We’re happy to release a new version of our beloved Beetle. As of today, our chatbots detect the users’ language automatically, and switches to it (if it is part of the supported idioms).

This means that if a user loads the chatbot in French, but actually asks a question in English, the conversation will continue smoothly without hiccups. A simple yet useful time-saver.

We’re excited to bring you the integration of ChatGPT into our chatbot platform.

Thanks to a simple action call, you can provide an answer to your customers that comes from the AI engine of OpenAI. This can be very useful when you own bot doesn’t have a foreseen answer on a new or unrecognised topic: you can have a fallback onto ChatGPT.

The integration is still currently in beta, whilst we gain experience as to how it behaves in all cases. Don’t hesitate to reach out to us, and we’ll be very happy to grant you access to this latest technological improvement.

Hello,

As the end of the year approaches and our product has grown, it seemed appropriate to take stock and launch our first Deeplink newsletter. Don’t panic, no spam, just an episodic email to keep in touch and inform you about the evolutions of our solution.

Enjoy your reading, and have a great holiday season!

Jérôme Berthier – CEO Deeplink

News

Bacon!

It’s well known 😉 that good stories start with bacon 🥓: this is the name of our new backend. Crisp and tasty (but above all robust, multi-channel and secure), this new base promises great innovations for 2023.

New front-end

This year we rewrote the chatbot. The modern interface is light to load and flexible. Above all, it has been designed primarily for smartphones, which account for more than 60% of conversations

Integration for Microsoft Teams

2022 saw the launch of our connector for MS Teams. This allows access to an existing bot from the communication tool, enhancing the bot’s services

After months of development, Deeplink has launched a new type of site with its AI Storyteller. Reversing the codes, our tool places the bot at the centre of the site, and generates an immersive experience thanks to “scrollytelling”

---

The year was also an opportunity to review the bot’s integration possibilities, to add an automatic content enrichment mechanism for the AI, to correct bugs and to launch the Live Chat function.

This week we’ve got things running smoother with our AI. We’ve coupled last release’s “Automated Data Enrichment” with an automatic training of the AI engine, and deployment to production. This means that every time your data gets better in one language, we not only enrich it synthetically in the other idioms, but we also make sure the AI follows and remains up-to-date. In parallel, we’ve updated our AI engines.

We’ve also made a few adjustments to the analytics dashboard so that is fits better on laptops, and added the chatbot’s ID to the settings page.

We’re very happy to bring you several big changes this week.

First, we’ve overhauled our mechanism for bot duplication. A copy of a bot now not only keeps the dialog in it’s complete form, it also contains all the intents and associated sentences for a ready-to-go AI engine.

Secondly, our chatbot’s layout on smartphones has been improved. You can now easily close the bot and return to the background website. Upon entering textual answers, iOS now presents a more natural and intuitive UI. And we’ve also fine-tuned the graphics of the slider components to reflect the bot’s graphical design settings.

We’re also super excited to launch a new feature: Automated Data Enrichment. What’s hiding behind that name? Well, we know how time-consuming it can be to collect data to train your AI engine, and regularly have to classify fresh sentences from users. And this is even worse, when you have several languages. We’ve come up with a new automatic processing to save you time: our system will add AI-generated synthetic data from other languages, which our advanced machine learning engine has detected as carrying high added value. This feature can be activated on your project’s Settings page.

Several optimisations were also carried out:

• we’ve release the 3rd version of our analytics dashboard. In a bid to help you better understand your users’ behaviour, we’ve now added some satisfaction measures, and allow to drill-down according to device and languages. We’ve also taken this opportunity to display more intent-related stats, and fixes a few minor bugs.
• the versioning page (and related actions) is now much faster to use.
• admin users can now edit/rename/modify intents
• we’ve changed the UI of the Settings page to be more user-friendly

We took a break for a bit, but now, we’re back in the saddle!

This week we fixed a few things. First of all, our spotlight feature has been improved: a new handling of layer-ordre reduces chances of overlapping content, and a bug preventing closure has been squashed. Then a bug related to the design of bots was ironed out: you may again choose to remove the banner image, without breaking your chatbot. We also tweaked the default design values, to improve the contrasts when reading. And the classification page now indicates the language in which a sentence was entered.

For the most curious of you, we’re super happy to announce that we’ve introduced a cool feature allowing to switch between bots. But we’ll cover this soon in a post on our blog, so stay tuned for exciting news.

These last weeks, we’ve been working on the AI engine, making it more powerful in all languages, including custom languages (i.e. dialects). Variables can no be left empty. You can use conditional logic to check is there is any value assigned: just compare the result with empty, by not putting anything. And we’ve ironed out some minor bugs. Intents are no longer corrupt upon bot duplication; bot URLs remain consistant upon reset in all parts, including referenced cards; API environment is now automatically passed to API calls.

Time for an update on our latest developments. We’ve been working on the platform’s stabilisation & optimisation, especially on the automatic translation tools. It’s an ongoing task. We’ve also got some novelty: we’re launching two new features today. The first allows you to open a Deeplink chatbot in the middle of a dialog. This is extremely useful when you want to switch from one bot to another, and dive right into the topic of interest. To do so, you must first define an anchor on the first message of a topic. This is done in our dialog editor, in the advanced parameters pane, where you set an “External Anchor”. Then you can call you bot by referencing that anchor in the URL. With a fullscreen bot, add ?state=yourAnchor to the URL. With a bot as widget, you’ll need a JS function to setState(). And option to reset the dialog exists as second parameter. An example is given at the bottom of this release not. The second new feature is destined at helping you onboard new customers: you can now spotlight an element on a page, whilst your chatbot explains it’s meaning & functioning. The spotlight darkens the whole page except the HTML element you want to highlight. You select an element, you must enter it’s CSS ID in the advanced pane of a message. This feature is currently in BETA release – your feedback is welcome.

Example: 
document.getElementById('chatbot').__vue_custom_element__.$children[0].setState(state, resetDialog)

We’ve been silent for a few weeks, hard at work on one of our platform’s core features. Today we have very happy to present you with a new version of our Data Collection Tool. In our push to make AI more accessible, we’ve rethought the way you create a campaign, the way people complete the polling survey, and how easily you classify the data. We’ve also taken this opportunity to rethink the navigation on the platform, choosing words that are more natural for all. We are also launching a new action, in “beta” mode: you may now interconnect two bots. Said otherwise: as a user interacts with a first bot, you can redirect him to a second bot and launch right into the middle of the dialog, at any point you have decided.

What’s New

In this release, we’ve addressed several small issues.

• Announced in the last release, the sending of card’s had been empowered, to allow the specification of one or several recipient(s), a subject and a sender name. Note that these fields can be dynamic, if you use variables. We’ve also fixed a few bugs, linked to variable substitution and images in emails.
• The Dialog Editor now warns of concurrent edition of the chatbot: no more risks of erasing a colleague’s work. We’ve also taken this opportunity to improve copy-past of objects.
• Finally, just for the fans of Star Trek we’ve introduced teleportation! Well, it’s only half of a joke: there is now a new action which allows you to switch from one bot to another, if you believe this other can better answer users’ questions. It’s a first step, and we will continue to work on this. But as the captain of the starship USS Enterprise would say: “Things are only impossible until they’re not“