

  • Fixed text overflow when deleting resource configuration

  • Resolved copy-pasting issues in Ava

  • Fixed Draft agents and tools functionality in agent builder

  • Addressed camelCase and snake_case inconsistencies for tools

Improvements

  • Enhanced preview visualization on node steps

  • Improved code display in test center/Ava environment

  • Added prevention mechanism to stop Ava from adding nodes after End node

06-02-2025 | HALO

Introducing HALO (previously known as GenAI). HALO introduces a new way of working together with AI.

This major release introduces several new concepts:

  • Agents:

    • AI Agents are autonomous programs capable of understanding context, making decisions, and using tools to achieve their goals. In Halo, you can utilize agents to automate nearly any process imaginable. Your agents are defined using natural language and can collaborate with other agents while using tools to carry out their tasks, such as querying databases or generating content.

    • Available through the HALO Studio application

  • Tools:

    • Tools within HALO's agent ecosystem are designed to enhance agent functionalities by facilitating seamless integration with external APIs or services. These tools enable agents to effectively communicate with external systems, execute tasks automatically, and interact with complex datasets, thereby expanding their operational capabilities and improving overall efficiency.

    • Available through the HALO Studio application

  • Governance:

    • Introduces concepts like AI audit logs, and comprehensive approval flows that allow you to map any organizational structure and integrate it into your agents and tools for human-in-the-loop approval. The governance application provides full transparency into what your AI is doing, when it’s doing it, and why.

    • Available through the HALO Governance application

As well as some changes to existing previously existing concepts:

  • Style (previously known as “behaviour”):

    • The Style feature in HALO allows seamless customization of AI communication to reflect your brand's identity, ensuring consistent messaging and enhanced user engagement. By adjusting writing style, tone of voice, and response structure, businesses can create interactions that resonate with users, foster trust, and maintain brand integrity. This personalization leads to effective and welcoming communication, optimizing the overall user experience with HALO's AI agents.

    • Available through the HALO Studio application

  • Knowledge:

    • Knowledge within HALO's agent ecosystem is designed to enhance agent capabilities by providing access to business-specific information and context. By utilizing Knowledge, agents can deliver accurate and relevant responses, drawing from pre-curated sources such as documents (PDF and JSON) and web crawled data. This empowers agents to conduct informed interactions, automate complex decision-making, and ensure consistency across communications, thereby enhancing operational effectiveness and user satisfaction.

    • Available through the HALO Studio application

For more information take a look at CM.com's knowledge center.

26-08-2024 | HALO

Improvement | Increased Number of Sources

Previously when asking a question, the top-3 knowledge source chunks were taken into account by the LLM to answer the question. We have increased that number to 5, such that more relevant information is considered and answers will become (even) more relevant. Good to note is the additional 4th & 5th source will only be shown if they have (at least some) relevance to the question; otherwise still only the top-3 sources will be shown.

21-08-2024 | HALO

Feature | Feedback Mechanism

Overview
The Feedback Mechanism is a new feature within the HALO engine that allows you to provide feedback on AI-generated answers. By refining the way HALO interacts with questions, you can significantly enhance the accuracy and relevance of the responses provided. The Feedback Mechanism is a part of our Quality Assurance features, enabling you to gain greater insights and control over the HALO engine.
Tip: you can also check out a Discover Video on the Feedback Mechanism here!

Navigating Quality Assurance & Feedback Mechanism

Two new tabs have been added to the left side of the HALO dashboard: “Conversations” and “Feedback”. The “Conversations” tab displays all the interactions in which HALO has participated. Here, you can view general metrics of these conversations, such as the topic and recognition rate—the latter being the percentage of messages to which the HALO engine was able to provide a response.

By clicking on a specific conversation, you can review all the messages exchanged during that conversation. This view allows you to like or dislike the HALO engine's responses.

82b12ee3-5514-46e2-8783-13c5539da90e-image

 

Types of Feedback

If you are satisfied with a HALO-generated answer, you can like it. We're expecting to leverage responses marked as "liked" for future features to ensure consistency in your knowledge base over time, and to run model finetunings for your specific environment. Additionally, liking an answer serves as a reminder that you've already rated the response and don’t need to evaluate it again.

If you are dissatisfied with a HALO answer, you can dislike it and choose from several options to specify why you didn’t like the response: “I don’t like the answer,” “Wrong answer,” or “Do not answer.”

3e9efebf-9149-4f8e-b746-2ee16c33d8ca-image

Wrong answer

Select the “Wrong answer” option if the HALO engine provided an incorrect response. When you choose this option, you’ll be prompted to review the knowledge base used to generate the answer, as an incorrect response often indicates that the knowledge base contains incorrect, vague, or outdated information.

Do not answer

You can also select “Do not answer” as a feedback reason. This option ensures that similar questions in the future will not be answered by the HALO engine. Use this option when you prefer that certain user questions, such as those on sensitive topics or other company-specific contexts, should not be addressed by AI. When this feedback is applied, future similar questions will either prompt the end user to be transferred to a live agent, or, if you are using the HALO basic package, the question will simply remain unanswered.

I don’t like the answer

If the answer provided by the HALO engine is mostly correct, but you are not fully satisfied with how it’s phrased, formatted, or structured, you can select the “I don’t like the answer” feedback type.

When you choose "I don’t like the answer," you’ll be taken to a new view where you can provide detailed feedback using the Feedback Mechanism. Here, you’ll see the original question and answer, along with a field to enter your feedback. In this field, you can specify how you’d like the answer to be modified, for example: “make it shorter,” “remove the last sentence,” or “format as a numbered list.” After providing your feedback, you can use the “Test” button to see how the feedback alters the generated answer. Once you are satisfied with the changes, you can save your feedback.

515715e6-cfca-4a67-942a-4c0224e141e8-image

When you’re satisfied with how the feedback has improved the HALO response, you can save your feedback. This ensures that the feedback will be applied when a similar question is encountered by the HALO engine in the future, improving the answer to your liking.

Add Knowledge

Another reason you might be dissatisfied with a response could be due to missing knowledge in your Knowledge base. Previously, adding missing knowledge required modifying a PDF, website, or other document. Now, it’s much easier to add knowledge directly to the HALO engine via the Conversations Screen. When viewing a conversation, you will now find an “Add Knowledge” button next to the Like and Dislike buttons beneath a HALO response. Clicking this button allows you to enter new knowledge into a text field and save it to your Knowledge base, ensuring better responses in the future.

c61c9a9a-d9d1-4029-98f1-fe6181e46204-image

After adding your first piece of knowledge this way, you can view it in the updated 'Knowledge' tab. From this tab, you can also manually add additional pieces of knowledge.

Please note that manually added knowledge is treated the same as knowledge added through your website, a PDF, or another source. It is considered a piece of Knowledge, not Feedback. This means it may be used for any relevant question, not just those similar to the one where the knowledge was initially added.

View and edit your provided Feedback

In the Feedback tab, you can view all the feedback you’ve provided to the HALO engine. Here, you can see the type of feedback given, review the details, and use the three dots under the “Actions” column to edit or delete your feedback.

0ae14dc9-c262-41ea-9708-28f1cab4c41f-image

How to provide proper Feedback for “I don’t like the answer”

Here are several tips to ensure your feedback is useful for our HALO engine:

  • Provide feedback only if the answer is mostly factually correct but you’re not completely satisfied with it. but you’re not completely satisfied with it. If the answer is incorrect, it’s better to update your Knowledge base rather than trying to correct the mistake with Feedback.

  • Do not reference the customer question when giving Feedback. It's our job to match your feedback with new incoming questions, so you don’t need to worry about this. For example, feedback like “On similar questions, do not mention our email address” isn’t helpful. Instead, simply state: “Do not mention our email address.”

  • State how you want your answer to be changed, instead of rewriting the answer. For instance, if you want the last sentence removed from a HALO response, provide feedback like: “remove the last sentence” rather than rewriting the entire answer with the last sentence removed.

  • Avoid referring to your Knowledge base in the Feedback. Our Feedback Mechanism doesn’t “see” the Knowledge used when processing feedback. Its sole purpose is to improve the given answer by incorporating your feedback without having knowledge of the entire conversation context.

  • Use simple, imperative language in your feedback. Examples of good, usable feedback include: “Write HALO as Generative AI,” “Format as a numbered list,” “Do not mention our Email Channel,” “Use simpler language,” or “Add an emoji to the end of the sentence.”

19-08-2024 | HALO

Improvement | Conversational

We have implemented two updates that should make conversations with the HALO Engine feel more conversational:

  • Previously all (safe) interactions were processed in the RAG flow, where the best-matching knowledge was found and passed along to the LLM for answering the question. We have now altered this flow such that purely conversational messages (e.g. 'thanks for your help') no longer follow this flow but immediately receive a conversational comment; this should make the Engine feel more conversational, and decreases the response time for such a message.

  • Previously the first interaction of a conversation required a minimum of 3 words in order for a reply to be given. We have removed this restriction to provide a more conversational interaction (e.g. consumers opening the conversation with a simple 'Hello' will now get a conversational response).

06-08-2024 | HALO

Improvement | Trial Account

We have enhanced the trial experience by adding valuable information on the next steps you should take to go live with HALO.

Improvement | Resource Panel

We've revamped the Knowledge Page’s appearance. Additionally, you can now visualize PDF and JSON files directly within the app.

Improvement | Sync button

We’ve simplified the sync flow. Now, you only need to press one button to sync all resources or just some of them. Pressing the button again will stop the sync. Additionally, the sync badge is now more informative than ever.

18-07-2024 | HALO

Feature | Start the tour from any point

We listened to your feedback! You can now start the tour from any point you choose. Plus, you can finish the tour right away by pressing the "End tour now" button.

Improvement | Japanese Added to the Behavior Page

You can now set Japanese as the default language for the AI on the Behavior page. To make this feature work even better, ensure your knowledge base is also in Japanese.

Bug Fix | Test Center Enter Key Submit Issue

We fixed a bug that caused the Test Center to submit even while you were still typing a word, especially when using IME (Input Method Editor). This should greatly improve your experience.

11-07-2024 | HALO

Feature | Customizable AI behaviour for unanswered questions

Great update! You can now specify the AI's behavior when it encounters a question it can't answer, similar to how you set the Tone of Voice. This replaces the standard, repetitive response with a customized one. This option is available to you in the Behaviour page.

Improvement | Test Center redesigning

We've revamped the Test Center to enhance user experience and give it a fresh look. You'll notice HALO messages are now in purple, and you can view and navigate to the original sources from which information was extracted.

Bug fix | Authentication issue resolved

We've implemented a minor fix for situations where a user is authenticated but hasn't selected an account.

11-06-2024 | HALO

Improvement | Output Formatting

We are excited to announce several enhancements to our HALO engine, aimed at improving its functionality and user experience. The updates include:

  • Improved Hyperlink Formatting: Hyperlinks are now formatted using markdown code, making them clickable for easier access and navigation.

  • Enhanced Tone of Voice Recognition: The engine now better recognizes and adapts to various tones of voice. For instance, if a response is requested with emojis, the output will appropriately include emojis.

  • Optimized Email Formatting: Responses generated for the email channel will now have improved formatting, ensuring a more polished and professional appearance.

  • Better Adherence to Language: The engine now more accurately follows the language defined in the profile.

We are confident that these improvements will enhance your interaction with our HALO engine, making it more intuitive and effective.

28-05-2024 | HALO

Feature release | Enhanced Rewrite Mechanism

We made a new feature release for our HALO engine, namely an improved rewrite mechanism. This new system allows us to retrieve more relevant context from our knowledge database by utilizing the conversation history more effectively. This is relevant for all questions where previous interactions occurred; the mechanism updates the incoming question with relevant information from the conversation history so far.

Example of the Enhanced Rewrite Mechanism in action:

Conversation History:
Question: "Do you sell pens?"
Answer: "Yes, we sell pens!"

Follow-up Question:
"Do you sell blue ones?"

In this example, the follow-up question refers to the entity "pens" mentioned earlier. Our improved rewrite mechanism now rewrites the original question to: "Do you sell blue pens?". This rewrite allows us to retrieve more relevant information from the knowledge, ensuring the follow-up question is accurately understood in context, rather than potentially referring to something else. This improvement enables our HALO engine to deliver more accurate and relevant responses compared to the previous rewriting mechanism.

23-05-2024 | HALO

Feature release | Quality Assurance Dashboard

Are you curious about how your chatbot is performing across different products or profiles? Wonder no more! We're releasing the Quality Assurance Dashboard, which gives you an insight into the performance of HALO (Basic). For each configured profile, you'll get detailed metrics on conversation stats and recognition rates. You can also switch between different time periods to see how these metrics have changed.

Improvement | Better Guided Tour Experience

We added extra steps to the Guided Tour to enhance the overall experience for new users.

Improvement | UI Consistency

We improved the UI for better consistency and readability.

Bug fix | Test Centre Misleading Source Type

We fixed a bug where the source type in the Diagnostic section of the Test Centre was always displayed as "Other".

08-05-2024 | HALO

Bug fix | Knowledge Preview

The preview of uploaded knowledge didn't show any enters. This has been updated now such that the formatting of the text follows the formatting of the website/file more closely.

07-05-2024 | HALO

Bug fix | Web Scraper

The web scraper was not able to scrape web pages with unexpected elements in the headings. This has been fixed now.

25-04-2024 | HALO

Feature release | Interactive Guided Tour

They say a picture is worth a thousand words, but what about an interactive guided tour? We are introducing the first interactive guided tour, which will allow you to get started faster with HALO (Basic). In less than five minutes, the tour walks you through the most important features: Profile selection, Knowledge base, Behaviour settings, and finally, the Chatbot Test Centre. You can start the tour at any moment in time using the 'Take the tour' button in the left corner of the screen.

blob

18-04-2024 | HALO

Feature release | Search Engine

We have improved our search engine resulting in better search performance, so every incoming question should retrieve the most relevant knowledge articles.


16-04-2024 | HALO

Feature release | Additional Configuration Options

We've introduced three new fields within the HALO application across two screens:

  • Behaviour:

    • Company name: the name of the organisation that the AI will represent.

  • Profile Settings:

    • API key: required for onboarding HALO profiles with products like Conversational AI Cloud.

    • Product token: required for onboarding channel based integrations with CM.com's Conversational Router.


15-04-2024 | HALO

Feature release | Behaviour

Introducing a new screen to allow for the configuration of the AI's behaviour under different circumstances, including:

  • Tone of voice (moved over from profile settings).

  • Base language selection

  • Welcome message

  • No answer message

  • Handover behaviour (enable/disable, questions asked)


11-04-2024 | HALO

Improvement | Chunking strategy

Various revisions have been done to improve the chunking results from the uploaded knowledge (both for scraped websites and uploaded files):

  • It is now possible to read in tables in PDFs

  • Chunk splitting happens in more logical places.

  • Furthermore, unreadable chunks are dropped from the database, and chunks that are too similar to already-existing chunks are not saved as knowledge

As a result, the database will contain more concise and distinctive chunks, leading to retrieval of more relevant knowledge and thus more relevant answers by the HALO Engine.


10-04-2024 | HALO

Feature release | Conversation clustering tracking

Previously every time a synchronisation was performed on conversation history, all already-existing clusters were deleted and overwritten by new clusters. This has been adjusted such that previously-existing clusters remain available, and are automatically mapped to newly-created clusters (if they are similar enough).

Bug Fix | Conversation history storer

The prompts that are used to transform conversation history into summarized question-answer pairs was accidentally hardcoded to the Dutch language instead of the original language, causing non-Dutch conversations to be dropped from the database. This bug is now fixed.


21-03-2024 | HALO

Released the HALO application for the CM.com platform. The HALO app is the core application for customers to manage their AI within the CM.com platform. It provides the following high-level features for customers:

  • Self-service profile management:

    • A profile is an isolated AI that can be fed distinctive data, and will learn within its own domain separate from other profiles in any customers environment.

    • There is no limitation to the amount of profiles a customer can create.

  • Knowledge sources, with the initial release we support:

    • File uploads (JSON and PDF).

    • Web data through our web scraper. Single URLs, or wildcard URLS (urls ending in /*) that crawl multiple pages at once.

    • Conversation history, a mechanism that automatically generates new content based on conversations between live agents and end-users.

  • Audiences:

    • A way for customers to indicate the intended audience of a specific data source, e.g. a customer might only want a particular file to be used for answer suggestion for agents, but not for the conversational AI.

  • Synchronisations:

    • A synchronisation is a process that fetches the latest data from all connected sources, e.g. configured URL's for the web crawler, and recently uploaded/updated files.

    • A fully transparant process that shows what data it extracted from what sources.

    • Synchronisation can be configured to run on an automated schedule, e.g. every day or every Sunday.

    • An optional web crawler blacklists that tells the system to avoid specific URL's when crawling configured web pages.

    • The ability to synchronise individual sources to allow for rapid iterations/testing.

  • Test center and diagnostics:

    • The ability to talk directly to the AI to test and validate the AI.

    • Full diagnostics of AI behaviour, including the reasoning of the AI when formulating a response, and the sources used to formulate that particular response.

Channel integrations:

  • WhatsApp Business

  • Apple Messages for Business

  • Viber for Business

  • Facebook Messenger

  • Instagram Messaging

  • Telegram

  • Google's Business Messages

  • RCS Business Messaging

  • Microsoft Teams

  • Slack

All CCaaS integrations made available through marketplace, to name a few:

  • Salesforce Service Cloud

  • Dynamics Omnichannel

  • Quandago

JavaScript errors detected

Please note, these errors can depend on your browser setup.

If this problem persists, please contact our support.